Coder/deepseek-harness
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/master' into fix/workspace-instruction-frame-metadata

Browse Source 
# Conflicts:
#	packages/support/acp-snapshot/README.i18n.yaml
#	packages/support/acp-snapshot/README.md
#	packages/support/acp-snapshot/README.zh.md
This commit is contained in:
ZiyaZhang
2026-07-28 08:59:32 -07:00
parent f688cd32af eb38a365dc
commit 15edb38a34
122 changed files with 3370 additions and 2451 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
.agents/notes/implemented/architecture/2026-06-14-session-persistence.i18n.yaml
View File

@@ -1,6 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write
2026-06-14-session-persistence.md: 75e13b860f621ed407849b3b4c62ff7287ab4812
2026-06-14-session-persistence.zh.md: a6bd400a053779c742940236737447d1687622de
# pnpm run verify-translation-pairing --write .agents/notes/implemented/architecture/2026-06-14-session-persistence.md
2026-06-14-session-persistence.md: 137b2b01126214629952812f3dd3b71985a3acda
2026-06-14-session-persistence.zh.md: 2846ee92349c297fb3a173ba9dd3e2ff3cd9ee1a

2
.agents/notes/implemented/architecture/2026-06-14-session-persistence.md
View File

@@ -29,7 +29,7 @@ Key choices recorded here because they are durable, contested, and surprising:
Each key choice above records its rejected alternative where the choice is stated: a **chunk-filtered canonical log** (Codex's `policy.rs` shape) — breaks the contiguous-seq contract; **truncating a crashed turn** — silently destroys a long autonomous run's real work; an **in-log `session/meta` event as line 0** — metadata is not replayable state; **finite fractional `createdAt` values** — have no producer and diverge from integer Unix-millisecond storage and query columns; **adopting a non-pristine unversioned SQLite file** — can overwrite unrelated objects or identity; **hard-injecting `sessionPersistence` into the loop** — would pend non-persistent demos forever.
Format versioning: the header carries a `version`; `load` rejects any non-current version (no migration — the pre-release session format is pinned at `SESSION_FORMAT_VERSION = 0` and absorbs shape churn, per the AGENTS.md pre-release stance). Stated honestly: append-only + flush is robust to partial trailing writes (tolerated on load) but not to fsync-less power loss mid-line; a DB/WAL backend is the stronger option later.
Format versioning: the header carries a `version`; `load` rejects any non-current version. The pre-release session format stays pinned at `SESSION_FORMAT_VERSION = 0` and carries no broad compatibility promise, while the coordinator may own an explicit narrow import upgrade when persisted user data requires it ([pre-identity message recovery](../bug-fix/2026-07-28-load-pre-identity-session-messages.md)). Stated honestly: append-only + flush is robust to partial trailing writes (tolerated on load) but not to fsync-less power loss mid-line; a DB/WAL backend is the stronger option later.
## Consequences

2
.agents/notes/implemented/architecture/2026-06-14-session-persistence.zh.md
View File

@@ -29,7 +29,7 @@ Status: implemented
上述每个关键选择都在陈述处记录了被否决的替代方案：**过滤分片的规范日志**（Codex 的 `policy.rs` 形式）破坏连续 seq 契约；**截断崩溃的轮次**会静默销毁长时间自主运行中的真实工作；**日志内 `session/meta` 事件作为第 0 行**——元数据不是可回放状态；**有限的非整数 `createdAt` 值**没有生产方，且与整数 Unix 毫秒存储及查询列不一致；**接受非全新的未版本化 SQLite 文件**可能覆盖无关对象或应用标识；**将 `sessionPersistence` 硬注入循环**会让非持久化的演示永远挂起。
格式版本控制：header 携带一个 `version`；`load` 拒绝任何非当前版本（不做迁移——预发布阶段的会话格式固定为 `SESSION_FORMAT_VERSION = 0` 并吸收形状变动，遵循 AGENTS.md 的预发布立场）。坦率地说：仅追加 + 刷写对部分尾部写入是健壮的（加载时容忍），但对行写入中途的无 fsync 断电不健壮；数据库/WAL 后端是后续更强的选项。
格式版本控制：header 携带一个 `version`；`load` 拒绝任何非当前版本。预发布阶段的会话格式仍固定为 `SESSION_FORMAT_VERSION = 0`，不承诺广泛兼容；当持久化用户数据确有需要时，协调器可以负责显式且范围受限的导入升级（[消息标识机制引入前的消息恢复](../bug-fix/2026-07-28-load-pre-identity-session-messages.md)）。坦率地说：仅追加 + 刷写对部分尾部写入是健壮的（加载时容忍），但对行写入中途的无 fsync 断电不健壮；数据库/WAL 后端是后续更强的选项。
## 后果

6
.agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.i18n.yaml Normal file
View File

@@ -0,0 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.md
2026-07-28-load-pre-identity-session-messages.md: 2901527658421b37576bdf5b49e66829104a3b41
2026-07-28-load-pre-identity-session-messages.zh.md: 61d57ac9f3318299b63faa659b6d155e8e89fae3

38
.agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.md Normal file
View File

@@ -0,0 +1,38 @@
# Agent Note: Load sessions persisted before message identity
Status: implemented
English | [中文](2026-07-28-load-pre-identity-session-messages.zh.md)
## Problem
The identified immutable message change replaced four durable event payloads with complete message values. Existing v0 JSONL and SQLite sessions still held the immediately preceding shapes: direct `content`/`source` on user and steering events, `content`/`provenance` on assistant events, and `callId`/`content`/`isError` on tool results. Their headers still matched `SESSION_FORMAT_VERSION`, but current-shape validation rejected them before resume could construct a live `Session`.
Changing the message representation without a version bump made those logs indistinguishable at the header level from current v0 logs. The runtime needs a narrow import rule that restores data created by the supported first-party backends without weakening validation for unrelated obsolete or malformed events.
## Decision
`PersistenceCoordinator` normalizes the four exact pre-identity message payloads after backend decoding and before current message validation. It wraps their existing semantic fields in the current role-specific message shape and assigns `legacy-message:<session-id>:<event-seq>` as the deterministic imported `MessageId`. A legacy `tool/result` content replacement inherits the imported id of its replacement target, preserving the current content-only rewrite invariant.
The same normalization runs for `load`, `inspect`, an ownerless loaded state claiming its live session, and HMR prefix adoption. Prefix comparisons therefore compare the live current-shape seed with the same normalized stored view. Current-looking wrappers with missing or invalid fields are not repaired, and unsupported event vocabulary, request headers, versions, and surface relations retain their existing rejection paths.
The upgrade is read-only. Stored legacy records remain unchanged; a resumed session appends only current-shape events after them. Deterministic identities make repeated loads and a mixed legacy/current log reproduce the same message ids without a backend-specific rewrite transaction.
## Alternatives considered
**Reject the logs under the pre-release compatibility stance.** This is the default for unrelated v0 churn, but it strands real first-party sessions even though every old field maps unambiguously to the current message representation.
**Rewrite the complete stored log in place.** This would canonicalize the artifact but violate the append-only storage contract, require separate atomic replacement mechanisms for JSONL and SQLite, and expand a read compatibility fix into a migration system.
**Mint random ids on each load.** The messages would satisfy the type shape but lose stable identity across inspect, resume, restart, and mixed legacy/current appends.
## Consequences
Pre-identity JSONL and SQLite sessions resume with their original message content, sources, provider provenance, tool correlation, errors, metadata, and surface replacements. The returned events are otherwise indistinguishable from current imported message snapshots and remain deeply frozen.
This is one explicit same-version import exception, not a general v0 compatibility layer. Adding another exception requires another complete, unambiguous mapping at the persistence boundary; malformed current data continues to fail rather than being guessed into validity. The shared coordinator contract exercises the upgrade against the in-memory reference, JSONL, and SQLite backends, including deterministic reload and tool-result replacement identity.
## Related
- [Create every message as an identified immutable value](../architecture/2026-07-28-identified-immutable-message-values.md) — owns the current message identity and immutability contract.
- [Session persistence as an abstract service](../architecture/2026-06-14-session-persistence.md) — owns the append-only backend and resume boundary.

38
.agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.zh.md Normal file
View File

@@ -0,0 +1,38 @@
# Agent Note: 加载消息标识机制引入前持久化的会话
Status: implemented
[English](2026-07-28-load-pre-identity-session-messages.md) | 中文
## 问题
带标识的不可变消息变更将四种持久事件载荷替换为完整消息值。现有的 v0 JSONL 和 SQLite 会话仍保留紧邻该变更之前的形状：用户事件和 steering（中途引导）事件直接携带 `content`/`source`，assistant 事件携带 `content`/`provenance`，工具结果则携带 `callId`/`content`/`isError`。这些会话的 header 仍与 `SESSION_FORMAT_VERSION` 匹配，但当前形状验证会拒绝它们，导致恢复流程无法构造实时 `Session`。
消息表示改变时没有提升版本，导致这些日志无法仅凭 header 与当前的 v0 日志区分。运行时需要一条范围受限的导入规则，既能恢复受支持的第一方后端所创建的数据，又不削弱对无关过时事件或格式错误事件的验证。
## 决策
`PersistenceCoordinator` 会在后端解码之后、当前消息验证之前，规范化消息标识机制引入前的四种特定消息载荷。它将载荷现有的语义字段包装进当前按角色区分的消息形状，并为其分配确定性的导入 `MessageId`：`legacy-message:<session-id>:<event-seq>`。旧版 `tool/result` 的内容替换会继承替换目标导入后的 id，从而保持当前仅改写内容的不变量。
同一项规范化也用于 `load`、`inspect`、无 owner 的已加载状态认领其实时会话，以及 HMR（热模块替换）前缀接管。因此，前缀比较会将实时的当前形状 seed 与同一份规范化存储视图进行比较。看似当前形状、但字段缺失或无效的包装层不会被修复；不受支持的事件词汇、请求 header、版本和 surface 关系仍沿用现有拒绝路径。
这项升级只发生在读取时。存储中的旧版记录保持不变；会话恢复后，只会在其后追加当前形状的事件。确定性标识使重复加载以及新旧形状混合的日志无需执行后端专用的重写事务，也能复现相同的消息 id。
## 考虑过的替代方案
**按照预发布兼容性立场拒绝这些日志。** 这是处理其他 v0 形状变动的默认方式，但即使每个旧字段都能明确映射到当前消息表示，它仍会导致真实的第一方会话无法恢复。
**就地重写完整的存储日志。** 这会使产物规范化，但违反仅追加存储契约，还需要为 JSONL 和 SQLite 分别实现原子替换机制，并将一次读取兼容性修复扩大为迁移系统。
**每次加载时随机生成 id。** 这些消息会满足类型形状，却无法在检查、恢复、重启以及新旧形状混合追加之间保持稳定标识。
## 后果
消息标识机制引入前的 JSONL 和 SQLite 会话可以恢复，并保留原始的消息内容、来源、提供方溯源信息、工具关联、错误、元数据和 surface 替换。除此之外，返回事件与当前导入的消息快照无法区分，并且仍然经过深度冻结。
这是一个显式的同版本导入例外，而非通用的 v0 兼容层。若要增加另一个例外，必须在持久化边界提供另一套完整且无歧义的映射；当前数据若格式错误，系统仍会拒绝，而不会猜测如何将其变成有效数据。共享协调器契约会针对内存参考实现、JSONL 和 SQLite 后端验证这项升级，包括重新加载时的确定性，以及工具结果替换时的标识继承。
## 相关
- [将每条消息创建为带标识的不可变值](../architecture/2026-07-28-identified-immutable-message-values.md)：该记录负责当前的消息标识与不可变性契约。
- [会话持久化作为抽象服务](../architecture/2026-06-14-session-persistence.md)：该记录负责仅追加后端与恢复边界。

6
.agents/notes/implemented/feature/2026-06-29-todo-write-tool.i18n.yaml
View File

@@ -1,6 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write
2026-06-29-todo-write-tool.md: 760373d64f462e3717a174d5793e6d47ab76b0b4
2026-06-29-todo-write-tool.zh.md: fd29e6049e6a729c2c7e78860ad7c065422da60e
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-06-29-todo-write-tool.md
2026-06-29-todo-write-tool.md: 288932f641a37c13ea6beeb069ac360c4a8447c1
2026-06-29-todo-write-tool.zh.md: 03dae6328e5c7f4379694ab01db3434b4469826c

2
.agents/notes/implemented/feature/2026-06-29-todo-write-tool.md
View File

@@ -18,7 +18,7 @@ The model sends the entire list every call; the new list replaces the old (last-
### State on the session log, not a service
The list is appended as a `todo/write` event carrying the full `{ todos }` snapshot. The harness is event-sourced — the LLM history, tool calls, and turn structure all live on the log — so the todo list lives there too. This buys durability, replay, and resume reconstruction for free: a reopened session re-derives the current list from the latest `todo/write`, with no separate persistence backend, in-memory service to rehydrate, or extra wiring. An in-memory `ctx.todos` service would have to reinvent all of that. (Full-log consumers get this reconstruction outright; the web client's paged window gets it from the tail history page's host-computed projection — see the [web todo display note](2026-07-23-web-todo-display.md).)
The list is appended as a `todo/write` event carrying the full `{ todos }` snapshot. The harness is event-sourced — the LLM history, tool calls, and turn structure all live on the log — so the todo list lives there too. This buys durability, replay, and resume reconstruction for free: a reopened session re-derives the standing plan from the latest `todo/write` that is not followed by a later `turn/start` ([plan strip lifetime](2026-07-28-todo-plan-clears-on-next-turn.md)), with no separate persistence backend, in-memory service to rehydrate, or extra wiring. An in-memory `ctx.todos` service would have to reinvent all of that. (Full-log consumers get this reconstruction outright; the web client's paged window gets it from the tail history page's host-computed projection — see the [web todo display note](2026-07-23-web-todo-display.md).)
### NOT a surface event

2
.agents/notes/implemented/feature/2026-06-29-todo-write-tool.zh.md
View File

@@ -18,7 +18,7 @@ harness 为模型提供了 bash 和 subagent 工具，却没有办法记录结
### 状态在会话日志上，而非服务
列表作为 `todo/write` 事件追加到日志，携带完整的 `{ todos }` 快照。harness 是事件溯源的——LLM（大语言模型）历史、工具调用和轮次结构都在日志上——所以 todo 列表也在那里。这免费获得了持久性、回放和恢复重建：重新打开的会话从最新的 `todo/write` 重新推导当前列表，无需独立的持久化后端、无需重新注水的内存服务、无需额外接线。一个内存中的 `ctx.todos` 服务需要重新发明以上所有。（全量 log 消费者直接获得这份重建；web 客户端的分页窗口则从尾页 history 携带的 host 计算投影获得——见 [web todo 展示 Note](2026-07-23-web-todo-display.md)。）
列表作为 `todo/write` 事件追加到日志，携带完整的 `{ todos }` 快照。harness 是事件溯源的——LLM（大语言模型）历史、工具调用和轮次结构都在日志上——所以 todo 列表也在那里。这免费获得了持久性、回放和恢复重建：重新打开的会话从「其后没有更晚 `turn/start`」的最近一次 `todo/write` 重新推导站立计划（[计划条生命周期](2026-07-28-todo-plan-clears-on-next-turn.md)），无需独立的持久化后端、无需重新注水的内存服务、无需额外接线。一个内存中的 `ctx.todos` 服务需要重新发明以上所有。（全量 log 消费者直接获得这份重建；web 客户端的分页窗口则从尾页 history 携带的 host 计算投影获得——见 [web todo 展示 Note](2026-07-23-web-todo-display.md)。）
### 不是 surface 事件

4
.agents/notes/implemented/feature/2026-07-23-web-todo-display.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-23-web-todo-display.md
2026-07-23-web-todo-display.md: 5fe08cc40c1d23ff3a9b8c6d766fea6d3694c30d
2026-07-23-web-todo-display.zh.md: c121ffc27e3d0a93707c2c22b2f180023ebae5be
2026-07-23-web-todo-display.md: 7223ff9adbf1fa6dca39c9eb4949b6d3861bdd6d
2026-07-23-web-todo-display.zh.md: ac773a3c587183aa9a152caff2812686bc79f3c7

4
.agents/notes/implemented/feature/2026-07-23-web-todo-display.md
View File

@@ -14,7 +14,7 @@ Consume `todo/write` as a Session side effect, not a surface node, and render it
### Side-effect channel, converging with window replay
`applyEventSideEffects` gains a `todo/write` case (whole list, last write wins). Unlike partial/openCalls, `rebuildDerivedFromWindow` deliberately does NOT reset it: the value is session-level — taken from the tail history page's full-log projection — and an arbitrary window may not contain the latest write, so an older-page prepend keeps it and only an in-window or live write overwrites it. Every `installWindow` caller is a tail request (`doOpen`, its gap re-pull, `repairGap`; `loadOlder` prepends without it), which the host answers with the projection or omits it only when the full log holds no `todo/write` — so an absent field is the authoritative empty list and is assigned as such. That distinction matters on rollback: a live write whose host crashed before persisting leaves the log empty, and preserving the prior value instead would strand the rolled-back plan on screen indefinitely. `ConversationSnapshot.todos` is the read surface. This follows the event's own contract ("log-only UI state; never derived history"): surfacing each write as a conversation node would render superseded lists as if they were still standing.
`applyEventSideEffects` gains a `todo/write` case (whole list, last write wins) and clears on `turn/start` ([turn-scoped plan lifetime](2026-07-28-todo-plan-clears-on-next-turn.md)). `rebuildDerivedFromWindow` sweeps the window from an empty plan and restores the tail-page seed only when the window never determined the plan (no `todo/write` and no `turn/start`); otherwise the in-window write/`turn/start` fold wins. Every `installWindow` caller is a tail request (`doOpen`, its gap re-pull, `repairGap`; `loadOlder` prepends without reseeding), which the host answers with the projection or omits it when no plan stands — so an absent field is the authoritative empty list and is assigned as such. That distinction matters on rollback: a live write whose host crashed before persisting leaves the log empty, and preserving the prior value instead would strand the rolled-back plan on screen indefinitely. `ConversationSnapshot.todos` is the read surface. This follows the event's own contract ("log-only UI state; never derived history"): surfacing each write as a conversation node would render superseded lists as if they were still standing.
### TodoPanel: the durable list as a persistent strip
@@ -33,4 +33,4 @@ The dedicated `todo_write` chat row is a plain registrant plugin (`todoToolview`
## Consequences
Replay correctness is owned by one code path: any future change to window rebuild keeps todos consistent for free, and the fixture (fx-alpha turn 65) plus the assembled keyless snapshot (`apps/web/tests/todo-display.snapshot.ts`) pin the full chain (row summary and state, dock panel content, collapse round-trip) over the built client graph. `todos` is a required `ConversationSnapshot` field, so scripted fakes in specs must carry it. The TUI panel is untouched (the automation-only ACP bridge deliberately omits todo presentation); the web surfaces render the same event, adding one wire field and no new event type. That field is how cold-load reconstruction stays host-backed: the tail history page carries `todos` — the full-log latest `todo/write`, computed independently of the page window (the same backscan posture the view pairing uses) — so a reopened session restores the plan even when the last write precedes the window; that value survives an older-page prepend, is overwritten by any later write, and resets to empty when a tail response carries no projection.
Replay correctness is owned by one code path: any future change to window rebuild keeps todos consistent for free, and the fixture (fx-alpha turn 65) plus the assembled keyless snapshot (`apps/web/tests/todo-display.snapshot.ts`) pin the full chain (row summary and state, dock panel content, collapse round-trip) over the built client graph. `todos` is a required `ConversationSnapshot` field, so scripted fakes in specs must carry it. The TUI panel shares the same turn-scoped lifetime (the automation-only ACP bridge deliberately omits todo presentation); the web surfaces render the same event, adding one wire field and no new event type. That field is how cold-load reconstruction stays host-backed: the tail history page carries `todos` — the full-log standing plan (latest `todo/write` with no later `turn/start`), computed independently of the page window (the same backscan posture the view pairing uses) — so a reopened session restores the plan when it still stands and the last write precedes the window; that value survives an older-page prepend, is overwritten by any later write, clears on a later `turn/start`, and resets to empty when a tail response carries no projection.

4
.agents/notes/implemented/feature/2026-07-23-web-todo-display.zh.md
View File

@@ -14,7 +14,7 @@ Status: implemented
### 副作用通道，与窗口回放收敛
`applyEventSideEffects` 新增一个 `todo/write` 分支（整份列表，后写覆盖先写）。与 partial/openCalls 不同，`rebuildDerivedFromWindow` 刻意不重置它：该值是会话级的——取自尾页 history 携带的全量 log 投影——而任意窗口未必包含最近一次写入，因此往前翻页保留它，只有窗口内或实时的写入才会覆盖。`installWindow` 的每个调用方都是尾页请求（`doOpen`、其补洞重拉、`repairGap`；`loadOlder` 只往前拼接、不走它），而 host 对尾页请求要么带上投影、要么仅在全量 log 没有任何 `todo/write` 时省略——因此字段缺失就是权威的空列表，直接照此赋值。这个区分在回滚场景上要紧：实时写入若在 host 持久化前崩溃，log 里就是空的，此时保留旧值会让已回滚的计划永远留在屏幕上。`ConversationSnapshot.todos` 是读取面。这遵循事件自身的契约（「仅日志 UI 状态，绝非派生历史」）：把每次写入作为对话节点呈现，会让已被取代的列表看起来仍然有效。
`applyEventSideEffects` 新增一个 `todo/write` 分支（整份列表，后写覆盖先写），并在 `turn/start` 清空（[按 turn 界定的计划生命周期](2026-07-28-todo-plan-clears-on-next-turn.md)）。`rebuildDerivedFromWindow` 从空计划扫过窗口，仅当窗口从未判定计划（无 `todo/write` 且无 `turn/start`）时恢复尾页种子；否则以窗口内写入／`turn/start` 折叠为准。`installWindow` 的每个调用方都是尾页请求（`doOpen`、其补洞重拉、`repairGap`；`loadOlder` 只往前拼接、不再播种），而 host 对尾页请求要么带上投影、要么在无站立计划时省略——因此字段缺失就是权威的空列表，直接照此赋值。这个区分在回滚场景上要紧：实时写入若在 host 持久化前崩溃，log 里就是空的，此时保留旧值会让已回滚的计划永远留在屏幕上。`ConversationSnapshot.todos` 是读取面。这遵循事件自身的契约（「仅日志 UI 状态，绝非派生历史」）：把每次写入作为对话节点呈现，会让已被取代的列表看起来仍然有效。
### TodoPanel：长驻列表作为一条常驻横条
@@ -33,4 +33,4 @@ Status: implemented
## Consequences
回放正确性由一条代码路径掌管：未来对窗口重建的任何改动都免费保持 todos 一致；fx-alpha 第 65 轮的 fixture（测试前置数据）加 assembled keyless snapshot（`apps/web/tests/todo-display.snapshot.ts`）在构建产物客户端全图上钉住整条链（行摘要与状态、dock 面板内容、折叠往返）。`todos` 是 `ConversationSnapshot` 的必填字段，所以 spec 里脚本化的 fake 必须带上它。TUI 面板未受改动（自动化专用的 ACP 桥接刻意不做 todo 呈现）；Web 各面渲染同一个事件，只新增一个协议字段，不新增事件类型。冷加载重建正是靠这个字段由 host 兜底：history 尾页附带 `todos`——全量 log 上最新一次 `todo/write` 的投影，独立于分页窗口计算（与 view 配对同一种 backscan 姿势）——因此重开会话时即使最后一次写入落在窗口之前，计划也照常恢复；该值跨往前翻页保留，之后的任何写入照常覆盖，而尾页响应不带投影时复位为空。
回放正确性由一条代码路径掌管：未来对窗口重建的任何改动都免费保持 todos 一致；fx-alpha 第 65 轮的 fixture（测试前置数据）加 assembled keyless snapshot（`apps/web/tests/todo-display.snapshot.ts`）在构建产物客户端全图上钉住整条链（行摘要与状态、dock 面板内容、折叠往返）。`todos` 是 `ConversationSnapshot` 的必填字段，所以 spec 里脚本化的 fake 必须带上它。TUI 面板共用同一按 turn 界定的生命周期（自动化专用的 ACP 桥接刻意不做 todo 呈现）；Web 各面渲染同一个事件，只新增一个协议字段，不新增事件类型。冷加载重建正是靠这个字段由 host 兜底：history 尾页附带 `todos`——全量 log 上的站立计划（其后没有更晚 `turn/start` 的最近一次 `todo/write`），独立于分页窗口计算（与 view 配对同一种 backscan 姿势）——因此重开会话时若计划仍站立且最后一次写入落在窗口之前，计划也照常恢复；该值跨往前翻页保留，之后的任何写入照常覆盖，更晚的 `turn/start` 会清空，而尾页响应不带投影时复位为空。

6
.agents/notes/implemented/feature/2026-07-28-todo-plan-clears-on-next-turn.i18n.yaml Normal file
View File

@@ -0,0 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-28-todo-plan-clears-on-next-turn.md
2026-07-28-todo-plan-clears-on-next-turn.md: 8a2808f5023a0800fbbeb65c5e28b3495e051dbc
2026-07-28-todo-plan-clears-on-next-turn.zh.md: 6ea6b2b4fafe4b2f695ad2d54561f164747e5694

31
.agents/notes/implemented/feature/2026-07-28-todo-plan-clears-on-next-turn.md Normal file
View File

@@ -0,0 +1,31 @@
# Agent Note: Todo plan strip clears on the next turn
Status: implemented
English | [中文](2026-07-28-todo-plan-clears-on-next-turn.zh.md)
## Problem
`todo_write` stores whole-list snapshots on the session log, and interactive hosts render the latest list as a plan strip (web TodoPanel via the `todos` projection, TUI Plan panel). After a turn finished, that strip stayed on screen into the next user turn — a completed or abandoned checklist from the previous task. Readers treat the strip as "what this turn is doing," so a stale list across the turn boundary is the wrong product lifetime. The [web todo display](2026-07-23-web-todo-display.md) and [`todo_write` tool](2026-06-29-todo-write-tool.md) notes still own event-sourcing and the two render surfaces; they described the standing plan as lasting for the whole session until the next write.
## Decision
The standing plan is the latest `todo/write` that is not followed by a later `turn/start`. `turn/end` keeps the list visible so the finished checklist remains while the user reads the answer; the next `turn/start` clears it until the model writes again.
### Host projection (web)
`dsh-tool-todo`'s `todos` projection unit folds the rule: `apply` takes the whole list from each `todo/write` and returns `null` on each `turn/start` (`stateVersion` 2). Carriers (`dsh-host-apiproxy`) serve that value on the history tail `projections` block and push `session/projection` frames; the web dock reads it through `useProjection('todos')`. The keyless fixture mirrors the same fold for assembled snapshots.
### TUI live path
The TUI `renderEvent` switch still clears its local plan panel on `turn/start` and replaces it on `todo/write` (TUI is not yet a projection carrier). The rebuild path resets the panel before replaying the log so cold resume converges on the same rule.
## Alternatives considered
- **Clear on `turn/end`** — hides the checklist while the user is still reading the just-finished answer; the strip's job at that moment is the completed plan, not an empty dock.
- **Clear only when every item is `completed`** — leaves abandoned or partial plans across turns; the strip would still show another task's work.
- **Append an empty `todo/write` on turn start** — mutates the log for a UI lifetime rule and invents a write the model never authored.
## Consequences
The host projection and the TUI panel share one lifetime rule; reopening a session restores a plan only when no later turn has started. Partial supersession of the session-long standing-plan wording in [web todo display](2026-07-23-web-todo-display.md) and [`todo_write` tool](2026-06-29-todo-write-tool.md): event-sourcing, last-write-wins replacement, and the two render surfaces stay there; this note owns turn-boundary clearance. Coverage: tool-todo projection specs for turn/start clear + turn/end keep, fixture push-frame clearance for the assembled web snapshot, plus the TUI snapshot that starts the next turn and pins the strip gone.

31
.agents/notes/implemented/feature/2026-07-28-todo-plan-clears-on-next-turn.zh.md Normal file
View File

@@ -0,0 +1,31 @@
# Agent Note: 下一轮开始时清空 todo 计划条
Status: implemented
[English](2026-07-28-todo-plan-clears-on-next-turn.md) | 中文
## 问题
`todo_write` 在会话日志中存储整表快照，交互式宿主把最新列表渲染为计划条（web TodoPanel 经 `todos` 投影，TUI Plan 面板）。一轮结束后，该条仍留在下一用户轮次的屏幕上——上一任务已完成或已放弃的清单。读者把计划条理解为「本轮正在做什么」，因此跨轮次的陈旧列表是错误的产品生命周期。[web todo 展示](2026-07-23-web-todo-display.md)与 [`todo_write` 工具](2026-06-29-todo-write-tool.md)笔记仍拥有事件溯源与两个渲染面；它们把站立计划描述为持续整段会话直至下一次写入。
## 决策
站立计划是其后没有更晚 `turn/start` 的最近一次 `todo/write`。`turn/end` 保留列表可见，以便用户阅读回答时仍能看到刚完成的清单；下一次 `turn/start` 将其清空，直至模型再次写入。
### Host 投影（web）
`dsh-tool-todo` 的 `todos` 投影单元折叠该规则：`apply` 从每个 `todo/write` 取整表，并在每个 `turn/start` 返回 `null`（`stateVersion` 2）。载体（`dsh-host-apiproxy`）在历史尾页的 `projections` 块与 `session/projection` 推送帧上供给该值；web dock 经 `useProjection('todos')` 读取。无密钥 fixture（测试前置数据）镜像同一折叠，供组装后的 snapshot 使用。
### TUI 实时路径
TUI 的 `renderEvent` 分支仍在 `turn/start` 清空本地计划面板、在 `todo/write` 替换之（TUI 尚非投影载体）。重建路径在回放日志前重置面板，使冷恢复收敛到同一规则。
## 考虑过的替代方案
- **在 `turn/end` 清空**——用户仍在阅读刚完成的回答时就隐藏清单；此时计划条的职责是已完成计划，而非空 dock。
- **仅在全部项为 `completed` 时清空**——会让放弃或部分完成的计划跨轮残留；计划条仍会显示另一任务的工作。
- **在 turn start 追加空的 `todo/write`**——为 UI 生命周期规则改写日志，并捏造模型从未写出的写入。
## 后果
Host 投影与 TUI 面板共用同一生命周期规则；重新打开会话仅在其后没有更晚轮次开始时恢复计划。部分取代 [web todo 展示](2026-07-23-web-todo-display.md)与 [`todo_write` 工具](2026-06-29-todo-write-tool.md)中「会话级站立计划」的表述：事件溯源、last-write-wins 替换与两个渲染面仍归那些笔记；本笔记拥有轮次边界清空。覆盖：tool-todo 投影对 turn/start 清空与 turn/end 保留的规格测试、供组装 web snapshot 的 fixture 推送帧清空，以及启动下一轮并钉住计划条消失的 TUI snapshot。

6
.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.i18n.yaml
View File

@@ -1,6 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write
2026-06-19-acp-snapshot-tests.md: 430441e633af1e487f19272900360a8ed2f595c9
2026-06-19-acp-snapshot-tests.zh.md: 243e431567b45d69e1fe16dda5d07ec058403b7c
# pnpm run verify-translation-pairing --write .agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md
2026-06-19-acp-snapshot-tests.md: ba46682b3087e3d2ff4c52d2ad22f54b7dac31db
2026-06-19-acp-snapshot-tests.zh.md: 58195c47889edb31d5122116328928f916ad09a9

5
.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md
View File

@@ -53,7 +53,7 @@ Replay uses a `cordis.snapshot.yml` overlay that replaces the real adapter with
A snapshot run asserts **two** normalized surfaces, because the harness's external surfaces are distinct:
1. The **stdout transcript** — the framed ACP JSON-RPC responses and committed-message updates an automation client receives. It catches regressions in the transport contract and is compared against a committed `stdout.expected.jsonl`.
2. The **re-persisted session JSONL**, normalized and compared with `session.jsonl`. The same fixture is both replay source and expected log. Prompt text is scrubbed; one scenario per header class pins readable prompt and tool content as described in the [header-pinning Agent Note](../../archived/testing/2026-07-06-pin-request-header-content-in-one-scenario.md). Override scenarios derive model behavior solely from their sidecar.
2. The **re-persisted session JSONL**, normalized and compared with `session.jsonl`. The same fixture is both replay source and expected log. Prompt and tool bulk are scrubbed; one scenario per header class pins the remaining header sequence. The pin owns readable prompt and tool-schema sidecars by default, or names another pin as either source when the complete sequence is identical, so each distinct sidecar version is committed once. Fixture guards reject duplicate sidecar content, and record/refresh rejects shared claimants that generate different bytes. The original header-pinning rationale is preserved in the [header-pinning Agent Note](../../archived/testing/2026-07-06-pin-request-header-content-in-one-scenario.md). Override scenarios derive model behavior solely from their sidecar.
The surfaces are complementary: stdout covers the minimal automation wire, while JSONL covers loop, tool, and boundary structure that the wire intentionally omits.
@@ -76,9 +76,10 @@ Tool determinism comes from a generated cwd, scrubbed environment, fresh non-log
- **A hand-authored `llm.json` of model chunks** — the earlier draft; reusing the real session log makes the fixture a genuine product of the system rather than a hand-built mock, and doubles it as a behavioral expected output.
- **A byte-level HTTP-record library (Polly/nock/MSW)** — rejected: adapter-specific, awkward with streaming SSE, and lower-level than the thing under test.
- **Synthesizing throw/cancel entries from `turn/end {kind:'error'|'aborted'}`** — rejected: it couples `llm-replay` to loop-internal turn-closing semantics, and the `turn/end` reason is lossy (it cannot distinguish a thrown 401 from a finish-error); the explicit `replay.override.json` sidecar is the cleaner seam.
- **Copying both request-header sidecars beside every class pin** — rejected: prompt and tool-schema composition vary independently, so a change to one shared component would churn byte-identical files across unrelated class pins. Explicit per-component sources retain one structural pin per class without duplicating content.
## Consequences
The tier adds reviewed per-scenario input, session, stdout, optional override, and optional workspace fixtures. Workspace seeds are copied into the generated cwd for both record and replay. In return the tier provides deterministic keyless coverage through the real Loader and tool composition. Most retained scenarios exercise the assembled backend rather than ACP; the [automation-only ACP decision](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary) keeps that corpus here and defers any move to a transport-neutral headless suite as an independent testing change (the suite-level FIXME marks it).
The tier adds reviewed per-scenario input, session, stdout, optional override, and optional workspace fixtures, plus one file for each distinct pinned prompt and tool-schema sequence. Workspace seeds are copied into the generated cwd for both record and replay. In return the tier provides deterministic keyless coverage through the real Loader and tool composition. Most retained scenarios exercise the assembled backend rather than ACP; the [automation-only ACP decision](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary) keeps that corpus here and defers any move to a transport-neutral headless suite as an independent testing change (the suite-level FIXME marks it).
This Agent Note relates to but does not supersede the [proposed determinism Agent Note](../../proposed/testing/2026-06-11-deterministic-and-stress-testing.md): that proposal's "universal replay fixture" re-derives session *message history* after every test (an internal-consistency invariant), whereas these snapshots pin assembled behavior plus the external automation output. They are complementary until the backend corpus moves off ACP.

5
.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.zh.md
View File

@@ -53,7 +53,7 @@ Status: implemented
快照运行断言**两个**归一化后的表面，因为 harness 的外部表面是不同的：
1. **stdout transcript**——自动化客户端收到的、经过 framing 的 ACP JSON-RPC 响应与已提交的消息更新。它捕获传输契约的回归，与已提交的 `stdout.expected.jsonl` 比较。
2. **重新持久化的会话 JSONL**，经过规范化后与 `session.jsonl` 比较。同一 fixture 同时作为重放来源和预期日志。提示词文本会被清理；按照[请求头固定 Agent Note](../../archived/testing/2026-07-06-pin-request-header-content-in-one-scenario.md)所述，每种请求头类别由一个场景固定可读提示词与工具内容。Override 场景仅从其 sidecar 派生模型行为。
2. **重新持久化的会话 JSONL**，经过规范化后与 `session.jsonl` 比较。同一 fixture 同时作为重放来源和预期日志。提示词与工具的主体内容会被清理；每种请求头类别由一个场景固定余下的请求头序列。该 pin 默认拥有可读的提示词与工具 schema sidecar；当完整的对应序列相同时，也可将另一个 pin 指定为其中任一来源，因此每个不同的 sidecar 版本只提交一次。fixture 保护会拒绝重复的 sidecar 内容，录制/刷新会拒绝生成不同字节的共享引用方。最初的请求头固定理由保留在[请求头固定 Agent Note](../../archived/testing/2026-07-06-pin-request-header-content-in-one-scenario.md)中。Override 场景仅从其 sidecar 派生模型行为。
两个表面互补：stdout 覆盖精简的自动化线协议，JSONL 覆盖线协议有意省略的 loop、工具和 boundary 结构。
@@ -76,9 +76,10 @@ Status: implemented
- **手工编写包含模型分片的 `llm.json`**——早期草案；复用真实会话日志，使 fixture 成为系统的真实产物而非手工构建的 mock，并让它同时充当行为预期输出。
- **字节级 HTTP 录制库（Polly/nock/MSW）**：否决。与适配器耦合，处理流式 SSE（Server-Sent Events）时笨拙，且层级低于被测对象。
- **从 `turn/end {kind:'error'|'aborted'}` 合成抛错/取消条目**：否决。这会将 `llm-replay` 耦合到 loop 内部的轮次关闭语义，且 `turn/end` 原因是有损的（无法区分抛出的 401 与 finish-error）；显式的 `replay.override.json` 伴随文件是更清晰的 seam。
- **在每个类别 pin 旁复制两个请求头 sidecar**：否决。提示词与工具 schema 的组合各自独立变化，因此一个共享组件发生变更，就会使不相关类别 pin 中字节完全相同的文件产生无意义改动。显式的分组件来源可在不重复内容的情况下，为每个类别保留一个结构性 pin。
## 后果
该测试层为每个场景增加经过评审的输入、会话、stdout、可选 override 和可选 workspace fixture。记录与重放都会把 workspace seed 复制到生成的 cwd。作为回报，该层通过真实 Loader 和工具组合提供确定性的无密钥覆盖。保留下来的大多数场景测试的是组装后的后端而非 ACP；[仅面向自动化的 ACP 决策](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary)将该语料保留在此处，并把向传输无关 headless 套件的任何迁移推迟为一项独立的测试变更（套件级 FIXME 标记了这一点）。
该测试层为每个场景增加经过评审的输入、会话、stdout、可选 override 和可选 workspace fixture，并为每个不同的已固定提示词序列、每个不同的已固定工具 schema 序列各增加一个文件。记录与重放都会把 workspace seed 复制到生成的 cwd。作为回报，该层通过真实 Loader 和工具组合提供确定性的无密钥覆盖。保留下来的大多数场景测试的是组装后的后端而非 ACP；[仅面向自动化的 ACP 决策](../simplification/2026-07-23-acp-automation-only-protocol.md#snapshot-boundary)将该语料保留在此处，并把向传输无关 headless 套件的任何迁移推迟为一项独立的测试变更（套件级 FIXME 标记了这一点）。
本 Agent Note 与[拟议的确定性 Agent Note](../../proposed/testing/2026-06-11-deterministic-and-stress-testing.md)相关，但不取代它：该提案的“通用重放 fixture”在每次测试后重新派生会话*消息历史*（内部一致性不变量），而这些快照固定组装后的行为与外部自动化输出。在后端语料迁出 ACP 之前，两者相互补充。

4
.agents/notes/proposed/architecture/2026-07-27-session-projection-and-command-log.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/proposed/architecture/2026-07-27-session-projection-and-command-log.md
2026-07-27-session-projection-and-command-log.md: 51cc60208ecafd55738c12f1887056c7b0427117
2026-07-27-session-projection-and-command-log.zh.md: 71f6f6ea944c7c1bdd7e560ec8f0dc2528522fc1
2026-07-27-session-projection-and-command-log.md: 6a073c956c27bbfc65cff2d4f44ca12023df0cd5
2026-07-27-session-projection-and-command-log.zh.md: 500f07968db049e4a174ff3b7a075bfe095283db

4
.agents/notes/proposed/architecture/2026-07-27-session-projection-and-command-log.md
View File

@@ -51,7 +51,7 @@ declare module 'cordis' {
- Values are wire JSON payloads; the same map typed end to end (host unit, wire block, React hook) via `import type` — no second DTO table, no separate client-side "views" map. How a value is *rendered* is the slot system's business, never the projection layer's.
- **The host is the only place a projection is computed.** The framework drives every registered unit forward eagerly: each committed session event passes through `apply`; a unit uninterested in an event returns the same state reference, and an unchanged reference (`Object.is`) produces no downstream work. Clients never fold domain events — they receive finished values (baseline block + push frame below). This removes the double-implementation trap (plan's two-event fold written once, on the host) and any client-side domain code.
- **State is always computed, never logged.** The log holds events only; the unit's state lives in the framework's per-session watermark cache (`{state, observedSeq}` per unit) and, in a later phase, in a **persisted projection cache** on the domain-KV storage seam: rows of `(sessionId, key, stateVersion, observedSeq, stateJson)`. A row is never wrong, only possibly stale — `observedSeq` says exactly how stale. The one read recipe, cold and live alike: take the cached state (or `init()`), forward-apply only the events past its watermark, `view` the result. Cold listings (every session's title across all workspaces) become an index read plus, at worst, a short tail replay; the session-persistence seam grows a read-from-seq primitive for that tail in the same later phase. Write policy: throttled (count/interval, configurable) plus two mandatory points — `turn/end` and detach (the live-to-cold moment). A crash between writes costs a longer tail replay, never a wrong value.
- **State is always computed, never logged.** The log holds events only; the unit's state lives in the framework's per-session watermark cache (`{state, observedSeq}` per unit) and, in a later phase, in a **persisted projection cache** on the domain-KV storage seam: rows of `(sessionId, key, ver, seq, val)` (`ver` = the unit's `stateVersion`, `seq` = the watermark, `val` = the state JSON). A row is never wrong, only possibly stale — its `seq` says exactly how stale. The one read recipe, cold and live alike: take the cached state (or `init()`), forward-apply only the events past its watermark, `view` the result. Cold listings (every session's title across all workspaces) become an index read plus, at worst, a short tail replay; the session-persistence seam grows a read-from-seq primitive for that tail in the same later phase. Write policy: throttled (count/interval, configurable) plus two mandatory points — `turn/end` and detach (the live-to-cold moment). A crash between writes costs a longer tail replay, never a wrong value.
- A domain's input event set is its own choice: todos folds `todo/write` alone; plan folds `plan/mode` plus its own `/plan` `command/run` records (see the plan section); goal folds `goal/change` metadata; session title folds its title events (retiring the bespoke `session/title` frame and the client's title-snapshot map — the fourth hand-rolled projection this seam absorbs).
- Registration is an effect (disposer with the fiber): an unloaded plugin's key disappears from subsequent responses and the client reads it as capability absence — HMR semantics for free. Duplicate keys throw. Domain plugins register under `ctx.inject(['sessionProjections'], …)` so headless assemblies without the registry stay unaffected.
- The package owns `./invariant` (every served key has a live registration).
@@ -133,7 +133,7 @@ Infrastructure first; the three in-flight PRs are left untouched and re-target a
2. **Client base**: the generic value store + `useProjection` seat; retire the per-domain cell machinery and, with title's unit registered, the `session/title` frame and title-snapshot map. Depends on 1 for the frame shape (fixtures feed synthetic frames meanwhile).
3. **Command channel**: the two events, executor logging, generic node + keyed slot, notice retirement, `{matched, commandId?}` admission. Parallel with 1.
4. **Domain re-targets** (after 1+2): todo (unit in `tool-todo`, drop the rider field), then plan (two-event unit, RPCs retired, toggle → `/plan`), then goal (`goal/change` unit, drop `goals.get`, move the six `Session` methods into the domain plugin's inject).
5. **Persisted projection cache** (later phase, after the domain-KV storage seam): the `(sessionId, key, stateVersion, observedSeq, state)` rows, throttled writes with turn/end + detach mandatory points, and the persistence read-from-seq primitive for cold tail replay.
5. **Persisted projection cache** (later phase, after the domain-KV storage seam): the `(sessionId, key, ver, seq, val)` rows, throttled writes with turn/end + detach mandatory points, and the persistence read-from-seq primitive for cold tail replay.
## Alternatives considered

4
.agents/notes/proposed/architecture/2026-07-27-session-projection-and-command-log.zh.md
View File

@@ -51,7 +51,7 @@ declare module 'cordis' {
- 值就是协议层的 JSON 载荷；同一张类型表经 `import type` 端到端贯通（host 侧单元、协议块、React 钩子）——没有第二张 DTO 表，也没有独立的客户端「views」表。值如何*渲染*是 slot 体系的事，永远不归投影层管。
- **host 是投影唯一的计算地点。** 框架正向驱动（eager drive）每个已注册的单元：每个已提交的会话事件都经过 `apply`；对某事件不感兴趣的单元返回同一个状态引用，而引用未变（`Object.is`）就不产生任何下游工作。客户端从不折叠领域事件——它们收到的是成品值（基线块 + 下文的推送帧）。这消除了双重实现陷阱（plan 的双事件折叠只在 host 写一遍），也消除了一切客户端侧领域代码。
- **状态永远靠计算得出，绝不入日志。** 日志只存事件；单元的状态住在框架的按会话水位线缓存里（每单元一份 `{state, observedSeq}`），并在后续阶段进入 domain-KV 存储 seam 上的**持久投影缓存（persisted projection cache）**：形如 `(sessionId, key, stateVersion, observedSeq, stateJson)` 的行。一行永远不会是错的，至多是陈旧的——`observedSeq` 精确说明陈旧到哪。冷读与活读共用同一套读取配方：取缓存状态（或 `init()`），只对超出其水位线的事件做正向 `apply`，再对结果做 `view`。冷列表（跨全部 workspace 列出每个会话的标题）变成一次索引读，至多外加一小段尾部回放；session-persistence seam 在同一后续阶段为这段尾部补一个按 seq 起读的原语。写入策略：节流（次数/间隔，可配置）外加两个强制点——`turn/end` 与 detach（由活转冷的时刻）。两次写入之间崩溃的代价是尾部回放更长一些，绝不会是值出错。
- **状态永远靠计算得出，绝不入日志。** 日志只存事件；单元的状态住在框架的按会话水位线缓存里（每单元一份 `{state, observedSeq}`），并在后续阶段进入 domain-KV 存储 seam 上的**持久投影缓存（persisted projection cache）**：形如 `(sessionId, key, ver, seq, val)` 的行（`ver` = 单元的 `stateVersion`，`seq` = 水位线，`val` = 状态 JSON）。一行永远不会是错的，至多是陈旧的——其 `seq` 精确说明陈旧到哪。冷读与活读共用同一套读取配方：取缓存状态（或 `init()`），只对超出其水位线的事件做正向 `apply`，再对结果做 `view`。冷列表（跨全部 workspace 列出每个会话的标题）变成一次索引读，至多外加一小段尾部回放；session-persistence seam 在同一后续阶段为这段尾部补一个按 seq 起读的原语。写入策略：节流（次数/间隔，可配置）外加两个强制点——`turn/end` 与 detach（由活转冷的时刻）。两次写入之间崩溃的代价是尾部回放更长一些，绝不会是值出错。
- 领域的输入事件集由领域自己选择：todos 只折叠 `todo/write`；plan 折叠 `plan/mode` 外加它自己的 `/plan` `command/run` 记录（见 plan 一节）；goal 折叠 `goal/change` 元数据；会话标题折叠其标题事件（顺带下线专设的 `session/title` 帧与客户端的标题快照表——这是该 seam 收编的第四个手工投影）。
- 注册是 effect（disposer 随 fiber 走）：插件卸载后其 key 从后续响应中消失，客户端将其读作能力缺失——HMR（热模块替换）语义随之自动成立。key 重复直接 throw。领域插件在 `ctx.inject(['sessionProjections'], …)` 下注册，因此不带注册表的 headless 组装完全不受影响。
- 该包拥有 `./invariant`（每个被服务的 key 都有一条存活的注册）。
@@ -133,7 +133,7 @@ host 侧命令执行器（`packages/ui/commands`）在调用处理器前追加 `
2. **客户端基座**：通用值仓 + `useProjection` 席位；下线按领域的 cell 机制，并在标题单元注册后一并下线 `session/title` 帧与标题快照表。帧的形状依赖 1（在此之前 fixture（测试前置数据）喂合成帧）。
3. **命令通道**：两个事件、执行器落日志、通用节点 + keyed slot、通知通道下线、`{matched, commandId?}` 准入。与 1 并行。
4. **领域重新对接**（在 1+2 之后）：先 todo（单元进 `tool-todo`，删掉搭载字段），再 plan（双事件单元、RPC 下线、开关改发 `/plan`），最后 goal（`goal/change` 单元，删掉 `goals.get`，把六个 `Session` 方法移入领域插件的 inject）。
5. **持久投影缓存**（后续阶段，待 domain-KV 存储 seam 就绪后）：`(sessionId, key, stateVersion, observedSeq, state)` 行、带 turn/end 与 detach 强制点的节流写入，以及持久化侧供冷尾部回放用的按 seq 起读原语。
5. **持久投影缓存**（后续阶段，待 domain-KV 存储 seam 就绪后）：`(sessionId, key, ver, seq, val)` 行、带 turn/end 与 detach 强制点的节流写入，以及持久化侧供冷尾部回放用的按 seq 起读原语。
## Alternatives considered

6
.agents/notes/proposed/architecture/2026-07-28-storage-root-and-derived-medium-recovery.i18n.yaml Normal file
View File

@@ -0,0 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write .agents/notes/proposed/architecture/2026-07-28-storage-root-and-derived-medium-recovery.md
2026-07-28-storage-root-and-derived-medium-recovery.md: 06fa98b10dc5ac3164d8905e7005a42d9e99ae92
2026-07-28-storage-root-and-derived-medium-recovery.zh.md: b7bd18ffdbfaf412d9a91940cf1770e273f5b847

57
.agents/notes/proposed/architecture/2026-07-28-storage-root-and-derived-medium-recovery.md Normal file
View File

@@ -0,0 +1,57 @@
# Agent Note: Storage root placement and derived-medium recovery
Status: proposed
English | [中文](2026-07-28-storage-root-and-derived-medium-recovery.zh.md)
## Problem
The persisted projection cache ([RFC](2026-07-27-session-projection-and-command-log.md), shipped as `dsh-session-projection-cache`) surfaced two gaps in the storage substrate it landed on. Both are properties of the domain-KV stack ([design](2026-07-24-domain-kv-storage-and-workspace.md)), not of the cache itself, and both bite the cache first because it is the first *derived* medium on that stack.
**Where the files actually live.** The shipped composition gives the json backend a relative root — `root: './.storages'` (apps/cli/cordis.yml) — and `AppCLIEntry.composePatches` patches only the session store's root to the global harness home (`$DSH_HOME/sessions`, default `~/.dsh/sessions`, profile-overridable via `persistenceRoot`); no equivalent patch or profile key exists for `storage-json`. `JsonStorageBackend` never resolves its root either — each unit open joins the still-relative path against whatever `process.cwd()` is at that moment (packages/storage/storage-json/src/index.ts) — the exact hazard the JSONL session backend resolves-once to prevent ("later process.cwd() changes cannot split one backend across roots", packages/session-persistence/session-persistence-jsonl/src/index.ts). Net effect: session logs are global across launch directories, but `workspace.json` and `session_projcache.json` land under `<launch dir>/.storages/`. Two launches from different directories share their sessions yet see different workspace registries and different projection caches — and the cache exists precisely to serve the cross-session cold listing, which now misses for every session last cached under another launch directory.
**How recovery works today.** Inside a healthy medium the cache is fully self-healing by design: a `stateVersion`-mismatched row is discarded and refolded, a log shrunk below a row's watermark is detected by the anchored restore floor and answered with one full re-read, and every background write is fail-soft. But at the *medium* level there is no recovery at all: a truncated, hand-edited, or version-bumped `session_projcache.json` fails `openJsonUnit` with `malformed-medium`/`version-mismatch` (packages/storage/storage-json/src/format.ts), a schema-drifted record fails domain open with `invalid-record` (packages/storage/storage-domain/src/index.ts), the rejection propagates through `SessionProjectionCache[Service.init]`, and under the CLI's fail-loud boot the assembly refuses to start. A file whose entire content is rebuildable from session logs can brick boot. This contradicts the cache package's own stated stance ("a stale or unreadable cache costs a longer tail replay, never a wrong value") and the cache domain spec's JSDoc ("version bumps discard the whole medium"), which today describes an aspiration, not the implementation. The same fail-loud path is *correct* for `workspace.json` — workspace records are authoritative, not derivable — so the missing concept is a per-domain declaration of authority, not a global behavior change.
## Proposal
Two independent changes, one per gap.
### One global storage root, resolved once
- `AppCLIEntry.composePatches` Source 0 additionally patches `storage-json.root` to `join(resolveDshHome(), 'storages')` — `~/.dsh/storages` by default, beside `~/.dsh/sessions` — and `PROFILE_MAPPINGS` gains `storageRoot` → (`storage-json`, `root`), mirroring `persistenceRoot` exactly. The yml keeps `./.storages` as the raw-composition engineering default (tests and bare Loader boots are unaffected), same layering as the session root today.
- `JsonStorageBackend` resolves its configured root once at construction (`resolve(config.root)`), adopting the JSONL backend's recorded rationale verbatim: a later `process.cwd()` change must not split one backend across roots. The SQLite storage backend already resolves its path.
- Pre-release stance applies: no migration shim. A deployment that cached under `<cwd>/.storages` re-derives everything (workspace re-bootstraps from the header index; the projection cache refolds lazily) or moves the two json files by hand once.
### Declared derived media: reset instead of reject
- `DomainSpec` gains `recovery?: 'reject' | 'reset'` (default `'reject'`). The spec object is already the single source of a domain's identity and layout; whether its medium is authoritative or derived is the same kind of fact and lives in the same place. `session_projcache` declares `'reset'`; `workspace` stays on the default.
- `KvFacet` gains one primitive: `destroy(descriptor): Promise<void>` — remove the unit's medium entirely (json: delete the file; sqlite: drop the unit's tables). Like `open`, it is a backend storage primitive, not policy.
- `DomainFacility.open`, when a spec declares `'reset'` and the open fails with exactly a damage-class error — `StorageError('version-mismatch' | 'malformed-medium')` or `DomainError('invalid-record')` — logs one warning naming the domain and the discarded medium, calls `destroy`, and opens again empty. Every other failure (`backend-not-found`, `facet-unsupported`, `already-open`, I/O errors) stays loud regardless of the declaration: misconfiguration and environmental faults are not medium damage. The retry is single-shot — a second failure propagates, so a persistently failing medium cannot loop.
- With this in place the cache domain spec's version field gains its intended meaning: bumping `version` (or letting zod reject drifted rows) genuinely discards the whole medium and the cache rebuilds through its normal write points and cold reads — the recovery ladder's outermost rung, matching the row-level rungs already shipped.
## Alternatives considered
**Keep per-launch-directory `.storages` (status quo)** — rejected: sessions are global, so every derived-from-sessions medium splits against its own source of truth; the cache's motivating scenario (one listing over all sessions) structurally misses rows, and the workspace registry indexes sessions it cannot see from another launch directory.
**Patch only the projection cache's route to a global root, leave `workspace.json` per-cwd** — rejected: the workspace registry has the identical global-vs-cwd mismatch, and the user decision that shaped the cache placed it deliberately beside `workspace.json` — one hub root keeps the media co-located and the mental model single.
**Cache-plugin-local recovery (catch damage errors in `SessionProjectionCache[Service.init]`, delete the file, reopen)** — rejected: the plugin cannot name the medium path without reaching around the backend abstraction, and every future derived domain would re-implement the same catch; the facility is the one place that already classifies open failures.
**Fall back to an ephemeral in-memory domain on damage** — rejected: it silently degrades to memory-only for the life of the process and the damaged file never heals; the next boot fails the same way.
**Rename the damaged medium aside (`<unit>.json.corrupt-<ts>`) instead of deleting** — not chosen: a derived medium's damaged bytes have no recovery value (the logs are the source of truth) and the litter accumulates unbounded; delete is the honest operation. Rename-aside remains the right choice if a future *authoritative* domain ever wants reset semantics — which is exactly why `recovery` is per-spec.
**A blanket auto-reset for every domain (no spec field)** — rejected outright: `workspace.json` is authoritative user data; silently resetting it on a version bump would destroy workspaces. Authority is a property of the domain and must be declared by its owner.
## Acceptance criteria
- `dsh` launched from any directory reads and writes the same `$DSH_HOME/storages/*.json` (default `~/.dsh/storages`); the profile key `storageRoot` overrides it; a raw Loader boot of the yml still lands in `./.storages` relative to the boot cwd, resolved once at backend construction.
- With a truncated, version-bumped, or schema-drifted `session_projcache.json`, the assembly boots clean: one warning names the discarded medium, the file is gone, the cache rebuilds through normal operation, and the cold listing column reappears as sessions are re-checkpointed.
- The same damage to `workspace.json` still fails boot loudly.
- Facility tests cover: each damage class resets a `'reset'` domain exactly once; non-damage failures stay loud on a `'reset'` domain; a `'reject'` domain propagates every failure; `destroy` removes the medium on both shipped backends.
## Risks
- **Auto-delete on a misclassified error destroys a healthy file.** Mitigated by the closed damage-class list: reset fires only on the three deterministic parse-time codes; ENOENT is already "empty unit", and every I/O error (EACCES, EIO) propagates loudly. The single-shot retry bounds the blast radius to one delete per open.
- **Root relocation changes where existing checkouts look.** Accepted under the pre-release stance (backends reject old formats, no external consumers); the note above records the one-time manual move for anyone who cares about a per-cwd `workspace.json`'s content.
- **`destroy` is a new destructive primitive on the storage seam.** Its only caller is the facility's declared-reset path; the backend contract documents it as facility-owned, and nothing model-facing or user-facing can reach it.

57
.agents/notes/proposed/architecture/2026-07-28-storage-root-and-derived-medium-recovery.zh.md Normal file
View File

@@ -0,0 +1,57 @@
# Agent Note：存储根目录落点与派生介质恢复
Status: proposed
[English](2026-07-28-storage-root-and-derived-medium-recovery.md) | 中文
## Problem
持久投影缓存（[RFC](2026-07-27-session-projection-and-command-log.md)，已作为 `dsh-session-projection-cache` 落地）暴露了它所依托的存储基座的两个缺口。二者都是 domain-KV 栈（[设计](2026-07-24-domain-kv-storage-and-workspace.md)）的属性而非缓存自身的问题，且都首先咬到缓存——因为它是这条栈上第一个*派生*介质。
**文件到底存在哪。** 出厂组合给 json 后端的是相对根目录——`root: './.storages'`（apps/cli/cordis.yml）——而 `AppCLIEntry.composePatches` 只把会话存储的根 patch 到全局 harness home（`$DSH_HOME/sessions`，默认 `~/.dsh/sessions`，可经 profile 键 `persistenceRoot` 覆盖）；`storage-json` 没有对应的 patch 也没有 profile 键。`JsonStorageBackend` 自己也从不 resolve 根——每次打开 unit 都把仍然相对的路径 join 到当时的 `process.cwd()` 上（packages/storage/storage-json/src/index.ts）——这正是 JSONL 会话后端用「构造时 resolve 一次」防住的那个隐患（"later process.cwd() changes cannot split one backend across roots"，packages/session-persistence/session-persistence-jsonl/src/index.ts）。净效果：会话日志跨启动目录全局共享，但 `workspace.json` 和 `session_projcache.json` 落在 `<启动目录>/.storages/` 下。从两个不同目录启动，会话相同，工作区注册表和投影缓存却各是一份——而缓存存在的意义恰恰是跨会话冷列表，如今凡是上次在别的启动目录下缓存过的会话全部 miss。
**现在是怎么恢复的。** 在健康介质内部，缓存按设计完全自愈：`stateVersion` 不匹配的行被丢弃重折，日志缩短到行水位以下由带锚的 restore floor 检出并以一次全量重读回答，每次后台写都是 fail-soft。但在*介质*层面完全没有恢复：被截断、被手改或版本被 bump 的 `session_projcache.json` 会让 `openJsonUnit` 以 `malformed-medium`/`version-mismatch` 失败（packages/storage/storage-json/src/format.ts），schema 漂移的记录让域 open 以 `invalid-record` 失败（packages/storage/storage-domain/src/index.ts），拒绝一路穿过 `SessionProjectionCache[Service.init]`，在 CLI 的 fail-loud 启动下整个组装拒绝启动。一个内容完全可从会话日志重建的文件能把启动搞死。这与缓存包自己声明的立场（"a stale or unreadable cache costs a longer tail replay, never a wrong value"）和缓存域 spec 的 JSDoc（"version bumps discard the whole medium"）相矛盾——后者今天描述的是愿望而非实现。同一条 fail-loud 路径对 `workspace.json` 却是*正确*的——工作区记录是权威数据，不可派生——所以缺的概念是按域声明权威性，而不是全局改行为。
## Proposal
两个独立改动，一个缺口一个。
### 全局唯一存储根，构造时 resolve 一次
- `AppCLIEntry.composePatches` 的 Source 0 追加把 `storage-json.root` patch 到 `join(resolveDshHome(), 'storages')`——默认 `~/.dsh/storages`，与 `~/.dsh/sessions` 并肩——并且 `PROFILE_MAPPINGS` 增加 `storageRoot` →（`storage-json`，`root`），与 `persistenceRoot` 完全镜像。yml 保留 `./.storages` 作为裸组合的工程默认（测试和裸 Loader 启动不受影响），分层方式与今天的会话根相同。
- `JsonStorageBackend` 在构造时对配置根 `resolve` 一次，原样采纳 JSONL 后端已记录的理由：后续 `process.cwd()` 变化不得把一个后端劈到多个根下。SQLite 存储后端已经 resolve 其路径。
- 适用 pre-release 立场：不做迁移垫片。曾在 `<cwd>/.storages` 下缓存过的部署要么全部重新派生（工作区从 header 索引重新 bootstrap；投影缓存惰性重折），要么手动把两个 json 文件挪一次。
### 声明派生介质：损坏时重置而非拒绝
- `DomainSpec` 增加 `recovery?: 'reject' | 'reset'`（默认 `'reject'`）。spec 对象已经是一个域的身份与布局的单一来源；其介质是权威还是派生属于同类事实，落在同一处。`session_projcache` 声明 `'reset'`；`workspace` 保持默认。
- `KvFacet` 增加一个原语：`destroy(descriptor): Promise<void>`——整体移除该 unit 的介质（json：删文件；sqlite：drop 该 unit 的表）。与 `open` 一样，它是后端存储原语，不是策略。
- `DomainFacility.open` 在 spec 声明 `'reset'` 且 open 恰以损坏类错误失败时——`StorageError('version-mismatch' | 'malformed-medium')` 或 `DomainError('invalid-record')`——记一条命名该域和被丢弃介质的警告，调用 `destroy`，再空开一次。其余一切失败（`backend-not-found`、`facet-unsupported`、`already-open`、I/O 错误）无论声明与否都保持大声：配置错误和环境故障不是介质损坏。重试单发——第二次失败原样传播，持续失败的介质不会成环。
- 有了这个，缓存域 spec 的 version 字段才获得其本意：bump `version`（或让 zod 拒绝漂移行）真正丢弃整个介质，缓存经正常写点和冷读重建——恢复阶梯的最外一档，与已落地的行级各档对齐。
## Alternatives considered
**保持按启动目录的 `.storages`（现状）**——拒绝：会话是全局的，所以每个从会话派生的介质都与自己的真源劈叉；缓存的动机场景（一次列出全部会话）结构性丢行，工作区注册表索引着从另一个启动目录看不见的会话。
**只把投影缓存的 route 指到全局根，`workspace.json` 留在 per-cwd**——拒绝：工作区注册表有一模一样的全局 vs per-cwd 错位，而且塑造缓存的用户决策就是刻意把它放在 `workspace.json` 旁边——一个 hub 根让介质同址、心智模型单一。
**缓存插件本地恢复（在 `SessionProjectionCache[Service.init]` 捕获损坏错误、删文件、重开）**——拒绝：插件不越过后端抽象就叫不出介质路径，且未来每个派生域都要重抄同一段 catch；facility 是唯一已经在分类 open 失败的地方。
**损坏时退到内存态临时域**——拒绝：进程余生静默降级为仅内存，损坏文件永不自愈；下次启动照样失败。
**把损坏介质改名旁置（`<unit>.json.corrupt-<ts>`）而非删除**——未选：派生介质的损坏字节没有恢复价值（日志才是真源），残骸无界累积；删除才是诚实的操作。若未来某个*权威*域想要重置语义，旁置改名才是对的——这正是 `recovery` 按 spec 声明的理由。
**所有域一律自动重置（不加 spec 字段）**——断然拒绝：`workspace.json` 是权威用户数据；版本 bump 时静默重置会毁掉工作区。权威性是域的属性，必须由其所有者声明。
## Acceptance criteria
- 从任意目录启动 `dsh` 都读写同一份 `$DSH_HOME/storages/*.json`（默认 `~/.dsh/storages`）；profile 键 `storageRoot` 可覆盖；裸 Loader 启动 yml 仍落在相对启动 cwd 的 `./.storages`,并在后端构造时 resolve 一次。
- `session_projcache.json` 被截断、版本 bump 或 schema 漂移时，组装干净启动：一条警告命名被丢弃的介质，文件消失，缓存经正常运转重建，冷列表列随会话重新 checkpoint 逐步回归。
- 同样的损坏发生在 `workspace.json` 上仍大声拒绝启动。
- facility 测试覆盖：每个损坏类恰好重置一次 `'reset'` 域；非损坏失败在 `'reset'` 域上保持大声；`'reject'` 域传播一切失败；`destroy` 在两个出厂后端上都移除介质。
## Risks
- **错误分类失误导致自动删除健康文件。** 由封闭的损坏类清单缓解：重置只在三个确定性解析期代码上触发；ENOENT 本来就是「空 unit」,一切 I/O 错误（EACCES、EIO）大声传播。单发重试把爆炸半径限定为每次 open 至多一删。
- **根迁移改变既有 checkout 的查找位置。** 在 pre-release 立场下接受（后端拒绝旧格式、无外部消费者）；上文为在乎 per-cwd `workspace.json` 内容的人记录了一次性手动搬移。
- **`destroy` 是存储 seam 上新增的破坏性原语。** 唯一调用方是 facility 的声明重置路径；后端契约将其记档为 facility 专属,任何面向模型或面向用户的路径都触不到它。

10
apps/cli/cordis.yml
View File

@@ -116,6 +116,16 @@
- id: workspace
name: '@deepseek-ai/dsh-workspace'
# Persisted projection cache: durable per-session checkpoints of every
# registered projection unit (json backend → ./.storages/session_projcache.json,
# beside workspace.json), throttled between the two mandatory points
# (turn/end + detach), serving cold listings without full-log loads.
- id: session-projection-cache
name: '@deepseek-ai/dsh-session-projection-cache'
config:
writeEveryEvents: 200
writeIntervalMs: 5000
# Managed child-process groups for the bash executor (spawn/kill/output plumbing).
- id: subprocess
name: '@deepseek-ai/dsh-subprocess-local'

1
apps/cli/package.json
View File

@@ -57,6 +57,7 @@
"@deepseek-ai/dsh-session": "workspace:^",
"@deepseek-ai/dsh-session-persistence-jsonl": "workspace:^",
"@deepseek-ai/dsh-session-projection": "workspace:^",
"@deepseek-ai/dsh-session-projection-cache": "workspace:^",
"@deepseek-ai/dsh-session-title": "workspace:^",
"@deepseek-ai/dsh-session-title-first-message-llm": "workspace:^",
"@deepseek-ai/dsh-skill": "workspace:^",

33
apps/web/tests/todo-display.snapshot.ts
View File

@@ -4,8 +4,9 @@
// Opens the fixture history session and pins the todo_write turn's two
// surfaces: the dedicated TodoRow in the chat flow (keyed toolview, summary
// derived from the call args) and the TodoPanel plan strip riding the
// 'conversation.input.dock' slot (fed by ConversationSnapshot.todos, seeded
// by the tail history page), including the collapse interaction.
// 'conversation.input.dock' slot (fed by the host `todos` projection via
// useProjection, seeded by the tail history page), including the collapse
// interaction and the next-turn clearance of the standing plan.
import { readFileSync } from 'node:fs'
import { join } from 'node:path'
import { act, cleanup, fireEvent, screen, waitFor, within } from '@testing-library/react'
@@ -189,3 +190,31 @@ it('collapses the plan strip to the count summary and restores it', async () =>
fireEvent.click(header)
expect(panel.querySelectorAll('li')).toHaveLength(3)
})
it('hides the plan strip when the next turn starts', async () => {
boot()
await openFixtureSession()
expect(document.querySelector('[data-testid="todo-panel"]')).not.toBeNull()
const composer = await screen.findByPlaceholderText('Message the agent', {}, { timeout: 10_000 })
fireEvent.change(composer, { target: { value: '下一轮清空计划' } })
fireEvent.keyDown(composer, { key: 'Enter' })
await screen.findByText('下一轮清空计划', { exact: true }, { timeout: 10_000 })
await waitFor(() => {
expect(document.querySelector('[data-testid="todo-panel"]')).toBeNull()
}, { timeout: 10_000 })
expect({
promptVisible: screen.getByText('下一轮清空计划', { exact: true }).textContent,
panelGone: document.querySelector('[data-testid="todo-panel"]') === null,
// Historical todo_write row stays in the flow; only the dock strip clears.
rowStillPresent: document.querySelector('[data-sample="todo-row"]') !== null,
}).toMatchInlineSnapshot(`
{
"panelGone": true,
"promptVisible": "下一轮清空计划",
"rowStillPresent": true,
}
`)
})

5
docs/capability-seams.md
View File

@@ -77,6 +77,8 @@ flowchart LR
pkg_session_projection["session-projection"]
svc_sessionProjections["ctx.sessionProjections<br/>Session projection units"]
pkg_host_apiproxy["host-apiproxy"]
pkg_session_projection_cache["session-projection-cache"]
svc_sessionProjectionCache["ctx.sessionProjectionCache<br/>Persisted projection cache"]
svc_tui["ctx.tui<br/>Mounted-terminal interaction service"]
pkg_skill["skill"]
svc_skills["ctx.skills<br/>Skill provider registry"]
@@ -184,6 +186,7 @@ flowchart LR
pkg_session_persistence_jsonl --> svc_sessionPersistence
pkg_session_persistence_sqlite --> svc_sessionPersistence
pkg_session_projection --> svc_sessionProjections
pkg_session_projection_cache --> svc_sessionProjectionCache
pkg_session_query --> svc_sessionQuery
pkg_session_query_sqlite --> svc_sessionQuery
pkg_session_reference --> svc_sessionReferences
@@ -261,6 +264,7 @@ flowchart LR
svc_sessionPersistence --> pkg_session_query
svc_sessionPersistence --> pkg_session_query_sqlite
svc_sessionPersistence --> pkg_tool_bash
svc_sessionProjectionCache --> pkg_host_apiproxy
svc_sessionProjections --> pkg_host_apiproxy
svc_sessionProjections --> pkg_session_title
svc_sessionProjections --> pkg_tool_todo
@@ -336,6 +340,7 @@ flowchart LR
| `ctx.planMode` | `core` | [`plan-mode`](../packages/plan/plan-mode) | - | - | - | Folds logged plan/mode state, flushes user selections at turn boundaries, renders deployment-owned guidance, registers /plan, and keeps the plan-exit schema stable across transitions. |
| `ctx.commands` | `core` | [`commands`](../packages/ui/commands) | - | [`tui`](../packages/ui/tui) | - | Plugins register direct human commands; TUI consumes the effective per-agent catalog without sending invocations to the model. |
| `ctx.sessionProjections` | `core` | [`session-projection`](../packages/session-projection/session-projection) | - | [`tool-todo`](../packages/todo/tool-todo), [`session-title`](../packages/session-title/session-title), [`host-apiproxy`](../packages/host/apiproxy) | - | Domains register state-driven fold units; the eager drive keeps per-session watermark states and api-proxy serves baselines and pushes changed values. |
| `ctx.sessionProjectionCache` | `core` | [`session-projection-cache`](../packages/session-projection/session-projection-cache) | - | [`host-apiproxy`](../packages/host/apiproxy) | - | Durably checkpoints projection unit states per session (throttled + turn/end/detach mandatory points) and serves the cold-read ladder: cache row + persistence tail replay, so listings never load full logs. |
| `ctx.tui` | `bundle` | [`tui`](../packages/ui/tui) | - | - | - | One TUI front door provides a FIFO overlay host; injected plugins receive caller-fiber ownership without access to pi-tui or terminal lifecycle state. |
| `ctx.skills` | `seam` | [`skill`](../packages/skill/skill) | [`skill-local`](../packages/skill/skill-local) | [`tool-skill`](../packages/skill/tool-skill) | - | Merges provider skill catalogs; tool-skill renders the session-prefix catalog and loads complete skill bodies. |
| `ctx.agents` | `core` | [`agent`](../packages/core/agent) | - | [`agent-loop`](../packages/core/agent-loop), [`acp`](../packages/acp/acp), [`cli-demo`](../packages/examples/cli-demo), [`subagent-inprocess`](../packages/subagent/subagent-inprocess), [`tui-demo`](../packages/examples/tui-demo) | - | Owns live Agent handles, the create/resume factory seam, and process-local initiator propagation. |

21
docs/config-catalog.md
View File

@@ -1048,6 +1048,27 @@ export type JournalMode = 'wal' | 'delete' | 'truncate' | 'persist'
Source: [`packages/session-persistence/session-persistence-sqlite/src/index.ts:58`](../packages/session-persistence/session-persistence-sqlite/src/index.ts)
## `@deepseek-ai/dsh-session-projection-cache`
Requires: `storageDomain` · `sessionProjections` · `sessionPersistence` · `sessions`
```ts config-catalog
/**
* Plugin config. Both throttle triggers are deployment choices with no
* universally correct value, so the composition states them explicitly
* (cordis.yml); the two mandatory write points (`turn/end` and session
* disposal) are policy, not tunables, and always fire.
*/
export interface Config {
/** Committed events per session that force a durable checkpoint write between mandatory points. */
writeEveryEvents: number
/** Longest time (milliseconds) a dirty checkpoint may stay unwritten between mandatory points. */
writeIntervalMs: number
}
```
Source: [`packages/session-projection/session-projection-cache/src/index.ts:42`](../packages/session-projection/session-projection-cache/src/index.ts)
## `@deepseek-ai/dsh-session-query-sqlite`
Requires: `sessions`

150
docs/cordis-catalog/services.md
View File

@@ -1042,7 +1042,9 @@ abstract append(id: SessionId, events: readonly SessionEvent[]): Promise<void>
* A coordinator-backed cold load reserves the identity across storage awaits,
* so concurrent publication of a same-id live Session rejects.
* Returned events are detached, and every identified message is deeply
* frozen; malformed identified messages reject before any stored event is returned.
* frozen. Coordinator-backed implementations upgrade supported pre-identity
* message events before validation; other malformed messages reject before
* any stored event is returned.
* @param id - the persisted session to reload.
* @returns the header and a log ending on a balanced `turn/end`.
*/
@@ -1052,14 +1054,34 @@ abstract load(id: SessionId): Promise<{ meta: SessionHeader; events: SessionEven
* Inspect a header and its valid contiguous stored prefix without repairing
* a torn tail, closing an interrupted turn, or publishing coordinator state.
* This read is serialized with writes for the same id and returns detached
* values with deeply frozen identified messages, so observers cannot mutate message
* identity/content or backend-owned state. Malformed identified messages reject.
* values with upgraded, deeply frozen identified messages, so observers
* cannot mutate message identity/content or backend-owned state. Other
* malformed messages reject.
* @param id - the persisted session to inspect.
* @param signal - optional cancellation for queued and backend read work.
* @returns the header and valid stored event prefix exactly as observed.
*/
abstract inspect(id: SessionId, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }>
/**
* Read the stored events from `fromSeq` onward — the read-from-seq
* primitive for read models that resume from a watermark (e.g. a persisted
* projection cache folding only the tail past its checkpoint). Like
* {@link inspect} it is non-mutating and detached: no torn-tail truncation,
* no synthetic closers, no coordinator-state publication; only events from
* the valid contiguous stored prefix are returned, so a torn fragment never
* reaches the caller. `fromSeq` at or beyond the stored prefix returns an
* empty event list (never an error). Backends whose medium can seek by seq
* (SQLite) read only the suffix; sequential media (JSONL, both encodings)
* still parse the whole artifact and skip forward — the primitive bounds
* what is RETURNED and refolded, not every backend's physical read.
* @param id - the persisted session to read.
* @param fromSeq - first event seq to include; a non-negative safe integer.
* @param signal - optional cancellation for queued and backend read work.
* @returns the header and the stored events with `seq >= fromSeq`.
*/
abstract readFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }>
/**
* Lightweight listing from metadata, without a full-log parse.
* @param signal - optional cancellation for backend listing work.
@@ -1084,6 +1106,54 @@ Types: [SessionEvent](../core-data-structures/core.md) · [SessionHeader](../cor
Source: [`packages/session-persistence/session-persistence/src/index.ts:52`](../../packages/session-persistence/session-persistence/src/index.ts)
## `ctx.sessionProjectionCache` — `SessionProjectionCache`
The persisted projection cache service. Opens the `session_projcache` domain at init, checkpoints live sessions on a throttled write-behind (count/interval triggers from Config) plus two mandatory points — `turn/end` and session disposal (the live-to-cold moment) — and serves the cold-read ladder: cached row, persistence `readFrom` tail, registry `restore`, durable write-back. Every durable write is fail-soft: failures log a warning and the cache self-heals on the next write or cold read.
```ts cordis-catalog
/**
* The zero-I/O listing read: whole values viewed straight from the stored
* rows (version-matching keys only), each cut carried with its watermark
* so a client value store can seed under its higher-seq-wins rule — as
* stale as the last durable checkpoint but never wrong, and never from an
* unrelated log (the caller's header is the identity witness). Fresher
* paths (the history tail baseline, {@link coldSnapshot}) supersede these
* values whenever a session is actually opened.
* @param meta - the listed session's header (identity witness; no log read).
* @returns the cut (`asOfSeq` = lowest served-row watermark), or
* `undefined` when no usable row exists for this lifecycle.
*/
cachedSnapshot(meta: SessionHeader): ProjectionSnapshot | undefined
/**
* Durably checkpoint one live session NOW (both mandatory points call
* this; tests and carriers may too). The registry cut is snapshotted at
* this boundary (states are live references), then the whole record is
* replaced. NOT fail-soft — callers on the fail-soft paths contain it.
* @param session - the live session to checkpoint.
* @returns resolution after durability and event emission.
*/
async write(session: Session): Promise<void>
/**
* Cold-read one persisted session's projections with zero full-log load:
* cached rows + a persistence `readFrom` tail from the registry's restore
* floor, refolded by the registry and written back (fail-soft) so the next
* cold read starts closer. A cache row invalidated by a shrunk log
* (crash-repair truncation) triggers one full re-read from seq 0 — the
* ladder's slow rung, still no crash. Rejects when the session has no
* persisted log (`not found` from the persistence seam).
* @param id - the persisted session to read.
* @param signal - optional cancellation for the persistence reads.
* @returns the snapshot cut at the stored log end.
*/
async coldSnapshot(id: SessionId, signal?: AbortSignal): Promise<ProjectionSnapshot>
```
Types: [Session](../core-data-structures/session.md) · [SessionHeader](../core-data-structures/persistence.md) · [SessionId](../core-data-structures/core.md)
Source: [`packages/session-projection/session-projection-cache/src/index.ts:71`](../../packages/session-projection/session-projection-cache/src/index.ts)
## `ctx.sessionProjections` — `SessionProjectionRegistry`
`ctx.sessionProjections`: the projection unit table and its drive. The service subscribes to `session/event` once; every committed event passes every registered unit's `apply` (eager drive), and a changed state reference notifies the change feed with the schema-validated view. Cells build lazily — a unit registered after events flowed, or a session older than the registry, folds `init` over the in-memory log on first touch (event or read). Registration is an effect (disposer rides the calling fiber): an unloaded domain plugin's key disappears from snapshots and clients read it as capability absence. Duplicate keys throw. Domain plugins register under `ctx.inject(['sessionProjections'], …)` so headless assemblies without the registry stay unaffected.
@@ -1116,11 +1186,81 @@ onChanged(listener: ProjectionChangeListener): () => void
* @returns the snapshot; `values` is empty when no unit is registered.
*/
snapshot(session: Session): ProjectionSnapshot
/**
* State-level checkpoint of every registered unit for one session, read
* from the watermark cache (missing cells fold lazily over the in-memory
* log). This is the write side of the persisted projection cache: the
* returned rows are the `(key → {ver, seq, val})` part of the durable
* `(sessionId, key, ver, seq, val)`
* rows. Every `val` is a DETACHED structured clone — never the live
* cell reference: the watermark cache is this registry's authoritative
* mutable state, and a caller reaching the live reference could corrupt
* every subsequent snapshot and frame through it (plain JSON by the unit
* contract, so the clone is total).
* @param session - the session whose unit states are checkpointed.
* @returns one row per registered key; empty when no unit is registered.
*/
checkpoint(session: Session): ProjectionCheckpoint
/**
* The stored seq a {@link restore} tail read over `checkpoint` must start
* at: one event BELOW the lowest usable watermark (a row is usable when
* its `ver` matches the live unit's `stateVersion`; an absent or mismatched row
* pulls the floor to `0` — that key must refold the full log). The
* one-below anchor is load-bearing: the tail then proves how far the
* stored log still extends, so {@link restore} can detect a log that
* shrank below a row's watermark (crash-repair truncation) instead of
* serving the stale row as current — an empty tail read from the anchor
* yields an end below every watermark and the restore rejects for a full
* re-read.
* @param checkpoint - persisted rows for one session (possibly stale or empty).
* @returns the seq to hand the persistence `readFrom`, or `undefined`
* when no unit is registered (no read needed — {@link restore} would
* serve empty values regardless).
*/
restoreFloor(checkpoint: ProjectionCheckpoint): number | undefined
/**
* View a checkpoint's rows without any log read: for every registered
* unit whose row's `ver` matches, serve the schema-validated
* `view` of the stored state; mismatched or absent rows leave their key
* absent (a cold or listing consumer treats it as not-yet-available and a
* fuller read path refolds it). The zero-I/O rung of the read ladder —
* values are as stale as their rows, never wrong.
* @param checkpoint - persisted rows for one session (possibly stale or empty).
* @returns whole values per key with a usable row; empty when none.
*/
viewCheckpoint(checkpoint: ProjectionCheckpoint): Partial<SessionProjectionMap>
/**
* Cold read: fold every registered unit over a stored log suffix, seeding
* each from its checkpoint row when usable — the one read recipe (cached
* state + forward tail replay + `view`) applied without a live `Session`.
* Call with the events returned by a persistence
* `readFrom(id, restoreFloor(checkpoint))` and that same floor as
* `baseSeq`; the floor's one-below anchor makes the supplied end honest,
* so a shrunk log is detected here. A row is usable iff its
* `ver` matches the live unit's `stateVersion`, it does not predate `baseSeq`
* (`seq >= baseSeq - 1`), and it does not claim events past the
* supplied end (`seq <= endSeq`); an unusable row is discarded
* and its key refolds from `init` — which is only sound over the full
* log, so a discarded row with `baseSeq > 0` throws (the caller re-reads
* from seq 0, e.g. after a crash-repair truncation shrank the log below
* a row's watermark).
* @param checkpoint - persisted rows for one session (possibly stale or empty).
* @param events - the stored events with `seq >= baseSeq`, in seq order.
* @param baseSeq - the seq `events` starts at (its first event's seq when non-empty).
* @returns the snapshot cut at the supplied log end (`asOfSeq` is the last
* supplied event's seq, `baseSeq - 1` for an empty tail) plus the
* refreshed checkpoint rows at that cut, ready for a durable write-back.
*/
restore(checkpoint: ProjectionCheckpoint, events: readonly SessionEvent[], baseSeq: number): { snapshot: ProjectionSnapshot; checkpoint: ProjectionCheckpoint }
```
Types: [Session](../core-data-structures/session.md)
Types: [Session](../core-data-structures/session.md) · [SessionEvent](../core-data-structures/core.md)
Source: [`packages/session-projection/session-projection/src/index.ts:136`](../../packages/session-projection/session-projection/src/index.ts)
Source: [`packages/session-projection/session-projection/src/index.ts:156`](../../packages/session-projection/session-projection/src/index.ts)
## `ctx.sessionQuery` — `SessionQueryService` (abstract seam)

4
docs/event-producer-consumer.md
View File

@@ -32,8 +32,8 @@ This matrix shows which packages dispatch each harness-owned event and which pac
| `goal/changed` | `emit` | [`packages/goal/goal/src/types.ts:169`](../packages/goal/goal/src/types.ts) | [`goal`](../packages/goal/goal) (`emit`) | [`goal-session`](../packages/goal/goal-session) |
| `llm/stream` | `waterfall` | [`packages/llm/llm/src/index.ts:58`](../packages/llm/llm/src/index.ts) | [`llm`](../packages/llm/llm) (`waterfall`) | [`agent-loop`](../packages/core/agent-loop), [`llm`](../packages/llm/llm), [`llm-replay`](../packages/support/llm-replay), [`session-checkpoint-policy`](../packages/session-persistence/session-checkpoint-policy), [`session-title`](../packages/session-title/session-title) |
| `session/created` | `emit` | [`packages/core/session/src/index.ts:71`](../packages/core/session/src/index.ts) | [`session`](../packages/core/session) (`events.dispatch`) | `apiproxy`, [`compact`](../packages/compact/compact), [`goal`](../packages/goal/goal), [`hook-protocol`](../packages/hooks/hook-protocol), [`jsonrpc`](../packages/ui/jsonrpc), [`llm-retry`](../packages/llm/llm-retry), [`plan-mode`](../packages/plan/plan-mode), [`session`](../packages/core/session), [`session-persistence`](../packages/session-persistence/session-persistence), [`session-telemetry`](../packages/telemetry/session-telemetry), [`tools`](../packages/core/tools), [`user-approval`](../packages/ui/user-approval) |
| `session/disposed` | `emit` | [`packages/core/session/src/index.ts:81`](../packages/core/session/src/index.ts) | [`session`](../packages/core/session) (`events.dispatch`) | [`agent-loop`](../packages/core/agent-loop), `apiproxy`, [`session-persistence`](../packages/session-persistence/session-persistence), [`session-telemetry`](../packages/telemetry/session-telemetry), [`session-title`](../packages/session-title/session-title) |
| `session/event` | `emit` | [`packages/core/session/src/index.ts:93`](../packages/core/session/src/index.ts) | [`session`](../packages/core/session) (`events.dispatch`) | [`acp`](../packages/acp/acp), `apiproxy`, [`cli-demo`](../packages/examples/cli-demo), [`compact`](../packages/compact/compact), [`compact-basic`](../packages/compact/compact-basic), [`goal`](../packages/goal/goal), [`goal-session`](../packages/goal/goal-session), [`hook-protocol`](../packages/hooks/hook-protocol), [`jsonrpc`](../packages/ui/jsonrpc), [`plan-mode`](../packages/plan/plan-mode), [`session`](../packages/core/session), [`session-persistence`](../packages/session-persistence/session-persistence), [`session-projection`](../packages/session-projection/session-projection), [`session-telemetry`](../packages/telemetry/session-telemetry), [`session-title`](../packages/session-title/session-title), [`token-meter`](../packages/llm/token-meter), [`tools`](../packages/core/tools), [`tui`](../packages/ui/tui), [`user-approval`](../packages/ui/user-approval), [`workspace-context`](../packages/context/workspace-context) |
| `session/disposed` | `emit` | [`packages/core/session/src/index.ts:81`](../packages/core/session/src/index.ts) | [`session`](../packages/core/session) (`events.dispatch`) | [`agent-loop`](../packages/core/agent-loop), `apiproxy`, [`session-persistence`](../packages/session-persistence/session-persistence), [`session-projection-cache`](../packages/session-projection/session-projection-cache), [`session-telemetry`](../packages/telemetry/session-telemetry), [`session-title`](../packages/session-title/session-title) |
| `session/event` | `emit` | [`packages/core/session/src/index.ts:93`](../packages/core/session/src/index.ts) | [`session`](../packages/core/session) (`events.dispatch`) | [`acp`](../packages/acp/acp), `apiproxy`, [`cli-demo`](../packages/examples/cli-demo), [`compact`](../packages/compact/compact), [`compact-basic`](../packages/compact/compact-basic), [`goal`](../packages/goal/goal), [`goal-session`](../packages/goal/goal-session), [`hook-protocol`](../packages/hooks/hook-protocol), [`jsonrpc`](../packages/ui/jsonrpc), [`plan-mode`](../packages/plan/plan-mode), [`session`](../packages/core/session), [`session-persistence`](../packages/session-persistence/session-persistence), [`session-projection`](../packages/session-projection/session-projection), [`session-projection-cache`](../packages/session-projection/session-projection-cache), [`session-telemetry`](../packages/telemetry/session-telemetry), [`session-title`](../packages/session-title/session-title), [`token-meter`](../packages/llm/token-meter), [`tools`](../packages/core/tools), [`tui`](../packages/ui/tui), [`user-approval`](../packages/ui/user-approval), [`workspace-context`](../packages/context/workspace-context) |
| `session/flush` | `parallel` | [`packages/core/session/src/index.ts:103`](../packages/core/session/src/index.ts) | [`session`](../packages/core/session) (`events.dispatch`) | [`session-persistence`](../packages/session-persistence/session-persistence), [`session-telemetry`](../packages/telemetry/session-telemetry) |
| `slash/input-begin-command` | `bail` | [`packages/client/ui-slash/src/types.ts:230`](../packages/client/ui-slash/src/types.ts) | - | `ui-conversation` |
| `slash/input-consume-token` | `bail` | [`packages/client/ui-slash/src/types.ts:244`](../packages/client/ui-slash/src/types.ts) | - | `ui-conversation` |

7
docs/module-graph.md
View File

@@ -211,6 +211,7 @@ flowchart TD
end
subgraph group_session_projection["packages/session-projection"]
pkg_session_projection["session-projection"]
pkg_session_projection_cache["session-projection-cache"]
end
subgraph group_storage["packages/storage"]
pkg_storage["storage"]
@@ -502,6 +503,11 @@ flowchart TD
pkg_pty --> pkg_invariants
pkg_scripts --> pkg_app_boot
pkg_scripts --> pkg_invariants
pkg_session_projection_cache --> pkg_invariants
pkg_session_projection_cache --> pkg_session
pkg_session_projection_cache --> pkg_session_persistence
pkg_session_projection_cache --> pkg_session_projection
pkg_session_projection_cache --> pkg_storage_domain
pkg_tasks --> pkg_agent
pkg_tasks --> pkg_brand
pkg_tasks --> pkg_invariants
@@ -1003,6 +1009,7 @@ flowchart TD
| [`time-context`](../packages/context/time-context) | `context` | [`agent`](../packages/core/agent), [`invariants`](../packages/support/invariants), [`session`](../packages/core/session) |
| [`pty`](../packages/pty/pty) | `pty` | [`agent`](../packages/core/agent), [`brand`](../packages/util/brand), [`invariants`](../packages/support/invariants) |
| [`scripts`](../packages/sdk/scripts) | `sdk` | [`app-boot`](../packages/ui/app-boot), [`invariants`](../packages/support/invariants) |
| [`session-projection-cache`](../packages/session-projection/session-projection-cache) | `session-projection` | [`invariants`](../packages/support/invariants), [`session`](../packages/core/session), [`session-persistence`](../packages/session-persistence/session-persistence), [`session-projection`](../packages/session-projection/session-projection), [`storage-domain`](../packages/storage/storage-domain) |
| [`tasks`](../packages/tasks/tasks) | `tasks` | [`agent`](../packages/core/agent), [`brand`](../packages/util/brand), [`invariants`](../packages/support/invariants), [`session`](../packages/core/session) |
| [`session-telemetry`](../packages/telemetry/session-telemetry) | `telemetry` | [`agent`](../packages/core/agent), [`invariants`](../packages/support/invariants), [`session`](../packages/core/session) |
| [`workflow`](../packages/workflow/workflow) | `workflow` | [`agent`](../packages/core/agent), [`brand`](../packages/util/brand), [`invariants`](../packages/support/invariants), [`llm`](../packages/llm/llm), [`session`](../packages/core/session) |

28
examples/acp-agent/tests/acp.snapshot.ts
View File

@@ -83,8 +83,8 @@ function snapshotModeFromEnv(value: string | undefined): SnapshotSuiteOptions['m
const SCENARIOS: Scenario[] = [
{ name: 'handshake', hasModelTurn: false, recorded: false },
{ name: 'reject-extra-dirs', hasModelTurn: false, recorded: false },
// text-turn is the pinned-header scenario: the minimal single text turn.
// Its prompt and tool-schema sidecars pin the composed header.
// text-turn is the default header pin and owns the prompt and tool-schema
// sidecars reused by alternate classes with identical component sequences.
{ name: 'text-turn', hasModelTurn: true, recorded: true, pinsHeader: true },
{
name: 'session-title-after-turn',
@@ -126,7 +126,15 @@ const SCENARIOS: Scenario[] = [
},
{ name: 'bash-tool-turn', hasModelTurn: true, recorded: true },
{ name: 'todo-write', hasModelTurn: true, recorded: true },
{ name: 'skill-load', hasModelTurn: true, recorded: false, pinsHeader: true, headerClass: 'skill' },
{
name: 'skill-load',
hasModelTurn: true,
recorded: false,
pinsHeader: true,
headerClass: 'skill',
systemPromptSource: 'text-turn',
toolSchemasSource: 'text-turn',
},
{ name: 'lsp-definition', hasModelTurn: true, recorded: false, pinsHeader: true, headerClass: 'lsp', configPath: LSP_CONFIG },
// web_fetch markdown rendering end to end: the overlay's loopback fixture
// server supplies deterministic HTML (entities, a GFM table, nesting), the
@@ -176,6 +184,7 @@ const SCENARIOS: Scenario[] = [
overridden: true,
pinsHeader: true,
headerClass: 'workspace-context',
toolSchemasSource: 'text-turn',
configPath: WORKSPACE_CONTEXT_CONFIG,
prepareWorkspace: prepareDelimiterPathWorkspace,
posixOnly: true,
@@ -250,9 +259,19 @@ const SCENARIOS: Scenario[] = [
recorded: true,
pinsHeader: true,
headerClass: 'code-workspace-context',
systemPromptSource: 'code-mode-turn',
toolSchemasSource: 'code-mode-turn',
configPath: CODE_MODE_WORKSPACE_CONTEXT_CONFIG,
},
{ name: 'both-mode-turn', hasModelTurn: true, recorded: true, pinsHeader: true, headerClass: 'both', configPath: BOTH_MODE_CONFIG },
{
name: 'both-mode-turn',
hasModelTurn: true,
recorded: true,
pinsHeader: true,
headerClass: 'both',
systemPromptSource: 'code-mode-turn',
configPath: BOTH_MODE_CONFIG,
},
// Machine permission scenarios use an explicit deployment policy; there is
// no session-scoped UI picker on the automation protocol.
{
@@ -261,6 +280,7 @@ const SCENARIOS: Scenario[] = [
recorded: true,
pinsHeader: true,
headerClass: 'sandbox',
toolSchemasSource: 'text-turn',
env: { DSH_PERMISSION_MODE: 'workspace-write' },
},
{

401
examples/acp-agent/tests/snapshots/both-mode-turn/system-prompt.expected.md
View File

@@ -1,401 +0,0 @@
You are an AI agent powered by the DeepSeek Harness SDK.
You are a coding assistant powered by the deepseek-v4-flash model. Your working directory is {{cwd}}.
Verify your work by running the code or tests. Keep answers brief and factual.
Use the read tool — not shell commands like cat — to inspect text files. Results include line numbers. Use offset and limit to continue reading large files.
Use the write tool to create files or completely replace file contents. Existing files are overwritten, so read an existing file first (the default fs-policy requires it) and prefer edit for targeted changes.
Use the edit tool for targeted changes to existing UTF-8 text files. It replaces literal old_string with new_string; by default old_string must appear exactly once. If old_string appears multiple times, provide a more specific old_string or set replace_all to true. Read the file first (the default fs-policy requires it), unless you just created or edited it in this session.
Check the [exit code: N] marker on every bash result; investigate failures before moving on.
Track every background task id you start. You are notified in-session when a task finishes — do not busy-poll or sleep on one; keep working on independent steps and do not duplicate a running task's work. Before giving a final answer, collect every still-relevant task with task_output (set wait: true only when you are genuinely blocked on it), and task_kill tasks that stopped mattering.
Use goal tools for one long-running completion objective in the current session. create_goal may infer goal intent from a direct human request in any language; do not create a goal for routine single-turn work. Call get_goal before update_goal and copy its exact goal_id and revision. After session resume or fork, an active goal is disarmed: when a human asks to continue or resume in any wording or language, use update_goal action resume to rearm it. Mark complete only when the objective is actually achieved. Mark blocked only after the same blocking condition persists for at least 3 consecutive goal rounds, and report that concrete condition in blocked_reason; difficulty, uncertainty, or useful remaining work is not blocked.
Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).
<!-- dsh-user-approval-policy:never -->
Use the workflow tool ONLY when the user explicitly asks for a workflow or for large multi-agent orchestration: you write a JavaScript script (the tool description documents the exact format) that fans work out across many subagents with phases and structured results. For one or two delegations, prefer plain subagent calls.
Use the ralph tool ONLY when the direct human explicitly asks for a Ralph loop or fresh-agent iterative execution. Each Ralph round starts a fresh child with no conversation seed and uses the shared workspace as durable memory. Completion and blockers are worker reports, not independent evaluation. Use same-session goal tools for ordinary long-running objectives, and plain subagents or workflows for bounded delegation and fan-out.
## Writing code for run_code
Pass `run_code` the body of an async TypeScript function (erasable syntax only — no `enum` or namespaces; type annotations are advisory, the code runs type-stripped). Inside the program:
- Call tools as `await tools.name(args)` — quoted access for exotic names: `tools["my-tool"](args)`. Every call resolves to the tool's typed canonical JSON value. Tool arguments must be lossless JSON.
- A FAILED tool call rejects with `ToolCallError`, whose `toolName` identifies the failed tool and whose `message` is human-readable — `try/catch` it to handle and continue.
- Independent read-only calls MAY overlap under `Promise.all` (safe calls run concurrently; mutating calls run alone, in submission order). Sequence dependent work with `await`.
- Emit results with `return` and/or `console.log(...)`. ONLY what you print or return comes back to you — intermediate tool results never enter the conversation, so extract just what you need.
The available tools:
```ts
type JsonValue = null | boolean | number | string | JsonValue[] | { [key: string]: JsonValue }
interface ToolArgsMap {
/** Execute a bash command (`bash -c`) and return its stdout/stderr. Each call runs in a fresh shell: no state (cwd, variables, functions) persists between calls — pass `workdir` instead of using `cd`. Non-zero exits are reported as `[exit code: N]`. Current harness environment facts are exposed through managed `$DSH_*` variables; inspect them when needed. Commands may run under a file sandbox; a blocked file operation is reported as `[sandbox: file access denied under <mode> mode]` — a policy denial, not a bug in the command; do not retry another way. Long output is truncated to its tail; the full output is saved to a file whose path is reported when available. Set `run_in_background: true` for long-running commands: the call returns a task id immediately; read its output with `task_output` and stop it with `task_kill`. Attempting a command the sandbox may deny is safe and expected: run it and read the marker rather than assuming the denial. When a command is denied and a wider mode would let it succeed, escalate immediately in the same turn — the one sanctioned exception to a denial: retry the exact same command once with `sandbox_permissions` (the narrowest wider mode that suffices) plus a one-sentence `justification`. Do not detour through chat to ask permission first — the approval prompt raised by that retry is how the user consents. If the session states approval prompts are disabled, there is no exception: a denial is final — do not set `sandbox_permissions`. Never escalate speculatively: ground the request in a real denial — normally the one this command just hit; escalating up front is fine only when this session already denied the same access. A rejected escalation is final for that command — stop and explain, never work around it — but it does not forbid attempting or escalating other commands later. */
bash: {
/** The bash command to execute. */
command: string;
/** Clear, concise description of what this command does in active voice, 5-10 words (shown in the UI). Examples: "ls" → "List files in current directory"; "git status" → "Show working tree status"; "npm install" → "Install package dependencies". */
description: string;
/** Timeout in milliseconds. The executor applies its configured default and cap, and kills the command on expiry. */
timeoutMs?: number;
/** Working directory for this command. Defaults to the session workspace; a relative path is resolved against it. */
workdir?: string;
/** Run in the background and return a task id immediately (collect with task_output, stop with task_kill). No timeout applies. */
run_in_background?: boolean;
/** The wider sandbox mode this command needs. Only valid as a one-shot retry of a command the sandbox just denied; requires justification and user approval. */
sandbox_permissions?: "workspace-write" | "danger-full-access";
/** Required with sandbox_permissions: one sentence for the user explaining why this exact command needs the wider access. */
justification?: string;
} & Record<string, JsonValue>;
/** Create one persisted same-session completion goal when the current direct human request is a long-running objective that should continue across autonomous goal rounds. You may infer that intent without requiring the user to say "create a goal". Do not use this for trivial single-turn work. Execution rejects non-human and subagent authority. */
create_goal: {
/** The concrete completion objective inferred from the direct human request. */
objective: string;
/** Optional positive safe-integer limit on automatic continuation rounds. */
max_goal_rounds?: number;
} & Record<string, JsonValue>;
/** Edit an existing UTF-8 text file by replacing literal text. */
edit: {
/** Path to edit, resolved by the filesystem backend. */
file_path: string;
/** Literal text to replace. Must match exactly. */
old_string: string;
/** Literal replacement text. Use an empty string to delete the match. */
new_string: string;
/** Replace all matches. Defaults to false; when false, old_string must appear exactly once. */
replace_all?: boolean;
/** The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval. */
sandbox_permissions?: "workspace-write" | "danger-full-access";
/** Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access. */
justification?: string;
} & Record<string, JsonValue>;
/** Read the current same-session goal, including its exact id/revision, objective, phase, completed continuation rounds, round limit, blocker reason when present, and whether another continuation is armed. Call this before updating a goal. */
get_goal: Record<string, JsonValue>;
/** Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools. */
ralph: {
/** The immutable completion objective for every fresh Ralph round. */
objective: string;
/** Optional positive safe-integer round cap, bounded by the deployment ceiling. */
maxRounds?: number;
} & Record<string, JsonValue>;
/** Read a UTF-8 text file and return line-numbered content. */
read: {
/** Path to read, resolved by the filesystem backend. */
file_path: string;
/** 1-based first line to return. Defaults to 1. */
offset?: number;
/** Maximum number of lines to return. Defaults to 2000. */
limit?: number;
} & Record<string, JsonValue>;
/** Load the full instructions for an available skill. Call this with the exact skill name from the session skill catalog before acting on a task that names or clearly matches that skill. */
skill: {
/** The exact skill name from the available skills list. */
name: string;
} & Record<string, JsonValue>;
/** Delegate a self-contained task to a subagent (a separate agent that works in its own context) and return its final result. Use this to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent runs to completion and you receive only its final answer, not its intermediate steps. Give it a complete, standalone prompt: it does not see this conversation. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`. */
subagent: {
/** A short (3-5 word) description of the delegated task, for display. */
description: string;
/** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */
prompt: string;
/** Run as a background task and return its id; collect with task_output or stop with task_kill. */
run_in_background?: boolean;
} & Record<string, JsonValue>;
/** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn), returning only its final result. Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive only its final answer, not its intermediate steps. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`. */
subagent_fork: {
/** A short (3-5 word) description of the delegated task, for display. */
description: string;
/** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */
prompt: string;
/** Run as a background task and return its id; collect with task_output or stop with task_kill. */
run_in_background?: boolean;
} & Record<string, JsonValue>;
/** Request cancellation of a running background task by task id. Returns immediately; the task settles as killed once its work actually stops. */
task_kill: {
/** Task id returned by the tool that started the background work. */
task_id: string;
/** Optional short reason, recorded in the log and forwarded to the task. */
reason?: string;
} & Record<string, JsonValue>;
/** List your background tasks (running and finished) with their ids, kinds, and statuses. */
task_list: Record<string, JsonValue>;
/** Read a background task. Stream tasks return only output since the previous read; final-output tasks return their result after settlement. Every response ends with `[status: ...]`. Reads are non-blocking unless `wait: true`, which waits up to the configured cap. */
task_output: {
/** Task id returned by the tool that started the background work. */
task_id: string;
/** Block until the task reaches a terminal status or the timeout expires. A timed-out wait returns [status: running] and leaves the task alive. */
wait?: boolean;
/** Max wait in milliseconds (only meaningful with wait: true). Defaults to the configured wait timeout; capped by the configured maximum. */
timeout_ms?: number;
} & Record<string, JsonValue>;
/** Record and update a structured task list for the current work. Send the ENTIRE list every call — it REPLACES the previous list (there are no partial updates, no per-item edits). Use it to plan multi-step work and show progress: add one todo per concrete step before you start. Keep AT MOST ONE todo `in_progress` at a time; while work remains, exactly one active task should be `in_progress`. Mark a todo `completed` the moment it is done (do not batch completions), and allow no `in_progress` item only once all work is complete. Skip the list for trivial single-step tasks. Statuses: `pending` (not started), `in_progress` (being worked on now), `completed` (finished). */
todo_write: {
/** The COMPLETE task list, replacing any previous list. */
todos: ({
/** What the task is — a short imperative line. */
content: string;
/** pending (not started) | in_progress (now) | completed (done). */
status: "pending" | "in_progress" | "completed";
})[];
} & Record<string, JsonValue>;
/** Update the exact current goal revision. edit, pause, and resume require a direct top-level human request. During an automatic continuation of the current goal, complete and blocked are also allowed. blocked is rejected before the configured minimum round count; the model remains responsible for judging that the same condition persisted across those rounds and must explain it in blocked_reason. */
update_goal: {
/** Exact id returned by get_goal. */
goal_id: string;
/** Exact positive revision returned by get_goal. */
revision: number;
/** edit | pause | resume | complete | blocked */
action: "edit" | "pause" | "resume" | "complete" | "blocked";
/** Replacement objective; valid only with action edit. */
objective?: string;
/** Replacement cap; valid only with action edit. */
max_goal_rounds?: number;
/** Concrete blocking condition; required only with action blocked. */
blocked_reason?: string;
} & Record<string, JsonValue>;
/** Run a JavaScript workflow script that orchestrates subagents at scale. Use this for work that fans out across many independent pieces — an audit over many files, a migration, multi-angle research, adversarial verification of findings — where you write the orchestration as a script instead of delegating turn by turn. The workflow's identity rides the `meta` parameter as JSON: required `name` (short kebab-case) and `description` strings, optional `whenToUse` string and `phases` array (`{title, detail?, provider?, model?}`). The `script` parameter is the plain JavaScript body ONLY (NOT TypeScript, and NO `export const meta` statement — meta is a parameter, not code), running with top-level await; end with `return <value>` — the value must be JSON-serializable and is this tool's result. Script-body hooks: - `agent(prompt, opts?): Promise<any>` — run one subagent to completion. Without `opts.schema` it resolves to the child's final text; with `opts.schema` (an object-rooted JSON Schema using ONLY type/properties/required/additionalProperties/items/enum/const/oneOf — no pattern/format/numeric bounds) it resolves to the validated object. Resolves `null` when the child fails (filter with `.filter(Boolean)`). Other opts: `label` (display), `phase` (progress group), and independent `provider`/`model` LLM target overrides (either may be provided alone). Anything else (`effort`/`isolation`/`agentType`) is rejected loudly. - `pipeline(items, ...stages): Promise<any[]>` — run each item through the stages independently with NO barrier between stages (prefer this for multi-stage work). Each stage receives `(prev, item, index)`. An ordinary stage throw drops that ITEM to `null` and skips its remaining stages. - `parallel(thunks): Promise<any[]>` — run zero-argument functions concurrently and await ALL of them (a barrier; use only when a stage genuinely needs every prior result together). A throwing thunk resolves to `null`. - `phase(title)` — start a progress phase; `log(message)` — narrate progress; `args` — the tool call's `args` input, verbatim. Misused hooks (bad arguments, unknown options, unsupported schemas, tripped caps) throw errors that ALWAYS kill the script — they never dissolve into a per-item `null`. Constraints: concurrency and total-agent caps apply; no filesystem, network, timers, or Node.js APIs are provided — the agents do the work, the script only coordinates them. The run executes in the foreground: this call returns when the whole script finishes. */
workflow: {
/** The plain-JS workflow script body (top-level await allowed; NO `export const meta` statement; end with `return <json-value>`). */
script: string;
/** The workflow identity block (plain JSON — never code). */
meta: {
/** Short kebab-case workflow name. */
name: string;
/** One-line description of what the workflow does. */
description: string;
/** Optional guidance on when this workflow applies. */
whenToUse?: string;
/** Optional phase declarations matched by phase() calls. */
phases?: ({
/** The phase title phase() calls match by exact string. */
title: string;
/** Optional one-line description of the phase. */
detail?: string;
/** Optional provider override this phase is expected to use. */
provider?: string;
/** Optional model override this phase is expected to use. */
model?: string;
} & Record<string, JsonValue>)[];
} & Record<string, JsonValue>;
/** Optional JSON input exposed to the script as the `args` global (wrap a bare list as a field, e.g. {"files": [...]}). */
args?: Record<string, JsonValue>;
} & Record<string, JsonValue>;
/** Create or fully replace a UTF-8 text file. */
write: {
/** Path to write, resolved by the filesystem backend. */
file_path: string;
/** Full UTF-8 text content to write. */
content: string;
/** The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval. */
sandbox_permissions?: "workspace-write" | "danger-full-access";
/** Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access. */
justification?: string;
} & Record<string, JsonValue>;
}
interface ToolOutputMap {
bash: {
kind: "background";
taskId: string;
} | {
kind: "foreground";
exitCode: number | null;
signal: string | null;
timedOut: boolean;
aborted: boolean;
timeoutMs: number;
stdout: {
text: string;
truncated: boolean;
spillPath?: string;
};
stderr: {
text: string;
truncated: boolean;
spillPath?: string;
};
sandbox?: {
mode: string;
denied: boolean;
enforcement?: string;
runnerFailed?: boolean;
};
};
create_goal: {
goal: null;
} | {
goal: {
id: string;
revision: number;
objective: string;
phase: "active" | "paused" | "blocked" | "complete";
roundsStarted: number;
maxGoalRounds: number;
blockedReason?: {
code: string;
message: string;
};
};
activation: "armed" | "disarmed";
};
edit: {
path: string;
before: string;
after: string;
};
get_goal: {
goal: null;
} | {
goal: {
id: string;
revision: number;
objective: string;
phase: "active" | "paused" | "blocked" | "complete";
roundsStarted: number;
maxGoalRounds: number;
blockedReason?: {
code: string;
message: string;
};
};
activation: "armed" | "disarmed";
};
ralph: {
runId: string;
agentsStarted: number;
result: JsonValue;
};
read: {
path: string;
offset: number;
lines: {
number: number;
text: string;
}[];
totalLines: number;
};
skill: {
name: string;
provider: string;
resourceBase?: {
kind: "directory";
path: string;
} | {
kind: "url";
url: string;
} | {
kind: "opaque";
description: string;
};
content: string;
};
subagent: {
kind: "background";
taskId: string;
} | {
kind: "foreground";
runId: string;
output: JsonValue[];
};
subagent_fork: {
kind: "background";
taskId: string;
} | {
kind: "foreground";
runId: string;
output: JsonValue[];
};
task_kill: {
outcome: "cancellation-requested" | "already-finished";
task: {
id: string;
kind: string;
label: string;
status: "running" | "stopping" | "completed" | "killed" | "failed";
detail?: string;
startedAt: number;
finishedAt?: number;
};
};
task_list: ({
id: string;
kind: string;
label: string;
status: "running" | "stopping" | "completed" | "killed" | "failed";
detail?: string;
startedAt: number;
finishedAt?: number;
})[];
task_output: {
text: string;
task: {
id: string;
kind: string;
label: string;
status: "running" | "stopping" | "completed" | "killed" | "failed";
detail?: string;
startedAt: number;
finishedAt?: number;
};
};
todo_write: {
todos: ({
content: string;
status: "pending" | "in_progress" | "completed";
})[];
counts: {
pending: number;
inProgress: number;
completed: number;
};
};
update_goal: {
goal: null;
} | {
goal: {
id: string;
revision: number;
objective: string;
phase: "active" | "paused" | "blocked" | "complete";
roundsStarted: number;
maxGoalRounds: number;
blockedReason?: {
code: string;
message: string;
};
};
activation: "armed" | "disarmed";
};
workflow: {
runId: string;
agentsStarted: number;
result: JsonValue;
};
write: {
path: string;
operation: "create" | "update";
before: string | null;
after: string;
};
}
type ToolName = keyof ToolOutputMap
declare class ToolCallError extends Error {
readonly name: "ToolCallError";
readonly toolName: ToolName;
}
declare const tools: {
[K in ToolName]: (args: ToolArgsMap[K]) => Promise<ToolOutputMap[K]>;
}
```

401
examples/acp-agent/tests/snapshots/code-mode-workspace-context/system-prompt.expected.md
View File

@@ -1,401 +0,0 @@
You are an AI agent powered by the DeepSeek Harness SDK.
You are a coding assistant powered by the deepseek-v4-flash model. Your working directory is {{cwd}}.
Verify your work by running the code or tests. Keep answers brief and factual.
Use the read tool — not shell commands like cat — to inspect text files. Results include line numbers. Use offset and limit to continue reading large files.
Use the write tool to create files or completely replace file contents. Existing files are overwritten, so read an existing file first (the default fs-policy requires it) and prefer edit for targeted changes.
Use the edit tool for targeted changes to existing UTF-8 text files. It replaces literal old_string with new_string; by default old_string must appear exactly once. If old_string appears multiple times, provide a more specific old_string or set replace_all to true. Read the file first (the default fs-policy requires it), unless you just created or edited it in this session.
Check the [exit code: N] marker on every bash result; investigate failures before moving on.
Track every background task id you start. You are notified in-session when a task finishes — do not busy-poll or sleep on one; keep working on independent steps and do not duplicate a running task's work. Before giving a final answer, collect every still-relevant task with task_output (set wait: true only when you are genuinely blocked on it), and task_kill tasks that stopped mattering.
Use goal tools for one long-running completion objective in the current session. create_goal may infer goal intent from a direct human request in any language; do not create a goal for routine single-turn work. Call get_goal before update_goal and copy its exact goal_id and revision. After session resume or fork, an active goal is disarmed: when a human asks to continue or resume in any wording or language, use update_goal action resume to rearm it. Mark complete only when the objective is actually achieved. Mark blocked only after the same blocking condition persists for at least 3 consecutive goal rounds, and report that concrete condition in blocked_reason; difficulty, uncertainty, or useful remaining work is not blocked.
Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).
<!-- dsh-user-approval-policy:never -->
Use the workflow tool ONLY when the user explicitly asks for a workflow or for large multi-agent orchestration: you write a JavaScript script (the tool description documents the exact format) that fans work out across many subagents with phases and structured results. For one or two delegations, prefer plain subagent calls.
Use the ralph tool ONLY when the direct human explicitly asks for a Ralph loop or fresh-agent iterative execution. Each Ralph round starts a fresh child with no conversation seed and uses the shared workspace as durable memory. Completion and blockers are worker reports, not independent evaluation. Use same-session goal tools for ordinary long-running objectives, and plain subagents or workflows for bounded delegation and fan-out.
## Writing code for run_code
Pass `run_code` the body of an async TypeScript function (erasable syntax only — no `enum` or namespaces; type annotations are advisory, the code runs type-stripped). Inside the program:
- Call tools as `await tools.name(args)` — quoted access for exotic names: `tools["my-tool"](args)`. Every call resolves to the tool's typed canonical JSON value. Tool arguments must be lossless JSON.
- A FAILED tool call rejects with `ToolCallError`, whose `toolName` identifies the failed tool and whose `message` is human-readable — `try/catch` it to handle and continue.
- Independent read-only calls MAY overlap under `Promise.all` (safe calls run concurrently; mutating calls run alone, in submission order). Sequence dependent work with `await`.
- Emit results with `return` and/or `console.log(...)`. ONLY what you print or return comes back to you — intermediate tool results never enter the conversation, so extract just what you need.
The available tools:
```ts
type JsonValue = null | boolean | number | string | JsonValue[] | { [key: string]: JsonValue }
interface ToolArgsMap {
/** Execute a bash command (`bash -c`) and return its stdout/stderr. Each call runs in a fresh shell: no state (cwd, variables, functions) persists between calls — pass `workdir` instead of using `cd`. Non-zero exits are reported as `[exit code: N]`. Current harness environment facts are exposed through managed `$DSH_*` variables; inspect them when needed. Commands may run under a file sandbox; a blocked file operation is reported as `[sandbox: file access denied under <mode> mode]` — a policy denial, not a bug in the command; do not retry another way. Long output is truncated to its tail; the full output is saved to a file whose path is reported when available. Set `run_in_background: true` for long-running commands: the call returns a task id immediately; read its output with `task_output` and stop it with `task_kill`. Attempting a command the sandbox may deny is safe and expected: run it and read the marker rather than assuming the denial. When a command is denied and a wider mode would let it succeed, escalate immediately in the same turn — the one sanctioned exception to a denial: retry the exact same command once with `sandbox_permissions` (the narrowest wider mode that suffices) plus a one-sentence `justification`. Do not detour through chat to ask permission first — the approval prompt raised by that retry is how the user consents. If the session states approval prompts are disabled, there is no exception: a denial is final — do not set `sandbox_permissions`. Never escalate speculatively: ground the request in a real denial — normally the one this command just hit; escalating up front is fine only when this session already denied the same access. A rejected escalation is final for that command — stop and explain, never work around it — but it does not forbid attempting or escalating other commands later. */
bash: {
/** The bash command to execute. */
command: string;
/** Clear, concise description of what this command does in active voice, 5-10 words (shown in the UI). Examples: "ls" → "List files in current directory"; "git status" → "Show working tree status"; "npm install" → "Install package dependencies". */
description: string;
/** Timeout in milliseconds. The executor applies its configured default and cap, and kills the command on expiry. */
timeoutMs?: number;
/** Working directory for this command. Defaults to the session workspace; a relative path is resolved against it. */
workdir?: string;
/** Run in the background and return a task id immediately (collect with task_output, stop with task_kill). No timeout applies. */
run_in_background?: boolean;
/** The wider sandbox mode this command needs. Only valid as a one-shot retry of a command the sandbox just denied; requires justification and user approval. */
sandbox_permissions?: "workspace-write" | "danger-full-access";
/** Required with sandbox_permissions: one sentence for the user explaining why this exact command needs the wider access. */
justification?: string;
} & Record<string, JsonValue>;
/** Create one persisted same-session completion goal when the current direct human request is a long-running objective that should continue across autonomous goal rounds. You may infer that intent without requiring the user to say "create a goal". Do not use this for trivial single-turn work. Execution rejects non-human and subagent authority. */
create_goal: {
/** The concrete completion objective inferred from the direct human request. */
objective: string;
/** Optional positive safe-integer limit on automatic continuation rounds. */
max_goal_rounds?: number;
} & Record<string, JsonValue>;
/** Edit an existing UTF-8 text file by replacing literal text. */
edit: {
/** Path to edit, resolved by the filesystem backend. */
file_path: string;
/** Literal text to replace. Must match exactly. */
old_string: string;
/** Literal replacement text. Use an empty string to delete the match. */
new_string: string;
/** Replace all matches. Defaults to false; when false, old_string must appear exactly once. */
replace_all?: boolean;
/** The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval. */
sandbox_permissions?: "workspace-write" | "danger-full-access";
/** Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access. */
justification?: string;
} & Record<string, JsonValue>;
/** Read the current same-session goal, including its exact id/revision, objective, phase, completed continuation rounds, round limit, blocker reason when present, and whether another continuation is armed. Call this before updating a goal. */
get_goal: Record<string, JsonValue>;
/** Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools. */
ralph: {
/** The immutable completion objective for every fresh Ralph round. */
objective: string;
/** Optional positive safe-integer round cap, bounded by the deployment ceiling. */
maxRounds?: number;
} & Record<string, JsonValue>;
/** Read a UTF-8 text file and return line-numbered content. */
read: {
/** Path to read, resolved by the filesystem backend. */
file_path: string;
/** 1-based first line to return. Defaults to 1. */
offset?: number;
/** Maximum number of lines to return. Defaults to 2000. */
limit?: number;
} & Record<string, JsonValue>;
/** Load the full instructions for an available skill. Call this with the exact skill name from the session skill catalog before acting on a task that names or clearly matches that skill. */
skill: {
/** The exact skill name from the available skills list. */
name: string;
} & Record<string, JsonValue>;
/** Delegate a self-contained task to a subagent (a separate agent that works in its own context) and return its final result. Use this to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent runs to completion and you receive only its final answer, not its intermediate steps. Give it a complete, standalone prompt: it does not see this conversation. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`. */
subagent: {
/** A short (3-5 word) description of the delegated task, for display. */
description: string;
/** The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs. */
prompt: string;
/** Run as a background task and return its id; collect with task_output or stop with task_kill. */
run_in_background?: boolean;
} & Record<string, JsonValue>;
/** Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn), returning only its final result. Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive only its final answer, not its intermediate steps. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`. */
subagent_fork: {
/** A short (3-5 word) description of the delegated task, for display. */
description: string;
/** The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new. */
prompt: string;
/** Run as a background task and return its id; collect with task_output or stop with task_kill. */
run_in_background?: boolean;
} & Record<string, JsonValue>;
/** Request cancellation of a running background task by task id. Returns immediately; the task settles as killed once its work actually stops. */
task_kill: {
/** Task id returned by the tool that started the background work. */
task_id: string;
/** Optional short reason, recorded in the log and forwarded to the task. */
reason?: string;
} & Record<string, JsonValue>;
/** List your background tasks (running and finished) with their ids, kinds, and statuses. */
task_list: Record<string, JsonValue>;
/** Read a background task. Stream tasks return only output since the previous read; final-output tasks return their result after settlement. Every response ends with `[status: ...]`. Reads are non-blocking unless `wait: true`, which waits up to the configured cap. */
task_output: {
/** Task id returned by the tool that started the background work. */
task_id: string;
/** Block until the task reaches a terminal status or the timeout expires. A timed-out wait returns [status: running] and leaves the task alive. */
wait?: boolean;
/** Max wait in milliseconds (only meaningful with wait: true). Defaults to the configured wait timeout; capped by the configured maximum. */
timeout_ms?: number;
} & Record<string, JsonValue>;
/** Record and update a structured task list for the current work. Send the ENTIRE list every call — it REPLACES the previous list (there are no partial updates, no per-item edits). Use it to plan multi-step work and show progress: add one todo per concrete step before you start. Keep AT MOST ONE todo `in_progress` at a time; while work remains, exactly one active task should be `in_progress`. Mark a todo `completed` the moment it is done (do not batch completions), and allow no `in_progress` item only once all work is complete. Skip the list for trivial single-step tasks. Statuses: `pending` (not started), `in_progress` (being worked on now), `completed` (finished). */
todo_write: {
/** The COMPLETE task list, replacing any previous list. */
todos: ({
/** What the task is — a short imperative line. */
content: string;
/** pending (not started) | in_progress (now) | completed (done). */
status: "pending" | "in_progress" | "completed";
})[];
} & Record<string, JsonValue>;
/** Update the exact current goal revision. edit, pause, and resume require a direct top-level human request. During an automatic continuation of the current goal, complete and blocked are also allowed. blocked is rejected before the configured minimum round count; the model remains responsible for judging that the same condition persisted across those rounds and must explain it in blocked_reason. */
update_goal: {
/** Exact id returned by get_goal. */
goal_id: string;
/** Exact positive revision returned by get_goal. */
revision: number;
/** edit | pause | resume | complete | blocked */
action: "edit" | "pause" | "resume" | "complete" | "blocked";
/** Replacement objective; valid only with action edit. */
objective?: string;
/** Replacement cap; valid only with action edit. */
max_goal_rounds?: number;
/** Concrete blocking condition; required only with action blocked. */
blocked_reason?: string;
} & Record<string, JsonValue>;
/** Run a JavaScript workflow script that orchestrates subagents at scale. Use this for work that fans out across many independent pieces — an audit over many files, a migration, multi-angle research, adversarial verification of findings — where you write the orchestration as a script instead of delegating turn by turn. The workflow's identity rides the `meta` parameter as JSON: required `name` (short kebab-case) and `description` strings, optional `whenToUse` string and `phases` array (`{title, detail?, provider?, model?}`). The `script` parameter is the plain JavaScript body ONLY (NOT TypeScript, and NO `export const meta` statement — meta is a parameter, not code), running with top-level await; end with `return <value>` — the value must be JSON-serializable and is this tool's result. Script-body hooks: - `agent(prompt, opts?): Promise<any>` — run one subagent to completion. Without `opts.schema` it resolves to the child's final text; with `opts.schema` (an object-rooted JSON Schema using ONLY type/properties/required/additionalProperties/items/enum/const/oneOf — no pattern/format/numeric bounds) it resolves to the validated object. Resolves `null` when the child fails (filter with `.filter(Boolean)`). Other opts: `label` (display), `phase` (progress group), and independent `provider`/`model` LLM target overrides (either may be provided alone). Anything else (`effort`/`isolation`/`agentType`) is rejected loudly. - `pipeline(items, ...stages): Promise<any[]>` — run each item through the stages independently with NO barrier between stages (prefer this for multi-stage work). Each stage receives `(prev, item, index)`. An ordinary stage throw drops that ITEM to `null` and skips its remaining stages. - `parallel(thunks): Promise<any[]>` — run zero-argument functions concurrently and await ALL of them (a barrier; use only when a stage genuinely needs every prior result together). A throwing thunk resolves to `null`. - `phase(title)` — start a progress phase; `log(message)` — narrate progress; `args` — the tool call's `args` input, verbatim. Misused hooks (bad arguments, unknown options, unsupported schemas, tripped caps) throw errors that ALWAYS kill the script — they never dissolve into a per-item `null`. Constraints: concurrency and total-agent caps apply; no filesystem, network, timers, or Node.js APIs are provided — the agents do the work, the script only coordinates them. The run executes in the foreground: this call returns when the whole script finishes. */
workflow: {
/** The plain-JS workflow script body (top-level await allowed; NO `export const meta` statement; end with `return <json-value>`). */
script: string;
/** The workflow identity block (plain JSON — never code). */
meta: {
/** Short kebab-case workflow name. */
name: string;
/** One-line description of what the workflow does. */
description: string;
/** Optional guidance on when this workflow applies. */
whenToUse?: string;
/** Optional phase declarations matched by phase() calls. */
phases?: ({
/** The phase title phase() calls match by exact string. */
title: string;
/** Optional one-line description of the phase. */
detail?: string;
/** Optional provider override this phase is expected to use. */
provider?: string;
/** Optional model override this phase is expected to use. */
model?: string;
} & Record<string, JsonValue>)[];
} & Record<string, JsonValue>;
/** Optional JSON input exposed to the script as the `args` global (wrap a bare list as a field, e.g. {"files": [...]}). */
args?: Record<string, JsonValue>;
} & Record<string, JsonValue>;
/** Create or fully replace a UTF-8 text file. */
write: {
/** Path to write, resolved by the filesystem backend. */
file_path: string;
/** Full UTF-8 text content to write. */
content: string;
/** The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval. */
sandbox_permissions?: "workspace-write" | "danger-full-access";
/** Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access. */
justification?: string;
} & Record<string, JsonValue>;
}
interface ToolOutputMap {
bash: {
kind: "background";
taskId: string;
} | {
kind: "foreground";
exitCode: number | null;
signal: string | null;
timedOut: boolean;
aborted: boolean;
timeoutMs: number;
stdout: {
text: string;
truncated: boolean;
spillPath?: string;
};
stderr: {
text: string;
truncated: boolean;
spillPath?: string;
};
sandbox?: {
mode: string;
denied: boolean;
enforcement?: string;
runnerFailed?: boolean;
};
};
create_goal: {
goal: null;
} | {
goal: {
id: string;
revision: number;
objective: string;
phase: "active" | "paused" | "blocked" | "complete";
roundsStarted: number;
maxGoalRounds: number;
blockedReason?: {
code: string;
message: string;
};
};
activation: "armed" | "disarmed";
};
edit: {
path: string;
before: string;
after: string;
};
get_goal: {
goal: null;
} | {
goal: {
id: string;
revision: number;
objective: string;
phase: "active" | "paused" | "blocked" | "complete";
roundsStarted: number;
maxGoalRounds: number;
blockedReason?: {
code: string;
message: string;
};
};
activation: "armed" | "disarmed";
};
ralph: {
runId: string;
agentsStarted: number;
result: JsonValue;
};
read: {
path: string;
offset: number;
lines: {
number: number;
text: string;
}[];
totalLines: number;
};
skill: {
name: string;
provider: string;
resourceBase?: {
kind: "directory";
path: string;
} | {
kind: "url";
url: string;
} | {
kind: "opaque";
description: string;
};
content: string;
};
subagent: {
kind: "background";
taskId: string;
} | {
kind: "foreground";
runId: string;
output: JsonValue[];
};
subagent_fork: {
kind: "background";
taskId: string;
} | {
kind: "foreground";
runId: string;
output: JsonValue[];
};
task_kill: {
outcome: "cancellation-requested" | "already-finished";
task: {
id: string;
kind: string;
label: string;
status: "running" | "stopping" | "completed" | "killed" | "failed";
detail?: string;
startedAt: number;
finishedAt?: number;
};
};
task_list: ({
id: string;
kind: string;
label: string;
status: "running" | "stopping" | "completed" | "killed" | "failed";
detail?: string;
startedAt: number;
finishedAt?: number;
})[];
task_output: {
text: string;
task: {
id: string;
kind: string;
label: string;
status: "running" | "stopping" | "completed" | "killed" | "failed";
detail?: string;
startedAt: number;
finishedAt?: number;
};
};
todo_write: {
todos: ({
content: string;
status: "pending" | "in_progress" | "completed";
})[];
counts: {
pending: number;
inProgress: number;
completed: number;
};
};
update_goal: {
goal: null;
} | {
goal: {
id: string;
revision: number;
objective: string;
phase: "active" | "paused" | "blocked" | "complete";
roundsStarted: number;
maxGoalRounds: number;
blockedReason?: {
code: string;
message: string;
};
};
activation: "armed" | "disarmed";
};
workflow: {
runId: string;
agentsStarted: number;
result: JsonValue;
};
write: {
path: string;
operation: "create" | "update";
before: string | null;
after: string;
};
}
type ToolName = keyof ToolOutputMap
declare class ToolCallError extends Error {
readonly name: "ToolCallError";
readonly toolName: ToolName;
}
declare const tools: {
[K in ToolName]: (args: ToolArgsMap[K]) => Promise<ToolOutputMap[K]>;
}
```

26
examples/acp-agent/tests/snapshots/code-mode-workspace-context/tool-schemas.expected.json
View File

@@ -1,26 +0,0 @@
{
"initial": [
{
"name": "run_code",
"description": "Execute a TypeScript program against the available tools. Write the BODY of an async function (erasable syntax only; top-level `await` and `return` work) and call tools as `await tools.name(args)` per the declarations in the system prompt. Only what you print or return comes back — curate it.",
"parameters": {
"type": "object",
"properties": {
"code": {
"type": "string",
"description": "The program: the body of an async TypeScript function."
},
"description": {
"type": "string",
"description": "Clear, concise description of what this program does in active voice, 5-10 words (shown in the UI). Examples: \"Count TODO markers across packages\"; \"Read failing test and its fixture\"; \"Rename config key in every cordis.yml\"."
}
},
"required": [
"code",
"description"
]
}
}
],
"changes": []
}

473
examples/acp-agent/tests/snapshots/escalation-approved/tool-schemas.expected.json
View File

@@ -1,473 +0,0 @@
{
"initial": [
{
"name": "bash",
"description": "Execute a bash command (`bash -c`) and return its stdout/stderr. Each call runs in a fresh shell: no state (cwd, variables, functions) persists between calls — pass `workdir` instead of using `cd`. Non-zero exits are reported as `[exit code: N]`. Current harness environment facts are exposed through managed `$DSH_*` variables; inspect them when needed. Commands may run under a file sandbox; a blocked file operation is reported as `[sandbox: file access denied under <mode> mode]` — a policy denial, not a bug in the command; do not retry another way. Long output is truncated to its tail; the full output is saved to a file whose path is reported when available. Set `run_in_background: true` for long-running commands: the call returns a task id immediately; read its output with `task_output` and stop it with `task_kill`. Attempting a command the sandbox may deny is safe and expected: run it and read the marker rather than assuming the denial. When a command is denied and a wider mode would let it succeed, escalate immediately in the same turn — the one sanctioned exception to a denial: retry the exact same command once with `sandbox_permissions` (the narrowest wider mode that suffices) plus a one-sentence `justification`. Do not detour through chat to ask permission first — the approval prompt raised by that retry is how the user consents. If the session states approval prompts are disabled, there is no exception: a denial is final — do not set `sandbox_permissions`. Never escalate speculatively: ground the request in a real denial — normally the one this command just hit; escalating up front is fine only when this session already denied the same access. A rejected escalation is final for that command — stop and explain, never work around it — but it does not forbid attempting or escalating other commands later.",
"parameters": {
"type": "object",
"properties": {
"command": {
"type": "string",
"description": "The bash command to execute."
},
"description": {
"type": "string",
"description": "Clear, concise description of what this command does in active voice, 5-10 words (shown in the UI). Examples: \"ls\" → \"List files in current directory\"; \"git status\" → \"Show working tree status\"; \"npm install\" → \"Install package dependencies\"."
},
"timeoutMs": {
"type": "number",
"description": "Timeout in milliseconds. The executor applies its configured default and cap, and kills the command on expiry."
},
"workdir": {
"type": "string",
"description": "Working directory for this command. Defaults to the session workspace; a relative path is resolved against it."
},
"run_in_background": {
"type": "boolean",
"description": "Run in the background and return a task id immediately (collect with task_output, stop with task_kill). No timeout applies."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this command needs. Only valid as a one-shot retry of a command the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact command needs the wider access."
}
},
"required": [
"command",
"description"
]
}
},
{
"name": "create_goal",
"description": "Create one persisted same-session completion goal when the current direct human request is a long-running objective that should continue across autonomous goal rounds. You may infer that intent without requiring the user to say \"create a goal\". Do not use this for trivial single-turn work. Execution rejects non-human and subagent authority.",
"parameters": {
"type": "object",
"properties": {
"objective": {
"type": "string",
"description": "The concrete completion objective inferred from the direct human request."
},
"max_goal_rounds": {
"type": "number",
"description": "Optional positive safe-integer limit on automatic continuation rounds."
}
},
"required": [
"objective"
]
}
},
{
"name": "edit",
"description": "Edit an existing UTF-8 text file by replacing literal text.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to edit, resolved by the filesystem backend."
},
"old_string": {
"type": "string",
"description": "Literal text to replace. Must match exactly."
},
"new_string": {
"type": "string",
"description": "Literal replacement text. Use an empty string to delete the match."
},
"replace_all": {
"type": "boolean",
"description": "Replace all matches. Defaults to false; when false, old_string must appear exactly once."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access."
}
},
"required": [
"file_path",
"old_string",
"new_string"
]
}
},
{
"name": "get_goal",
"description": "Read the current same-session goal, including its exact id/revision, objective, phase, completed continuation rounds, round limit, blocker reason when present, and whether another continuation is armed. Call this before updating a goal.",
"parameters": {
"type": "object",
"properties": {}
}
},
{
"name": "ralph",
"description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.",
"parameters": {
"type": "object",
"properties": {
"objective": {
"type": "string",
"description": "The immutable completion objective for every fresh Ralph round."
},
"maxRounds": {
"type": "number",
"description": "Optional positive safe-integer round cap, bounded by the deployment ceiling."
}
},
"required": [
"objective"
]
}
},
{
"name": "read",
"description": "Read a UTF-8 text file and return line-numbered content.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to read, resolved by the filesystem backend."
},
"offset": {
"type": "number",
"description": "1-based first line to return. Defaults to 1."
},
"limit": {
"type": "number",
"description": "Maximum number of lines to return. Defaults to 2000."
}
},
"required": [
"file_path"
]
}
},
{
"name": "skill",
"description": "Load the full instructions for an available skill. Call this with the exact skill name from the session skill catalog before acting on a task that names or clearly matches that skill.",
"parameters": {
"type": "object",
"properties": {
"name": {
"type": "string",
"description": "The exact skill name from the available skills list."
}
},
"required": [
"name"
]
}
},
{
"name": "subagent",
"description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) and return its final result. Use this to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent runs to completion and you receive only its final answer, not its intermediate steps. Give it a complete, standalone prompt: it does not see this conversation. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`.",
"parameters": {
"type": "object",
"properties": {
"description": {
"type": "string",
"description": "A short (3-5 word) description of the delegated task, for display."
},
"prompt": {
"type": "string",
"description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs."
},
"run_in_background": {
"type": "boolean",
"description": "Run as a background task and return its id; collect with task_output or stop with task_kill."
}
},
"required": [
"description",
"prompt"
]
}
},
{
"name": "subagent_fork",
"description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn), returning only its final result. Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive only its final answer, not its intermediate steps. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`.",
"parameters": {
"type": "object",
"properties": {
"description": {
"type": "string",
"description": "A short (3-5 word) description of the delegated task, for display."
},
"prompt": {
"type": "string",
"description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new."
},
"run_in_background": {
"type": "boolean",
"description": "Run as a background task and return its id; collect with task_output or stop with task_kill."
}
},
"required": [
"description",
"prompt"
]
}
},
{
"name": "task_kill",
"description": "Request cancellation of a running background task by task id. Returns immediately; the task settles as killed once its work actually stops.",
"parameters": {
"type": "object",
"properties": {
"task_id": {
"type": "string",
"description": "Task id returned by the tool that started the background work."
},
"reason": {
"type": "string",
"description": "Optional short reason, recorded in the log and forwarded to the task."
}
},
"required": [
"task_id"
]
}
},
{
"name": "task_list",
"description": "List your background tasks (running and finished) with their ids, kinds, and statuses.",
"parameters": {
"type": "object",
"properties": {}
}
},
{
"name": "task_output",
"description": "Read a background task. Stream tasks return only output since the previous read; final-output tasks return their result after settlement. Every response ends with `[status: ...]`. Reads are non-blocking unless `wait: true`, which waits up to the configured cap.",
"parameters": {
"type": "object",
"properties": {
"task_id": {
"type": "string",
"description": "Task id returned by the tool that started the background work."
},
"wait": {
"type": "boolean",
"description": "Block until the task reaches a terminal status or the timeout expires. A timed-out wait returns [status: running] and leaves the task alive."
},
"timeout_ms": {
"type": "number",
"description": "Max wait in milliseconds (only meaningful with wait: true). Defaults to the configured wait timeout; capped by the configured maximum."
}
},
"required": [
"task_id"
]
}
},
{
"name": "todo_write",
"description": "Record and update a structured task list for the current work. Send the ENTIRE list every call — it REPLACES the previous list (there are no partial updates, no per-item edits). Use it to plan multi-step work and show progress: add one todo per concrete step before you start. Keep AT MOST ONE todo `in_progress` at a time; while work remains, exactly one active task should be `in_progress`. Mark a todo `completed` the moment it is done (do not batch completions), and allow no `in_progress` item only once all work is complete. Skip the list for trivial single-step tasks. Statuses: `pending` (not started), `in_progress` (being worked on now), `completed` (finished).",
"parameters": {
"type": "object",
"properties": {
"todos": {
"type": "array",
"description": "The COMPLETE task list, replacing any previous list.",
"items": {
"type": "object",
"additionalProperties": false,
"properties": {
"content": {
"type": "string",
"description": "What the task is — a short imperative line."
},
"status": {
"type": "string",
"description": "pending (not started) | in_progress (now) | completed (done).",
"enum": [
"pending",
"in_progress",
"completed"
]
}
},
"required": [
"content",
"status"
]
}
}
},
"required": [
"todos"
]
}
},
{
"name": "update_goal",
"description": "Update the exact current goal revision. edit, pause, and resume require a direct top-level human request. During an automatic continuation of the current goal, complete and blocked are also allowed. blocked is rejected before the configured minimum round count; the model remains responsible for judging that the same condition persisted across those rounds and must explain it in blocked_reason.",
"parameters": {
"type": "object",
"properties": {
"goal_id": {
"type": "string",
"description": "Exact id returned by get_goal."
},
"revision": {
"type": "number",
"description": "Exact positive revision returned by get_goal."
},
"action": {
"type": "string",
"description": "edit | pause | resume | complete | blocked",
"enum": [
"edit",
"pause",
"resume",
"complete",
"blocked"
]
},
"objective": {
"type": "string",
"description": "Replacement objective; valid only with action edit."
},
"max_goal_rounds": {
"type": "number",
"description": "Replacement cap; valid only with action edit."
},
"blocked_reason": {
"type": "string",
"description": "Concrete blocking condition; required only with action blocked."
}
},
"required": [
"goal_id",
"revision",
"action"
]
}
},
{
"name": "workflow",
"description": "Run a JavaScript workflow script that orchestrates subagents at scale. Use this for work that fans out across many independent pieces — an audit over many files, a migration, multi-angle research, adversarial verification of findings — where you write the orchestration as a script instead of delegating turn by turn.\n\nThe workflow's identity rides the `meta` parameter as JSON: required `name` (short kebab-case) and `description` strings, optional `whenToUse` string and `phases` array (`{title, detail?, provider?, model?}`). The `script` parameter is the plain JavaScript body ONLY (NOT TypeScript, and NO `export const meta` statement — meta is a parameter, not code), running with top-level await; end with `return <value>` — the value must be JSON-serializable and is this tool's result.\n\nScript-body hooks:\n- `agent(prompt, opts?): Promise<any>` — run one subagent to completion. Without `opts.schema` it resolves to the child's final text; with `opts.schema` (an object-rooted JSON Schema using ONLY type/properties/required/additionalProperties/items/enum/const/oneOf — no pattern/format/numeric bounds) it resolves to the validated object. Resolves `null` when the child fails (filter with `.filter(Boolean)`). Other opts: `label` (display), `phase` (progress group), and independent `provider`/`model` LLM target overrides (either may be provided alone). Anything else (`effort`/`isolation`/`agentType`) is rejected loudly.\n- `pipeline(items, ...stages): Promise<any[]>` — run each item through the stages independently with NO barrier between stages (prefer this for multi-stage work). Each stage receives `(prev, item, index)`. An ordinary stage throw drops that ITEM to `null` and skips its remaining stages.\n- `parallel(thunks): Promise<any[]>` — run zero-argument functions concurrently and await ALL of them (a barrier; use only when a stage genuinely needs every prior result together). A throwing thunk resolves to `null`.\n- `phase(title)` — start a progress phase; `log(message)` — narrate progress; `args` — the tool call's `args` input, verbatim.\n\nMisused hooks (bad arguments, unknown options, unsupported schemas, tripped caps) throw errors that ALWAYS kill the script — they never dissolve into a per-item `null`.\n\nConstraints: concurrency and total-agent caps apply; no filesystem, network, timers, or Node.js APIs are provided — the agents do the work, the script only coordinates them. The run executes in the foreground: this call returns when the whole script finishes.",
"parameters": {
"type": "object",
"properties": {
"script": {
"type": "string",
"description": "The plain-JS workflow script body (top-level await allowed; NO `export const meta` statement; end with `return <json-value>`)."
},
"meta": {
"type": "object",
"description": "The workflow identity block (plain JSON — never code).",
"additionalProperties": true,
"properties": {
"name": {
"type": "string",
"description": "Short kebab-case workflow name."
},
"description": {
"type": "string",
"description": "One-line description of what the workflow does."
},
"whenToUse": {
"type": "string",
"description": "Optional guidance on when this workflow applies."
},
"phases": {
"type": "array",
"description": "Optional phase declarations matched by phase() calls.",
"items": {
"type": "object",
"additionalProperties": true,
"properties": {
"title": {
"type": "string",
"description": "The phase title phase() calls match by exact string."
},
"detail": {
"type": "string",
"description": "Optional one-line description of the phase."
},
"provider": {
"type": "string",
"description": "Optional provider override this phase is expected to use."
},
"model": {
"type": "string",
"description": "Optional model override this phase is expected to use."
}
},
"required": [
"title"
]
}
}
},
"required": [
"name",
"description"
]
},
"args": {
"type": "object",
"description": "Optional JSON input exposed to the script as the `args` global (wrap a bare list as a field, e.g. {\"files\": [...]}).",
"additionalProperties": true
}
},
"required": [
"script",
"meta"
]
}
},
{
"name": "write",
"description": "Create or fully replace a UTF-8 text file.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to write, resolved by the filesystem backend."
},
"content": {
"type": "string",
"description": "Full UTF-8 text content to write."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access."
}
},
"required": [
"file_path",
"content"
]
}
}
],
"changes": []
}

25
examples/acp-agent/tests/snapshots/skill-load/system-prompt.expected.md
View File

@@ -1,25 +0,0 @@
You are an AI agent powered by the DeepSeek Harness SDK.
You are a coding assistant powered by the deepseek-v4-flash model. Your working directory is {{cwd}}. Your bash tool runs under a file sandbox — a `[sandbox: file access denied …]` result is policy, not a command bug.
Verify your work by running the code or tests. Keep answers brief and factual.
Use the read tool — not shell commands like cat — to inspect text files. Results include line numbers. Use offset and limit to continue reading large files.
Use the write tool to create files or completely replace file contents. Existing files are overwritten, so read an existing file first (the default fs-policy requires it) and prefer edit for targeted changes.
Use the edit tool for targeted changes to existing UTF-8 text files. It replaces literal old_string with new_string; by default old_string must appear exactly once. If old_string appears multiple times, provide a more specific old_string or set replace_all to true. Read the file first (the default fs-policy requires it), unless you just created or edited it in this session.
Check the [exit code: N] marker on every bash result; investigate failures before moving on.
Track every background task id you start. You are notified in-session when a task finishes — do not busy-poll or sleep on one; keep working on independent steps and do not duplicate a running task's work. Before giving a final answer, collect every still-relevant task with task_output (set wait: true only when you are genuinely blocked on it), and task_kill tasks that stopped mattering.
Use goal tools for one long-running completion objective in the current session. create_goal may infer goal intent from a direct human request in any language; do not create a goal for routine single-turn work. Call get_goal before update_goal and copy its exact goal_id and revision. After session resume or fork, an active goal is disarmed: when a human asks to continue or resume in any wording or language, use update_goal action resume to rearm it. Mark complete only when the objective is actually achieved. Mark blocked only after the same blocking condition persists for at least 3 consecutive goal rounds, and report that concrete condition in blocked_reason; difficulty, uncertainty, or useful remaining work is not blocked.
Approval prompts are disabled in this session: actions that require approval are rejected automatically — do not request sandbox escalation (do not set `sandbox_permissions`).
<!-- dsh-user-approval-policy:never -->
Use the workflow tool ONLY when the user explicitly asks for a workflow or for large multi-agent orchestration: you write a JavaScript script (the tool description documents the exact format) that fans work out across many subagents with phases and structured results. For one or two delegations, prefer plain subagent calls.
Use the ralph tool ONLY when the direct human explicitly asks for a Ralph loop or fresh-agent iterative execution. Each Ralph round starts a fresh child with no conversation seed and uses the shared workspace as durable memory. Completion and blockers are worker reports, not independent evaluation. Use same-session goal tools for ordinary long-running objectives, and plain subagents or workflows for bounded delegation and fan-out.

473
examples/acp-agent/tests/snapshots/skill-load/tool-schemas.expected.json
View File

@@ -1,473 +0,0 @@
{
"initial": [
{
"name": "bash",
"description": "Execute a bash command (`bash -c`) and return its stdout/stderr. Each call runs in a fresh shell: no state (cwd, variables, functions) persists between calls — pass `workdir` instead of using `cd`. Non-zero exits are reported as `[exit code: N]`. Current harness environment facts are exposed through managed `$DSH_*` variables; inspect them when needed. Commands may run under a file sandbox; a blocked file operation is reported as `[sandbox: file access denied under <mode> mode]` — a policy denial, not a bug in the command; do not retry another way. Long output is truncated to its tail; the full output is saved to a file whose path is reported when available. Set `run_in_background: true` for long-running commands: the call returns a task id immediately; read its output with `task_output` and stop it with `task_kill`. Attempting a command the sandbox may deny is safe and expected: run it and read the marker rather than assuming the denial. When a command is denied and a wider mode would let it succeed, escalate immediately in the same turn — the one sanctioned exception to a denial: retry the exact same command once with `sandbox_permissions` (the narrowest wider mode that suffices) plus a one-sentence `justification`. Do not detour through chat to ask permission first — the approval prompt raised by that retry is how the user consents. If the session states approval prompts are disabled, there is no exception: a denial is final — do not set `sandbox_permissions`. Never escalate speculatively: ground the request in a real denial — normally the one this command just hit; escalating up front is fine only when this session already denied the same access. A rejected escalation is final for that command — stop and explain, never work around it — but it does not forbid attempting or escalating other commands later.",
"parameters": {
"type": "object",
"properties": {
"command": {
"type": "string",
"description": "The bash command to execute."
},
"description": {
"type": "string",
"description": "Clear, concise description of what this command does in active voice, 5-10 words (shown in the UI). Examples: \"ls\" → \"List files in current directory\"; \"git status\" → \"Show working tree status\"; \"npm install\" → \"Install package dependencies\"."
},
"timeoutMs": {
"type": "number",
"description": "Timeout in milliseconds. The executor applies its configured default and cap, and kills the command on expiry."
},
"workdir": {
"type": "string",
"description": "Working directory for this command. Defaults to the session workspace; a relative path is resolved against it."
},
"run_in_background": {
"type": "boolean",
"description": "Run in the background and return a task id immediately (collect with task_output, stop with task_kill). No timeout applies."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this command needs. Only valid as a one-shot retry of a command the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact command needs the wider access."
}
},
"required": [
"command",
"description"
]
}
},
{
"name": "create_goal",
"description": "Create one persisted same-session completion goal when the current direct human request is a long-running objective that should continue across autonomous goal rounds. You may infer that intent without requiring the user to say \"create a goal\". Do not use this for trivial single-turn work. Execution rejects non-human and subagent authority.",
"parameters": {
"type": "object",
"properties": {
"objective": {
"type": "string",
"description": "The concrete completion objective inferred from the direct human request."
},
"max_goal_rounds": {
"type": "number",
"description": "Optional positive safe-integer limit on automatic continuation rounds."
}
},
"required": [
"objective"
]
}
},
{
"name": "edit",
"description": "Edit an existing UTF-8 text file by replacing literal text.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to edit, resolved by the filesystem backend."
},
"old_string": {
"type": "string",
"description": "Literal text to replace. Must match exactly."
},
"new_string": {
"type": "string",
"description": "Literal replacement text. Use an empty string to delete the match."
},
"replace_all": {
"type": "boolean",
"description": "Replace all matches. Defaults to false; when false, old_string must appear exactly once."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access."
}
},
"required": [
"file_path",
"old_string",
"new_string"
]
}
},
{
"name": "get_goal",
"description": "Read the current same-session goal, including its exact id/revision, objective, phase, completed continuation rounds, round limit, blocker reason when present, and whether another continuation is armed. Call this before updating a goal.",
"parameters": {
"type": "object",
"properties": {}
}
},
{
"name": "ralph",
"description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.",
"parameters": {
"type": "object",
"properties": {
"objective": {
"type": "string",
"description": "The immutable completion objective for every fresh Ralph round."
},
"maxRounds": {
"type": "number",
"description": "Optional positive safe-integer round cap, bounded by the deployment ceiling."
}
},
"required": [
"objective"
]
}
},
{
"name": "read",
"description": "Read a UTF-8 text file and return line-numbered content.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to read, resolved by the filesystem backend."
},
"offset": {
"type": "number",
"description": "1-based first line to return. Defaults to 1."
},
"limit": {
"type": "number",
"description": "Maximum number of lines to return. Defaults to 2000."
}
},
"required": [
"file_path"
]
}
},
{
"name": "skill",
"description": "Load the full instructions for an available skill. Call this with the exact skill name from the session skill catalog before acting on a task that names or clearly matches that skill.",
"parameters": {
"type": "object",
"properties": {
"name": {
"type": "string",
"description": "The exact skill name from the available skills list."
}
},
"required": [
"name"
]
}
},
{
"name": "subagent",
"description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) and return its final result. Use this to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent runs to completion and you receive only its final answer, not its intermediate steps. Give it a complete, standalone prompt: it does not see this conversation. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`.",
"parameters": {
"type": "object",
"properties": {
"description": {
"type": "string",
"description": "A short (3-5 word) description of the delegated task, for display."
},
"prompt": {
"type": "string",
"description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs."
},
"run_in_background": {
"type": "boolean",
"description": "Run as a background task and return its id; collect with task_output or stop with task_kill."
}
},
"required": [
"description",
"prompt"
]
}
},
{
"name": "subagent_fork",
"description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn), returning only its final result. Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive only its final answer, not its intermediate steps. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`.",
"parameters": {
"type": "object",
"properties": {
"description": {
"type": "string",
"description": "A short (3-5 word) description of the delegated task, for display."
},
"prompt": {
"type": "string",
"description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new."
},
"run_in_background": {
"type": "boolean",
"description": "Run as a background task and return its id; collect with task_output or stop with task_kill."
}
},
"required": [
"description",
"prompt"
]
}
},
{
"name": "task_kill",
"description": "Request cancellation of a running background task by task id. Returns immediately; the task settles as killed once its work actually stops.",
"parameters": {
"type": "object",
"properties": {
"task_id": {
"type": "string",
"description": "Task id returned by the tool that started the background work."
},
"reason": {
"type": "string",
"description": "Optional short reason, recorded in the log and forwarded to the task."
}
},
"required": [
"task_id"
]
}
},
{
"name": "task_list",
"description": "List your background tasks (running and finished) with their ids, kinds, and statuses.",
"parameters": {
"type": "object",
"properties": {}
}
},
{
"name": "task_output",
"description": "Read a background task. Stream tasks return only output since the previous read; final-output tasks return their result after settlement. Every response ends with `[status: ...]`. Reads are non-blocking unless `wait: true`, which waits up to the configured cap.",
"parameters": {
"type": "object",
"properties": {
"task_id": {
"type": "string",
"description": "Task id returned by the tool that started the background work."
},
"wait": {
"type": "boolean",
"description": "Block until the task reaches a terminal status or the timeout expires. A timed-out wait returns [status: running] and leaves the task alive."
},
"timeout_ms": {
"type": "number",
"description": "Max wait in milliseconds (only meaningful with wait: true). Defaults to the configured wait timeout; capped by the configured maximum."
}
},
"required": [
"task_id"
]
}
},
{
"name": "todo_write",
"description": "Record and update a structured task list for the current work. Send the ENTIRE list every call — it REPLACES the previous list (there are no partial updates, no per-item edits). Use it to plan multi-step work and show progress: add one todo per concrete step before you start. Keep AT MOST ONE todo `in_progress` at a time; while work remains, exactly one active task should be `in_progress`. Mark a todo `completed` the moment it is done (do not batch completions), and allow no `in_progress` item only once all work is complete. Skip the list for trivial single-step tasks. Statuses: `pending` (not started), `in_progress` (being worked on now), `completed` (finished).",
"parameters": {
"type": "object",
"properties": {
"todos": {
"type": "array",
"description": "The COMPLETE task list, replacing any previous list.",
"items": {
"type": "object",
"additionalProperties": false,
"properties": {
"content": {
"type": "string",
"description": "What the task is — a short imperative line."
},
"status": {
"type": "string",
"description": "pending (not started) | in_progress (now) | completed (done).",
"enum": [
"pending",
"in_progress",
"completed"
]
}
},
"required": [
"content",
"status"
]
}
}
},
"required": [
"todos"
]
}
},
{
"name": "update_goal",
"description": "Update the exact current goal revision. edit, pause, and resume require a direct top-level human request. During an automatic continuation of the current goal, complete and blocked are also allowed. blocked is rejected before the configured minimum round count; the model remains responsible for judging that the same condition persisted across those rounds and must explain it in blocked_reason.",
"parameters": {
"type": "object",
"properties": {
"goal_id": {
"type": "string",
"description": "Exact id returned by get_goal."
},
"revision": {
"type": "number",
"description": "Exact positive revision returned by get_goal."
},
"action": {
"type": "string",
"description": "edit | pause | resume | complete | blocked",
"enum": [
"edit",
"pause",
"resume",
"complete",
"blocked"
]
},
"objective": {
"type": "string",
"description": "Replacement objective; valid only with action edit."
},
"max_goal_rounds": {
"type": "number",
"description": "Replacement cap; valid only with action edit."
},
"blocked_reason": {
"type": "string",
"description": "Concrete blocking condition; required only with action blocked."
}
},
"required": [
"goal_id",
"revision",
"action"
]
}
},
{
"name": "workflow",
"description": "Run a JavaScript workflow script that orchestrates subagents at scale. Use this for work that fans out across many independent pieces — an audit over many files, a migration, multi-angle research, adversarial verification of findings — where you write the orchestration as a script instead of delegating turn by turn.\n\nThe workflow's identity rides the `meta` parameter as JSON: required `name` (short kebab-case) and `description` strings, optional `whenToUse` string and `phases` array (`{title, detail?, provider?, model?}`). The `script` parameter is the plain JavaScript body ONLY (NOT TypeScript, and NO `export const meta` statement — meta is a parameter, not code), running with top-level await; end with `return <value>` — the value must be JSON-serializable and is this tool's result.\n\nScript-body hooks:\n- `agent(prompt, opts?): Promise<any>` — run one subagent to completion. Without `opts.schema` it resolves to the child's final text; with `opts.schema` (an object-rooted JSON Schema using ONLY type/properties/required/additionalProperties/items/enum/const/oneOf — no pattern/format/numeric bounds) it resolves to the validated object. Resolves `null` when the child fails (filter with `.filter(Boolean)`). Other opts: `label` (display), `phase` (progress group), and independent `provider`/`model` LLM target overrides (either may be provided alone). Anything else (`effort`/`isolation`/`agentType`) is rejected loudly.\n- `pipeline(items, ...stages): Promise<any[]>` — run each item through the stages independently with NO barrier between stages (prefer this for multi-stage work). Each stage receives `(prev, item, index)`. An ordinary stage throw drops that ITEM to `null` and skips its remaining stages.\n- `parallel(thunks): Promise<any[]>` — run zero-argument functions concurrently and await ALL of them (a barrier; use only when a stage genuinely needs every prior result together). A throwing thunk resolves to `null`.\n- `phase(title)` — start a progress phase; `log(message)` — narrate progress; `args` — the tool call's `args` input, verbatim.\n\nMisused hooks (bad arguments, unknown options, unsupported schemas, tripped caps) throw errors that ALWAYS kill the script — they never dissolve into a per-item `null`.\n\nConstraints: concurrency and total-agent caps apply; no filesystem, network, timers, or Node.js APIs are provided — the agents do the work, the script only coordinates them. The run executes in the foreground: this call returns when the whole script finishes.",
"parameters": {
"type": "object",
"properties": {
"script": {
"type": "string",
"description": "The plain-JS workflow script body (top-level await allowed; NO `export const meta` statement; end with `return <json-value>`)."
},
"meta": {
"type": "object",
"description": "The workflow identity block (plain JSON — never code).",
"additionalProperties": true,
"properties": {
"name": {
"type": "string",
"description": "Short kebab-case workflow name."
},
"description": {
"type": "string",
"description": "One-line description of what the workflow does."
},
"whenToUse": {
"type": "string",
"description": "Optional guidance on when this workflow applies."
},
"phases": {
"type": "array",
"description": "Optional phase declarations matched by phase() calls.",
"items": {
"type": "object",
"additionalProperties": true,
"properties": {
"title": {
"type": "string",
"description": "The phase title phase() calls match by exact string."
},
"detail": {
"type": "string",
"description": "Optional one-line description of the phase."
},
"provider": {
"type": "string",
"description": "Optional provider override this phase is expected to use."
},
"model": {
"type": "string",
"description": "Optional model override this phase is expected to use."
}
},
"required": [
"title"
]
}
}
},
"required": [
"name",
"description"
]
},
"args": {
"type": "object",
"description": "Optional JSON input exposed to the script as the `args` global (wrap a bare list as a field, e.g. {\"files\": [...]}).",
"additionalProperties": true
}
},
"required": [
"script",
"meta"
]
}
},
{
"name": "write",
"description": "Create or fully replace a UTF-8 text file.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to write, resolved by the filesystem backend."
},
"content": {
"type": "string",
"description": "Full UTF-8 text content to write."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access."
}
},
"required": [
"file_path",
"content"
]
}
}
],
"changes": []
}

473
examples/acp-agent/tests/snapshots/workspace-context/tool-schemas.expected.json
View File

@@ -1,473 +0,0 @@
{
"initial": [
{
"name": "bash",
"description": "Execute a bash command (`bash -c`) and return its stdout/stderr. Each call runs in a fresh shell: no state (cwd, variables, functions) persists between calls — pass `workdir` instead of using `cd`. Non-zero exits are reported as `[exit code: N]`. Current harness environment facts are exposed through managed `$DSH_*` variables; inspect them when needed. Commands may run under a file sandbox; a blocked file operation is reported as `[sandbox: file access denied under <mode> mode]` — a policy denial, not a bug in the command; do not retry another way. Long output is truncated to its tail; the full output is saved to a file whose path is reported when available. Set `run_in_background: true` for long-running commands: the call returns a task id immediately; read its output with `task_output` and stop it with `task_kill`. Attempting a command the sandbox may deny is safe and expected: run it and read the marker rather than assuming the denial. When a command is denied and a wider mode would let it succeed, escalate immediately in the same turn — the one sanctioned exception to a denial: retry the exact same command once with `sandbox_permissions` (the narrowest wider mode that suffices) plus a one-sentence `justification`. Do not detour through chat to ask permission first — the approval prompt raised by that retry is how the user consents. If the session states approval prompts are disabled, there is no exception: a denial is final — do not set `sandbox_permissions`. Never escalate speculatively: ground the request in a real denial — normally the one this command just hit; escalating up front is fine only when this session already denied the same access. A rejected escalation is final for that command — stop and explain, never work around it — but it does not forbid attempting or escalating other commands later.",
"parameters": {
"type": "object",
"properties": {
"command": {
"type": "string",
"description": "The bash command to execute."
},
"description": {
"type": "string",
"description": "Clear, concise description of what this command does in active voice, 5-10 words (shown in the UI). Examples: \"ls\" → \"List files in current directory\"; \"git status\" → \"Show working tree status\"; \"npm install\" → \"Install package dependencies\"."
},
"timeoutMs": {
"type": "number",
"description": "Timeout in milliseconds. The executor applies its configured default and cap, and kills the command on expiry."
},
"workdir": {
"type": "string",
"description": "Working directory for this command. Defaults to the session workspace; a relative path is resolved against it."
},
"run_in_background": {
"type": "boolean",
"description": "Run in the background and return a task id immediately (collect with task_output, stop with task_kill). No timeout applies."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this command needs. Only valid as a one-shot retry of a command the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact command needs the wider access."
}
},
"required": [
"command",
"description"
]
}
},
{
"name": "create_goal",
"description": "Create one persisted same-session completion goal when the current direct human request is a long-running objective that should continue across autonomous goal rounds. You may infer that intent without requiring the user to say \"create a goal\". Do not use this for trivial single-turn work. Execution rejects non-human and subagent authority.",
"parameters": {
"type": "object",
"properties": {
"objective": {
"type": "string",
"description": "The concrete completion objective inferred from the direct human request."
},
"max_goal_rounds": {
"type": "number",
"description": "Optional positive safe-integer limit on automatic continuation rounds."
}
},
"required": [
"objective"
]
}
},
{
"name": "edit",
"description": "Edit an existing UTF-8 text file by replacing literal text.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to edit, resolved by the filesystem backend."
},
"old_string": {
"type": "string",
"description": "Literal text to replace. Must match exactly."
},
"new_string": {
"type": "string",
"description": "Literal replacement text. Use an empty string to delete the match."
},
"replace_all": {
"type": "boolean",
"description": "Replace all matches. Defaults to false; when false, old_string must appear exactly once."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access."
}
},
"required": [
"file_path",
"old_string",
"new_string"
]
}
},
{
"name": "get_goal",
"description": "Read the current same-session goal, including its exact id/revision, objective, phase, completed continuation rounds, round limit, blocker reason when present, and whether another continuation is armed. Call this before updating a goal.",
"parameters": {
"type": "object",
"properties": {}
}
},
{
"name": "ralph",
"description": "Run a foreground fresh-agent Ralph loop toward one immutable objective. Use only when the direct human explicitly asks for Ralph or fresh-agent iteration. Each round opens a new child with no parent conversation or prior child session; the shared workspace is long-term memory, and only a bounded structured report crosses rounds. The call returns when a worker reports completion or a concrete blocker, or at the round limit. Ordinary long-running same-session work belongs to goal tools.",
"parameters": {
"type": "object",
"properties": {
"objective": {
"type": "string",
"description": "The immutable completion objective for every fresh Ralph round."
},
"maxRounds": {
"type": "number",
"description": "Optional positive safe-integer round cap, bounded by the deployment ceiling."
}
},
"required": [
"objective"
]
}
},
{
"name": "read",
"description": "Read a UTF-8 text file and return line-numbered content.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to read, resolved by the filesystem backend."
},
"offset": {
"type": "number",
"description": "1-based first line to return. Defaults to 1."
},
"limit": {
"type": "number",
"description": "Maximum number of lines to return. Defaults to 2000."
}
},
"required": [
"file_path"
]
}
},
{
"name": "skill",
"description": "Load the full instructions for an available skill. Call this with the exact skill name from the session skill catalog before acting on a task that names or clearly matches that skill.",
"parameters": {
"type": "object",
"properties": {
"name": {
"type": "string",
"description": "The exact skill name from the available skills list."
}
},
"required": [
"name"
]
}
},
{
"name": "subagent",
"description": "Delegate a self-contained task to a subagent (a separate agent that works in its own context) and return its final result. Use this to offload focused, independent work — research, a scoped implementation, an analysis — so it does not consume this conversation's context. The subagent runs to completion and you receive only its final answer, not its intermediate steps. Give it a complete, standalone prompt: it does not see this conversation. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`.",
"parameters": {
"type": "object",
"properties": {
"description": {
"type": "string",
"description": "A short (3-5 word) description of the delegated task, for display."
},
"prompt": {
"type": "string",
"description": "The complete, self-contained task for the subagent. It does not share this conversation's context, so include everything it needs."
},
"run_in_background": {
"type": "boolean",
"description": "Run as a background task and return its id; collect with task_output or stop with task_kill."
}
},
"required": [
"description",
"prompt"
]
}
},
{
"name": "subagent_fork",
"description": "Delegate a task to a subagent that inherits this conversation: a child agent seeded with all completed turns so far (it does not see the current in-flight turn), returning only its final result. Use this when the subtask builds on this conversation's context — a follow-up analysis, a review, a continuation — without consuming this conversation's context for the work itself. You receive only its final answer, not its intermediate steps. Set `run_in_background: true` to return a task id; collect with `task_output` and stop with `task_kill`.",
"parameters": {
"type": "object",
"properties": {
"description": {
"type": "string",
"description": "A short (3-5 word) description of the delegated task, for display."
},
"prompt": {
"type": "string",
"description": "The task for the subagent. It already sees this conversation's completed turns, so build on them freely and state only what is new."
},
"run_in_background": {
"type": "boolean",
"description": "Run as a background task and return its id; collect with task_output or stop with task_kill."
}
},
"required": [
"description",
"prompt"
]
}
},
{
"name": "task_kill",
"description": "Request cancellation of a running background task by task id. Returns immediately; the task settles as killed once its work actually stops.",
"parameters": {
"type": "object",
"properties": {
"task_id": {
"type": "string",
"description": "Task id returned by the tool that started the background work."
},
"reason": {
"type": "string",
"description": "Optional short reason, recorded in the log and forwarded to the task."
}
},
"required": [
"task_id"
]
}
},
{
"name": "task_list",
"description": "List your background tasks (running and finished) with their ids, kinds, and statuses.",
"parameters": {
"type": "object",
"properties": {}
}
},
{
"name": "task_output",
"description": "Read a background task. Stream tasks return only output since the previous read; final-output tasks return their result after settlement. Every response ends with `[status: ...]`. Reads are non-blocking unless `wait: true`, which waits up to the configured cap.",
"parameters": {
"type": "object",
"properties": {
"task_id": {
"type": "string",
"description": "Task id returned by the tool that started the background work."
},
"wait": {
"type": "boolean",
"description": "Block until the task reaches a terminal status or the timeout expires. A timed-out wait returns [status: running] and leaves the task alive."
},
"timeout_ms": {
"type": "number",
"description": "Max wait in milliseconds (only meaningful with wait: true). Defaults to the configured wait timeout; capped by the configured maximum."
}
},
"required": [
"task_id"
]
}
},
{
"name": "todo_write",
"description": "Record and update a structured task list for the current work. Send the ENTIRE list every call — it REPLACES the previous list (there are no partial updates, no per-item edits). Use it to plan multi-step work and show progress: add one todo per concrete step before you start. Keep AT MOST ONE todo `in_progress` at a time; while work remains, exactly one active task should be `in_progress`. Mark a todo `completed` the moment it is done (do not batch completions), and allow no `in_progress` item only once all work is complete. Skip the list for trivial single-step tasks. Statuses: `pending` (not started), `in_progress` (being worked on now), `completed` (finished).",
"parameters": {
"type": "object",
"properties": {
"todos": {
"type": "array",
"description": "The COMPLETE task list, replacing any previous list.",
"items": {
"type": "object",
"additionalProperties": false,
"properties": {
"content": {
"type": "string",
"description": "What the task is — a short imperative line."
},
"status": {
"type": "string",
"description": "pending (not started) | in_progress (now) | completed (done).",
"enum": [
"pending",
"in_progress",
"completed"
]
}
},
"required": [
"content",
"status"
]
}
}
},
"required": [
"todos"
]
}
},
{
"name": "update_goal",
"description": "Update the exact current goal revision. edit, pause, and resume require a direct top-level human request. During an automatic continuation of the current goal, complete and blocked are also allowed. blocked is rejected before the configured minimum round count; the model remains responsible for judging that the same condition persisted across those rounds and must explain it in blocked_reason.",
"parameters": {
"type": "object",
"properties": {
"goal_id": {
"type": "string",
"description": "Exact id returned by get_goal."
},
"revision": {
"type": "number",
"description": "Exact positive revision returned by get_goal."
},
"action": {
"type": "string",
"description": "edit | pause | resume | complete | blocked",
"enum": [
"edit",
"pause",
"resume",
"complete",
"blocked"
]
},
"objective": {
"type": "string",
"description": "Replacement objective; valid only with action edit."
},
"max_goal_rounds": {
"type": "number",
"description": "Replacement cap; valid only with action edit."
},
"blocked_reason": {
"type": "string",
"description": "Concrete blocking condition; required only with action blocked."
}
},
"required": [
"goal_id",
"revision",
"action"
]
}
},
{
"name": "workflow",
"description": "Run a JavaScript workflow script that orchestrates subagents at scale. Use this for work that fans out across many independent pieces — an audit over many files, a migration, multi-angle research, adversarial verification of findings — where you write the orchestration as a script instead of delegating turn by turn.\n\nThe workflow's identity rides the `meta` parameter as JSON: required `name` (short kebab-case) and `description` strings, optional `whenToUse` string and `phases` array (`{title, detail?, provider?, model?}`). The `script` parameter is the plain JavaScript body ONLY (NOT TypeScript, and NO `export const meta` statement — meta is a parameter, not code), running with top-level await; end with `return <value>` — the value must be JSON-serializable and is this tool's result.\n\nScript-body hooks:\n- `agent(prompt, opts?): Promise<any>` — run one subagent to completion. Without `opts.schema` it resolves to the child's final text; with `opts.schema` (an object-rooted JSON Schema using ONLY type/properties/required/additionalProperties/items/enum/const/oneOf — no pattern/format/numeric bounds) it resolves to the validated object. Resolves `null` when the child fails (filter with `.filter(Boolean)`). Other opts: `label` (display), `phase` (progress group), and independent `provider`/`model` LLM target overrides (either may be provided alone). Anything else (`effort`/`isolation`/`agentType`) is rejected loudly.\n- `pipeline(items, ...stages): Promise<any[]>` — run each item through the stages independently with NO barrier between stages (prefer this for multi-stage work). Each stage receives `(prev, item, index)`. An ordinary stage throw drops that ITEM to `null` and skips its remaining stages.\n- `parallel(thunks): Promise<any[]>` — run zero-argument functions concurrently and await ALL of them (a barrier; use only when a stage genuinely needs every prior result together). A throwing thunk resolves to `null`.\n- `phase(title)` — start a progress phase; `log(message)` — narrate progress; `args` — the tool call's `args` input, verbatim.\n\nMisused hooks (bad arguments, unknown options, unsupported schemas, tripped caps) throw errors that ALWAYS kill the script — they never dissolve into a per-item `null`.\n\nConstraints: concurrency and total-agent caps apply; no filesystem, network, timers, or Node.js APIs are provided — the agents do the work, the script only coordinates them. The run executes in the foreground: this call returns when the whole script finishes.",
"parameters": {
"type": "object",
"properties": {
"script": {
"type": "string",
"description": "The plain-JS workflow script body (top-level await allowed; NO `export const meta` statement; end with `return <json-value>`)."
},
"meta": {
"type": "object",
"description": "The workflow identity block (plain JSON — never code).",
"additionalProperties": true,
"properties": {
"name": {
"type": "string",
"description": "Short kebab-case workflow name."
},
"description": {
"type": "string",
"description": "One-line description of what the workflow does."
},
"whenToUse": {
"type": "string",
"description": "Optional guidance on when this workflow applies."
},
"phases": {
"type": "array",
"description": "Optional phase declarations matched by phase() calls.",
"items": {
"type": "object",
"additionalProperties": true,
"properties": {
"title": {
"type": "string",
"description": "The phase title phase() calls match by exact string."
},
"detail": {
"type": "string",
"description": "Optional one-line description of the phase."
},
"provider": {
"type": "string",
"description": "Optional provider override this phase is expected to use."
},
"model": {
"type": "string",
"description": "Optional model override this phase is expected to use."
}
},
"required": [
"title"
]
}
}
},
"required": [
"name",
"description"
]
},
"args": {
"type": "object",
"description": "Optional JSON input exposed to the script as the `args` global (wrap a bare list as a field, e.g. {\"files\": [...]}).",
"additionalProperties": true
}
},
"required": [
"script",
"meta"
]
}
},
{
"name": "write",
"description": "Create or fully replace a UTF-8 text file.",
"parameters": {
"type": "object",
"properties": {
"file_path": {
"type": "string",
"description": "Path to write, resolved by the filesystem backend."
},
"content": {
"type": "string",
"description": "Full UTF-8 text content to write."
},
"sandbox_permissions": {
"type": "string",
"description": "The wider sandbox mode this file operation needs. Only valid as a one-shot retry of an operation the sandbox just denied; requires justification and user approval.",
"enum": [
"workspace-write",
"danger-full-access"
]
},
"justification": {
"type": "string",
"description": "Required with sandbox_permissions: one sentence for the user explaining why this exact file operation needs the wider access."
}
},
"required": [
"file_path",
"content"
]
}
}
],
"changes": []
}

36
packages/client/connection/src/client/fixture.ts
View File

@@ -309,20 +309,31 @@ function projectionValuesOf(log: readonly SessionEvent[]): Record<string, unknow
if (titleEvent !== undefined) {
values['title'] = (titleEvent as unknown as { data: { title: string } }).data.title
}
const todos = backscanTodos(log)
if (todos !== undefined) values['todos'] = todos
// Always present (tool-todo unit composed): null when no plan stands.
values['todos'] = backscanTodos(log) ?? null
return values
}
/** Host push-frame parallel: emit one session/projection frame per key the given event advanced. */
function projectionFramesOf(id: SessionId, log: readonly SessionEvent[], event: SessionEvent): Extract<MuxFrame, { type: 'session/projection' }>[] {
const type = (event as { type: string }).type
const key = type === 'session/title' ? 'title' : type === 'todo/write' ? 'todos' : undefined
if (key === undefined) return []
const values = projectionValuesOf(log)
/* v8 ignore next -- the advancing event is in the log, so its key always has a value. */
if (!Object.hasOwn(values, key)) return []
return [{ type: 'session/projection', sessionId: id, key, value: values[key], seq: event.seq }]
if (type === 'session/title') {
const values = projectionValuesOf(log)
/* v8 ignore next -- the advancing title event is in the log, so the key is present. */
if (!Object.hasOwn(values, 'title')) return []
return [{ type: 'session/projection', sessionId: id, key: 'title', value: values['title'], seq: event.seq }]
}
// Standing-plan fold: writes replace the list; turn/start clears it (null).
if (type === 'todo/write' || type === 'turn/start') {
return [{
type: 'session/projection',
sessionId: id,
key: 'todos',
value: backscanTodos(log) ?? null,
seq: event.seq,
}]
}
return []
}
/**
@@ -356,11 +367,16 @@ function pageOf(
return { events, hasMore: start > 0 }
}
/** Current todo projection over the full log (host parallel: latest todo/write, last write wins). */
/**
* Current plan projection over the full log (host parallel: latest todo/write
* with no later turn/start; a new turn retires the previous plan).
*/
function backscanTodos(log: readonly SessionEvent[]): TodoItem[] | undefined {
for (let i = log.length - 1; i >= 0; i--) {
const event = log[i]
if (event !== undefined && event.type === 'todo/write') return event.data.todos
if (event === undefined) continue
if (event.type === 'turn/start') return undefined
if (event.type === 'todo/write') return event.data.todos
}
return undefined
}

5
packages/client/connection/tests/fixture.spec.ts
View File

@@ -69,7 +69,10 @@ describe('createFixtureApi', () => {
// tail block still rides it — empty-log cut at -1, the host convention.
const empty = await api.sessions.history(req({ sessionId: sid('no-such'), maxMessages: 10 }))
if (!empty.result.ok) throw new Error('empty failed')
expect(empty.result.value).toEqual({ events: [], hasMore: false, projections: { asOfSeq: -1, values: {} } })
// Fixture composes the todos unit (host parallel when tool-todo is mounted): null before any write.
expect(empty.result.value).toEqual({
events: [], hasMore: false, projections: { asOfSeq: -1, values: { todos: null } },
})
})
it('serves grouped models and keeps a selected target for later history and fixture requests', async () => {

4
packages/client/runtime/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/client/runtime/README.md
README.md: 16c1124ec812b9f030ce8266a16cdb8f5db0e6cc
README.zh.md: a3d2a2dfdd1662afee65ec45e26b1ef1029f44b5
README.md: 25eb60e2c95059ae918669c9f5169b6b8e9c6816
README.zh.md: e3085f91750503aeaffda41d86c40c62943b4ba9

2
packages/client/runtime/README.md
View File

@@ -2,7 +2,7 @@
English | [中文](README.zh.md)
Client cordis boot and React-free object services: SlotsService wraps SlotCore and supplies renderer data sources; SessionsService owns Session objects, list/scope/history state; WorkspacesService depends on SessionsService and owns Workspace objects, list/actions, default-target derivation, and the New Session blank-reuse entry (`connectWorkspace`). The runtime fans the shared Host stream into both managers. Client sessions are always Host-born (Session+Agent+cwd in one `session.create`); the client holds no pre-entity session state — a session's Agent scope (the client mirror of host dsh-scope, keyed by the shared agent/session id) is born when its row enters the list mirror and dies with the prune. Contract: api-contracts v3 §4. `ConversationSnapshot` carries `todos` — the session's current todo projection: taken from the tail history page's full-log value (host-computed, independent of the page window), preserved across an older-page prepend, and overwritten by each live `todo/write` (last write wins). A tail response that omits the field means the log holds no `todo/write`, so the list resets to empty — a plan the log never kept (a write lost to a host crash) disappears on the next open or resync.
Client cordis boot and React-free object services: SlotsService wraps SlotCore and supplies renderer data sources; SessionsService owns Session objects, list/scope/history state; WorkspacesService depends on SessionsService and owns Workspace objects, list/actions, default-target derivation, and the New Session blank-reuse entry (`connectWorkspace`). The runtime fans the shared Host stream into both managers. Client sessions are always Host-born (Session+Agent+cwd in one `session.create`); the client holds no pre-entity session state — a session's Agent scope (the client mirror of host dsh-scope, keyed by the shared agent/session id) is born when its row enters the list mirror and dies with the prune. Contract: api-contracts v3 §4. Each `Session` holds a generic `ProjectionValueStore` seeded from the history-tail `projections` block and updated by `session/projection` frames under higher-seq-wins; domain keys (including `todos`) are read via `projections.faceOf` / `useProjection`, not via `ConversationSnapshot`.
## Workspace and Session lists

2
packages/client/runtime/README.zh.md
View File

@@ -2,7 +2,7 @@
[English](README.md) | 中文
客户端 cordis 启动与不依赖 React 的对象服务：SlotsService 包装 SlotCore 并提供 renderer 数据源；SessionsService 拥有 Session 对象、列表／scope／history 状态；WorkspacesService 依赖 SessionsService，拥有 Workspace 对象、列表／操作、默认目标派生，以及 New Session 空会话复用入口（`connectWorkspace`）。运行时把共享 Host 流分发给两个 manager。客户端 Session 一律由 Host 出生（一次 `session.create` 同瞬产出 Session+Agent+cwd）；客户端不持有任何实体化之前的会话状态——Agent scope（host dsh-scope 的客户端镜像，以 agent/session 共用 id 为键）在会话行进入列表镜像时出生，随 prune 死亡。契约：api-contracts v3 §4。`ConversationSnapshot` 携带 `todos`——会话当前的 todo 投影：取自尾页 history 携带的全量 log 值（host 计算，独立于分页窗口），跨往前翻页保留，并被每次实时 `todo/write` 覆盖（后写胜出）。尾页响应省略该字段即表示 log 中没有任何 `todo/write`，因此列表复位为空——log 从未留下的计划（写入因 host 崩溃丢失）会在下一次打开或 resync 时消失。
客户端 cordis 启动与不依赖 React 的对象服务：SlotsService 包装 SlotCore 并提供 renderer 数据源；SessionsService 拥有 Session 对象、列表／scope／history 状态；WorkspacesService 依赖 SessionsService，拥有 Workspace 对象、列表／操作、默认目标派生，以及 New Session 空会话复用入口（`connectWorkspace`）。运行时把共享 Host 流分发给两个 manager。客户端 Session 一律由 Host 出生（一次 `session.create` 同瞬产出 Session+Agent+cwd）；客户端不持有任何实体化之前的会话状态——Agent scope（host dsh-scope 的客户端镜像，以 agent/session 共用 id 为键）在会话行进入列表镜像时出生，随 prune 死亡。契约：api-contracts v3 §4。每个 `Session` 持有一个通用的 `ProjectionValueStore`，由历史尾页的 `projections` 块播种，并经 `session/projection` 帧按 seq 高者胜更新；领域键（含 `todos`）经 `projections.faceOf`／`useProjection` 读取，不经 `ConversationSnapshot`。
## Workspace 与 Session 列表

13
packages/client/runtime/src/client/sessions/manager.ts
View File

@@ -212,6 +212,19 @@ export class SessionManager {
session.handleBlank(s.blank)
session.handleRunning(s.running)
}
// Seed each row's projection baseline into the per-session value
// store (cold titles surface without opening the session). Per-key
// apply, not seed(): the list block is a partial baseline — the
// cold cache serves only version-matching keys — so an absent key
// must not clear; higher-seq-wins still keeps a stale list block
// from overwriting a newer push frame or tail baseline.
for (const s of result.value.items) {
const block = s.projections
if (block === undefined) continue
const store = this.projectionStore(s.sessionId)
const values = block.values as Record<string, unknown>
for (const key of Object.keys(values)) store.apply(key, values[key], block.asOfSeq)
}
} else {
this.listState = 'error'
this.listError = result.error

22
packages/client/runtime/tests/manager.spec.ts
View File

@@ -160,6 +160,28 @@ describe('list lifecycle', () => {
expect(manager.getListSnapshot().items.find(item => item.sessionId === S1)?.title).toBeUndefined()
})
it('seeds cold titles from the list rows\' projections block under higher-seq-wins', async () => {
const api = new FakeApiClient()
const manager = new SessionManager(api)
// A push frame landed before the list (S2's title is newer than the block's cut).
manager.handleMuxEnvelope({
rpcId: 'push-newer' as never,
payload: { type: 'session/projection', sessionId: S2, key: 'title', value: 'Pushed', seq: 9 } as never,
})
api.onList = () => Promise.resolve(ok({
items: [
{ ...summary(S1), projections: { asOfSeq: 4, values: { title: 'Cold cached' } } },
{ ...summary(S2, { updatedAt: 200 }), projections: { asOfSeq: 5, values: { title: 'List stale' } } },
] as never[],
}))
await manager.refreshList()
const items = manager.getListSnapshot().items
// Cold row: title surfaces straight from the list block — no open, no history.
expect(items.find(item => item.sessionId === S1)?.title).toBe('Cold cached')
// The stale list block (seq 5) cannot overwrite the newer push frame (seq 9).
expect(items.find(item => item.sessionId === S2)?.title).toBe('Pushed')
})
it('drops a projection row beyond the subscription baseline before accepting its durable replay', async () => {
const api = new FakeApiClient()
api.onList = () => Promise.resolve(ok({ items: [summary(S1)] as never[] }))

4
packages/client/ui-conversation/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/client/ui-conversation/README.md
README.md: a04c20f225c731581accbe8c12c52a5e7597029a
README.zh.md: f9e6a635ea6090c87a66f029785af214025b9bda
README.md: 51ddecf93240c2196483d3fb2bcfaca4104da31a
README.zh.md: d98cbcc69b875d2f426d9bdd9f2fa81874ec614a

2
packages/client/ui-conversation/README.md
View File

@@ -12,7 +12,7 @@ Generic tool rows classify the built-in bash, read, search, write, edit, and run
Tool rows are slots too — the standalone tool ring (`ToolViewRegistry`/`ctx.toolviews`/outlet) is retired. The chat entry declares the keyed `'conversation.chat.toolview'` hole (session scope; the key space is runtime-open); its render site dispatches per row via `entryKey: toolName` with `GenericToolCard` as the call-site `fallback`. The owner payload is the uniform `ToolRowOwnerProps` (`callId`/`toolName`/`block`/`openFile`) and `ToolRowProps` pre-composes it with the session standard kit. A registrant is a plain plugin: `ctx.slots.register({ name: 'conversation.chat.toolview', key: '<tool>', inject? }, Row)` with `inject: ['slots', 'conversation']` as the load-order seam (apply mounts ConversationService after the chat registration, so the service being present guarantees the slot is declared); session differentiation happens inside the component (`useSessions` reading `parentId` — the bash sample is the third-party-posture exemplar). Trajectory/waterfall toolview slots share this shape and land with their own render sites (RendersCheck rejects a declaration nobody renders).
The todo surfaces are two registrations over that shape, both plain registrant plugins with `inject: ['slots', 'conversation']`. `TodoRow` takes the `'conversation.chat.toolview'` key `todo_write` and summarizes what the call attempted (`<done>/<total> 已完成 · <active item>` parsed from its args, falling back to the generic summary on malformed or wrongly-shaped model JSON, and keeping the generic dot for non-ok execution states so a cancelled call never reads as a completed update). `TodoDock` takes the `'conversation.input.dock'` list slot at `order: -1` — above the queue rows — and is the durable plan strip: it selects `todos` off the session snapshot and renders `TodoPanel`, which takes the plain list, hides itself while the list is empty, and collapses to a header of title plus `"<done>/<total> tasks · <n> in progress"` (status glyphs are the figma check / progress / dashed-pending set). The dock adapter owns the selection so the panel stays a pure function of its props; the persistent list lives here rather than in the row so the row stays one line. Anything the input-zone composer chain hides (a `conversation.composer` takeover such as ui-question's) hides the whole dock, this strip included.
The todo surfaces are two registrations over that shape, both plain registrant plugins with `inject: ['slots', 'conversation']`. `TodoRow` takes the `'conversation.chat.toolview'` key `todo_write` and summarizes what the call attempted (`<done>/<total> 已完成 · <active item>` parsed from its args, falling back to the generic summary on malformed or wrongly-shaped model JSON, and keeping the generic dot for non-ok execution states so a cancelled call never reads as a completed update). `TodoDock` takes the `'conversation.input.dock'` list slot at `order: -1` — above the queue rows — and is the plan strip: it reads the host-computed `todos` projection via `useProjection` (standing plan: latest `todo/write` with no later `turn/start`) and renders `TodoPanel`, which takes the plain list, hides itself while the list is empty, and collapses to a header of title plus `"<done>/<total> tasks · <n> in progress"` (status glyphs are the figma check / progress / dashed-pending set). The dock adapter owns the selection so the panel stays a pure function of its props; the standing list lives here rather than in the row so the row stays one line. Anything the input-zone composer chain hides (a `conversation.composer` takeover such as ui-question's) hides the whole dock, this strip included.
Per-session UI state for selection and the active view lives in the declared chat store (`stores.ts` `createChatStore`); the InputHub owns the composer state machine and mirrors its draft into that store for persistence. Apply passes one store handle to the strict session subtree, chat view, and details registrations, so each session shares one instance and the framework owns its lifecycle. Components are pure: the framework standard kit supplies `useSession`/`sessionId`, global `useSessions`/`useWorkspaces`, and the input machine's `useInput`/`inputActions`; store faces and inject factories supply the remaining state and callbacks.

2
packages/client/ui-conversation/README.zh.md
View File

@@ -12,7 +12,7 @@
工具行同样是 slot：独立工具环（`ToolViewRegistry`／`ctx.toolviews`／outlet）已经退役。聊天配置项声明键控的 `'conversation.chat.toolview'` 空位（Session scope；key 空间在运行时开放）；其渲染点逐行通过 `entryKey: toolName` 分发，并以 `GenericToolCard` 作为调用点 `fallback`。owner 载荷是统一的 `ToolRowOwnerProps`（`callId`／`toolName`／`block`／`openFile`），`ToolRowProps` 则预先将其与 Session 标准工具包组合。注册方只是普通插件：`ctx.slots.register({ name: 'conversation.chat.toolview', key: '<tool>', inject? }, Row)`，以 `inject: ['slots', 'conversation']` 作为加载顺序 seam（apply 在聊天注册后挂载 ConversationService，因此服务存在即可保证 slot 已声明）；Session 区分在组件内部完成（`useSessions` 读取 `parentId`，bash 示例是第三方姿态的范例）。Trajectory/waterfall 工具视图 slot 共享此形状，并随各自的渲染点落地（RendersCheck 会拒绝没有任何渲染方的声明）。
todo 两个面就是在该形状上的两个注册项，都是普通注册方插件，`inject: ['slots', 'conversation']`。`TodoRow` 占用 `'conversation.chat.toolview'` 的 `todo_write` key，摘要该次调用「试图写入」的内容（从其 args 解析出 `<已完成>/<总数> 已完成 · <进行中条目>`；模型 JSON 残缺或形状不对时回落到通用摘要；非 ok 执行状态保留通用状态点，使被取消的调用绝不读成一次已完成的更新）。`TodoDock` 以 `order: -1` 占用 `'conversation.input.dock'` 列表 slot（位于队列行之上），是常驻的计划条：它从会话快照中选取 `todos` 并渲染 `TodoPanel`，后者接收纯列表，在列表为空时自我隐藏，折叠时收成标题加 `"<已完成>/<总数> tasks · <n> in progress"` 的表头（状态图标为 figma 的勾选／进行中／虚线未开始一组）。选取由 dock 适配器负责，因此面板保持为其 props 的纯函数；常驻列表放在此处而非行内，行才能保持单行。输入区 composer 链隐藏的一切（例如 ui-question 对 `conversation.composer` 的接管）也会隐藏整个 dock，包括这条计划条。
todo 两个面就是在该形状上的两个注册项，都是普通注册方插件，`inject: ['slots', 'conversation']`。`TodoRow` 占用 `'conversation.chat.toolview'` 的 `todo_write` key，摘要该次调用「试图写入」的内容（从其 args 解析出 `<已完成>/<总数> 已完成 · <进行中条目>`；模型 JSON 残缺或形状不对时回落到通用摘要；非 ok 执行状态保留通用状态点，使被取消的调用绝不读成一次已完成的更新）。`TodoDock` 以 `order: -1` 占用 `'conversation.input.dock'` 列表 slot（位于队列行之上），是计划条：它经 `useProjection` 读取 host 计算的 `todos` 投影（站立计划：其后没有更晚 `turn/start` 的最近一次 `todo/write`）并渲染 `TodoPanel`，后者接收纯列表，在列表为空时自我隐藏，折叠时收成标题加 `"<已完成>/<总数> tasks · <n> in progress"` 的表头（状态图标为 figma 的勾选／进行中／虚线未开始一组）。选取由 dock 适配器负责，因此面板保持为其 props 的纯函数；站立列表放在此处而非行内，行才能保持单行。输入区 composer 链隐藏的一切（例如 ui-question 对 `conversation.composer` 的接管）也会隐藏整个 dock，包括这条计划条。
逐 Session UI 状态中的选择与活跃视图位于已声明的聊天 store（`stores.ts` `createChatStore`）中；InputHub 拥有输入区状态机，并将草稿镜像到该 store 以便持久化。apply 将同一个 store handle 传给严格限定于会话的子树、聊天视图和详情注册，因此每个会话内共享一个实例，框架拥有其生命周期。组件保持纯粹：框架标准工具包提供 `useSession`／`sessionId`、全局 `useSessions`／`useWorkspaces`，以及输入状态机的 `useInput`／`inputActions`；store 表层与 inject factory 提供其余状态和回调。

12
packages/client/ui-conversation/src/client/skeleton/TodoPanel.tsx
View File

@@ -1,9 +1,9 @@
// TodoPanel: persistent plan strip above the composer (the web counterpart
// of the TUI plan panel). Renders the latest todo/write whole-list snapshot —
// no data of its own, hidden while the list is empty. Mounted through the
// 'conversation.input.dock' slot (QueueDock posture): the dock adapter does
// the selecting, so the panel takes the plain list and stays framework-free.
// Visual: figma 772:51905 (states) / 772:52972 (collapsed) / 772:53419 (expanded).
// TodoPanel: plan strip above the composer (the web counterpart of the TUI
// plan panel). Renders the standing todo/write whole-list snapshot (cleared on
// the next turn/start) — no data of its own, hidden while the list is empty.
// Mounted through the 'conversation.input.dock' slot (QueueDock posture): the
// dock adapter does the selecting, so the panel takes the plain list and stays
// framework-free. Visual: figma 772:51905 / 772:52972 / 772:53419.
import { useId, useState } from 'react'
import type { Context } from 'cordis'

50
packages/cordis/tool-cordis/src/api-catalog.ts
View File

@@ -518,11 +518,15 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
},
{
signature: 'abstract load(id: SessionId): Promise<{ meta: SessionHeader; events: SessionEvent[] }>',
jsDoc: '/**\n * Load a header and balanced contiguous log. A complete interrupted final\n * turn is preserved and durably closed with missing tool errors plus any open\n * step and turn boundaries; only a torn final record is discarded. Unknown\n * versions and corruption in the committed prefix reject. Implementations\n * MUST NOT crash-repair an identity still bound to a live Session: a balanced\n * live log may return with its stored header as a durable snapshot, while an\n * open live turn rejects.\n * A coordinator-backed cold load reserves the identity across storage awaits,\n * so concurrent publication of a same-id live Session rejects.\n * Returned events are detached, and every identified message is deeply\n * frozen; malformed identified messages reject before any stored event is returned.\n * @param id - the persisted session to reload.\n * @returns the header and a log ending on a balanced `turn/end`.\n */',
jsDoc: '/**\n * Load a header and balanced contiguous log. A complete interrupted final\n * turn is preserved and durably closed with missing tool errors plus any open\n * step and turn boundaries; only a torn final record is discarded. Unknown\n * versions and corruption in the committed prefix reject. Implementations\n * MUST NOT crash-repair an identity still bound to a live Session: a balanced\n * live log may return with its stored header as a durable snapshot, while an\n * open live turn rejects.\n * A coordinator-backed cold load reserves the identity across storage awaits,\n * so concurrent publication of a same-id live Session rejects.\n * Returned events are detached, and every identified message is deeply\n * frozen. Coordinator-backed implementations upgrade supported pre-identity\n * message events before validation; other malformed messages reject before\n * any stored event is returned.\n * @param id - the persisted session to reload.\n * @returns the header and a log ending on a balanced `turn/end`.\n */',
},
{
signature: 'abstract inspect(id: SessionId, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }>',
jsDoc: '/**\n * Inspect a header and its valid contiguous stored prefix without repairing\n * a torn tail, closing an interrupted turn, or publishing coordinator state.\n * This read is serialized with writes for the same id and returns detached\n * values with deeply frozen identified messages, so observers cannot mutate message\n * identity/content or backend-owned state. Malformed identified messages reject.\n * @param id - the persisted session to inspect.\n * @param signal - optional cancellation for queued and backend read work.\n * @returns the header and valid stored event prefix exactly as observed.\n */',
jsDoc: '/**\n * Inspect a header and its valid contiguous stored prefix without repairing\n * a torn tail, closing an interrupted turn, or publishing coordinator state.\n * This read is serialized with writes for the same id and returns detached\n * values with upgraded, deeply frozen identified messages, so observers\n * cannot mutate message identity/content or backend-owned state. Other\n * malformed messages reject.\n * @param id - the persisted session to inspect.\n * @param signal - optional cancellation for queued and backend read work.\n * @returns the header and valid stored event prefix exactly as observed.\n */',
},
{
signature: 'abstract readFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }>',
jsDoc: '/**\n * Read the stored events from `fromSeq` onward — the read-from-seq\n * primitive for read models that resume from a watermark (e.g. a persisted\n * projection cache folding only the tail past its checkpoint). Like\n * {@link inspect} it is non-mutating and detached: no torn-tail truncation,\n * no synthetic closers, no coordinator-state publication; only events from\n * the valid contiguous stored prefix are returned, so a torn fragment never\n * reaches the caller. `fromSeq` at or beyond the stored prefix returns an\n * empty event list (never an error). Backends whose medium can seek by seq\n * (SQLite) read only the suffix; sequential media (JSONL, both encodings)\n * still parse the whole artifact and skip forward — the primitive bounds\n * what is RETURNED and refolded, not every backend\'s physical read.\n * @param id - the persisted session to read.\n * @param fromSeq - first event seq to include; a non-negative safe integer.\n * @param signal - optional cancellation for queued and backend read work.\n * @returns the header and the stored events with `seq >= fromSeq`.\n */',
},
{
signature: 'abstract list(signal?: AbortSignal): Promise<SessionHeader[]>',
@@ -534,6 +538,24 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
},
],
},
{
key: 'sessionProjectionCache',
summary: 'The persisted projection cache service.',
methods: [
{
signature: 'cachedSnapshot(meta: SessionHeader): ProjectionSnapshot | undefined',
jsDoc: '/**\n * The zero-I/O listing read: whole values viewed straight from the stored\n * rows (version-matching keys only), each cut carried with its watermark\n * so a client value store can seed under its higher-seq-wins rule — as\n * stale as the last durable checkpoint but never wrong, and never from an\n * unrelated log (the caller\'s header is the identity witness). Fresher\n * paths (the history tail baseline, {@link coldSnapshot}) supersede these\n * values whenever a session is actually opened.\n * @param meta - the listed session\'s header (identity witness; no log read).\n * @returns the cut (`asOfSeq` = lowest served-row watermark), or\n * `undefined` when no usable row exists for this lifecycle.\n */',
},
{
signature: 'async write(session: Session): Promise<void>',
jsDoc: '/**\n * Durably checkpoint one live session NOW (both mandatory points call\n * this; tests and carriers may too). The registry cut is snapshotted at\n * this boundary (states are live references), then the whole record is\n * replaced. NOT fail-soft — callers on the fail-soft paths contain it.\n * @param session - the live session to checkpoint.\n * @returns resolution after durability and event emission.\n */',
},
{
signature: 'async coldSnapshot(id: SessionId, signal?: AbortSignal): Promise<ProjectionSnapshot>',
jsDoc: '/**\n * Cold-read one persisted session\'s projections with zero full-log load:\n * cached rows + a persistence `readFrom` tail from the registry\'s restore\n * floor, refolded by the registry and written back (fail-soft) so the next\n * cold read starts closer. A cache row invalidated by a shrunk log\n * (crash-repair truncation) triggers one full re-read from seq 0 — the\n * ladder\'s slow rung, still no crash. Rejects when the session has no\n * persisted log (`not found` from the persistence seam).\n * @param id - the persisted session to read.\n * @param signal - optional cancellation for the persistence reads.\n * @returns the snapshot cut at the stored log end.\n */',
},
],
},
{
key: 'sessionProjections',
summary: '`ctx.sessionProjections`: the projection unit table and its drive.',
@@ -550,6 +572,22 @@ export const SERVICE_API: readonly ServiceApiEntry[] = [
signature: 'snapshot(session: Session): ProjectionSnapshot',
jsDoc: '/**\n * One consistent cut over every registered unit for one session, read from\n * the watermark cache (missing cells fold lazily over the in-memory log).\n * Fully synchronous — every value and `asOfSeq` reflect the same log\n * position. Each value passes its unit\'s schema before leaving.\n * @param session - the session whose projection values are read.\n * @returns the snapshot; `values` is empty when no unit is registered.\n */',
},
{
signature: 'checkpoint(session: Session): ProjectionCheckpoint',
jsDoc: '/**\n * State-level checkpoint of every registered unit for one session, read\n * from the watermark cache (missing cells fold lazily over the in-memory\n * log). This is the write side of the persisted projection cache: the\n * returned rows are the `(key → {ver, seq, val})` part of the durable\n * `(sessionId, key, ver, seq, val)`\n * rows. Every `val` is a DETACHED structured clone — never the live\n * cell reference: the watermark cache is this registry\'s authoritative\n * mutable state, and a caller reaching the live reference could corrupt\n * every subsequent snapshot and frame through it (plain JSON by the unit\n * contract, so the clone is total).\n * @param session - the session whose unit states are checkpointed.\n * @returns one row per registered key; empty when no unit is registered.\n */',
},
{
signature: 'restoreFloor(checkpoint: ProjectionCheckpoint): number | undefined',
jsDoc: '/**\n * The stored seq a {@link restore} tail read over `checkpoint` must start\n * at: one event BELOW the lowest usable watermark (a row is usable when\n * its `ver` matches the live unit\'s `stateVersion`; an absent or mismatched row\n * pulls the floor to `0` — that key must refold the full log). The\n * one-below anchor is load-bearing: the tail then proves how far the\n * stored log still extends, so {@link restore} can detect a log that\n * shrank below a row\'s watermark (crash-repair truncation) instead of\n * serving the stale row as current — an empty tail read from the anchor\n * yields an end below every watermark and the restore rejects for a full\n * re-read.\n * @param checkpoint - persisted rows for one session (possibly stale or empty).\n * @returns the seq to hand the persistence `readFrom`, or `undefined`\n * when no unit is registered (no read needed — {@link restore} would\n * serve empty values regardless).\n */',
},
{
signature: 'viewCheckpoint(checkpoint: ProjectionCheckpoint): Partial<SessionProjectionMap>',
jsDoc: '/**\n * View a checkpoint\'s rows without any log read: for every registered\n * unit whose row\'s `ver` matches, serve the schema-validated\n * `view` of the stored state; mismatched or absent rows leave their key\n * absent (a cold or listing consumer treats it as not-yet-available and a\n * fuller read path refolds it). The zero-I/O rung of the read ladder —\n * values are as stale as their rows, never wrong.\n * @param checkpoint - persisted rows for one session (possibly stale or empty).\n * @returns whole values per key with a usable row; empty when none.\n */',
},
{
signature: 'restore(checkpoint: ProjectionCheckpoint, events: readonly SessionEvent[], baseSeq: number): { snapshot: ProjectionSnapshot; checkpoint: ProjectionCheckpoint }',
jsDoc: '/**\n * Cold read: fold every registered unit over a stored log suffix, seeding\n * each from its checkpoint row when usable — the one read recipe (cached\n * state + forward tail replay + `view`) applied without a live `Session`.\n * Call with the events returned by a persistence\n * `readFrom(id, restoreFloor(checkpoint))` and that same floor as\n * `baseSeq`; the floor\'s one-below anchor makes the supplied end honest,\n * so a shrunk log is detected here. A row is usable iff its\n * `ver` matches the live unit\'s `stateVersion`, it does not predate `baseSeq`\n * (`seq >= baseSeq - 1`), and it does not claim events past the\n * supplied end (`seq <= endSeq`); an unusable row is discarded\n * and its key refolds from `init` — which is only sound over the full\n * log, so a discarded row with `baseSeq > 0` throws (the caller re-reads\n * from seq 0, e.g. after a crash-repair truncation shrank the log below\n * a row\'s watermark).\n * @param checkpoint - persisted rows for one session (possibly stale or empty).\n * @param events - the stored events with `seq >= baseSeq`, in seq order.\n * @param baseSeq - the seq `events` starts at (its first event\'s seq when non-empty).\n * @returns the snapshot cut at the supplied log end (`asOfSeq` is the last\n * supplied event\'s seq, `baseSeq - 1` for an empty tail) plus the\n * refreshed checkpoint rows at that cut, ready for a durable write-back.\n */',
},
],
},
{
@@ -1855,6 +1893,14 @@ export const TYPE_API: readonly TypeApiEntry[] = [
name: 'ProjectionChangeListener',
declaration: 'export type ProjectionChangeListener = (session: Session, key: Extract<keyof SessionProjectionMap, string>, value: unknown, seq: number) => void;',
},
{
name: 'ProjectionCheckpoint',
declaration: 'export type ProjectionCheckpoint = Record<string, ProjectionCheckpointRow>;',
},
{
name: 'ProjectionCheckpointRow',
declaration: 'export interface ProjectionCheckpointRow {\n ver: number;\n seq: number;\n val: unknown;\n}',
},
{
name: 'ProjectionDefinition',
declaration: 'export interface ProjectionDefinition<K extends keyof SessionProjectionMap, S> {\n key: K;\n schema: ZodType<SessionProjectionMap[K]>;\n init(): S;\n apply(state: S, event: SessionEvent): S;\n view(state: S): SessionProjectionMap[K];\n stateVersion: number;\n}',

61
packages/core/agent-loop/tests/resume.spec.ts
View File

@@ -5,7 +5,7 @@ import { mkdtemp, rm } from 'node:fs/promises'
import { tmpdir } from 'node:os'
import { join } from 'node:path'
import LlmService from '@deepseek-ai/dsh-llm'
import SessionStore, { Session, SessionId } from '@deepseek-ai/dsh-session'
import SessionStore, { SESSION_FORMAT_VERSION, Session, SessionId } from '@deepseek-ai/dsh-session'
import type { SessionEvent } from '@deepseek-ai/dsh-session'
import SystemPrompt from '@deepseek-ai/dsh-system-prompt'
import ToolRegistry from '@deepseek-ai/dsh-tools'
@@ -77,6 +77,65 @@ function throwUnknown(value: unknown): never {
}
describe('the session-persistence Agent Note: AgentLoop factory create/resume', () => {
it('resumes a session persisted before messages gained identities', async () => {
const sessionId = SessionId('pre-identity-resume')
const first = await persistentHarness(new MockAdapter([]))
await first.ctx.sessionPersistence.create({
version: SESSION_FORMAT_VERSION,
id: sessionId,
createdAt: 1,
})
await first.ctx.sessionPersistence.append(sessionId, [
{ type: 'turn/start', seq: 0, time: 1, data: { turn: 1, trigger: { kind: 'message', source: { kind: 'user' } } } },
{
type: 'user/message',
seq: 1,
time: 2,
data: { content: [{ type: 'text', text: 'old question' }], source: { kind: 'user' } },
surfaceOp: 'append',
},
{ type: 'step/start', seq: 2, time: 3, data: { turn: 1, step: 1 } },
{
type: 'assistant/message',
seq: 3,
time: 4,
data: {
turn: 1,
step: 1,
content: [{ type: 'text', text: 'old answer' }],
provenance: { provider: 'mock', model: 'mock' },
},
surfaceOp: 'append',
},
{ type: 'step/end', seq: 4, time: 5, data: { turn: 1, step: 1 } },
{ type: 'turn/end', seq: 5, time: 6, data: { turn: 1, reason: { kind: 'completed' } } },
] as unknown as SessionEvent[])
await first.ctx.fiber.dispose()
const ctx = await mountPersistentHarness(first.root, new MockAdapter([textResponse('new answer')]))
const handle = await ctx.agents.resume({
resumeSessionId: sessionId,
agentOptions: { provider: 'mock', model: 'mock' },
})
expect(handle.agent.session.deriveMessages()).toMatchObject([
{ id: `legacy-message:${sessionId}:1`, role: 'user' },
{ id: `legacy-message:${sessionId}:3`, role: 'assistant' },
])
handle.agent.followup(createUserMessage({
content: [{ type: 'text', text: 'new question' }],
source: { kind: 'user' },
}))
await waitForIdle(ctx, handle.agent)
expect(handle.agent.session.deriveMessages()).toHaveLength(4)
expect(handle.agent.session.events.at(-1)).toMatchObject({
type: 'turn/end',
data: { reason: { kind: 'completed' } },
})
await handle.dispose()
await ctx.fiber.dispose()
})
it('normalizes a non-Error resume publication failure for rollback and rethrows it', async () => {
const sessionId = SessionId('unknown-resume-failure-s')
const root = await persistSession(sessionId)

4
packages/core/session/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/core/session/README.md
README.md: 40516d12180de9c30efd40fdffa873da20ddacb3
README.zh.md: 43842643a3434c741f219f7b6c26622cddfae8e7
README.md: a9b6905dcf2b8ef1f75595e567273f7a3150a412
README.zh.md: f1a5e97e32d1ad1abcd6ad96e6c621af9972e989

2
packages/core/session/README.md
View File

@@ -142,5 +142,5 @@ Logging causes no invalidation, and exact reconstruction preserves request-prefi
- **Session branching/tree** (pi-style entry tree) — deferred unless needed beyond boundary-based `fork()`.
- **`fork()` cuts only at stable boundaries of live sessions** — the selected prefix must end outside an open turn and the source must be in the store; forking a persisted-but-unloaded session is excluded from the [fork API](../../../.agents/notes/implemented/feature/2026-06-30-session-store-fork-api.md).
- **`SESSION_FORMAT_VERSION` stays pinned at `0`** — pre-release, no compatibility implied: a backend rejects any other version, and no migration path exists until the first release ([policy](../../../AGENTS.md)).
- **`SESSION_FORMAT_VERSION` stays pinned at `0`** — pre-release, no broad compatibility implied: `Session` accepts only current seed shapes and a backend rejects any other version. Narrow storage import upgrades belong to the persistence boundary ([policy](../../../AGENTS.md), [pre-identity message recovery](../../../.agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.md)).
- **`TurnEndReasonMap` omits the ACP-named `refusal` / `max_turn_requests` variants** — producer-gated: they land when an adapter or the loop first emits them.

2
packages/core/session/README.zh.md
View File

@@ -142,5 +142,5 @@
- **会话分支／树**（pi 风格条目树）：除非需要超越基于边界的 `fork()` 能力，否则暂缓。
- **`fork()` 仅在实时会话的稳定边界处切分**：所选前缀结束时不得有开放轮次，且源会话必须位于存储中；[fork API](../../../.agents/notes/implemented/feature/2026-06-30-session-store-fork-api.md) 不支持对已持久化但未加载的会话进行 fork。
- **`SESSION_FORMAT_VERSION` 固定为 `0`**：预发布阶段不承诺兼容性；后端会拒绝其他任何版本，首次发布前不提供迁移路径（[政策](../../../AGENTS.md)）。
- **`SESSION_FORMAT_VERSION` 固定为 `0`**：预发布阶段不承诺广泛兼容性；`Session` 只接受当前 seed 形状，后端会拒绝其他任何版本。范围受限的存储导入升级应由持久化边界负责（[政策](../../../AGENTS.md)、[消息标识机制引入前的消息恢复](../../../.agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.md)）。
- **`TurnEndReasonMap` 不含 ACP（Agent Client Protocol）命名的 `refusal`／`max_turn_requests` 变体**：受生产方约束；只有当适配器或循环首次产生这些变体时才加入。

4
packages/host/apiproxy/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/host/apiproxy/README.md
README.md: 6b4691164ceb82a68c94e5a71d2f44fcfa1634af
README.zh.md: b28b57097a251ee18295a3f9a1ae41b00efe2db2
README.md: c20887b73b9b9deb278db30d34d84df07257d664
README.zh.md: 18a2f97477e5f57127371429b0dd59ba01f04341

2
packages/host/apiproxy/README.md
View File

@@ -22,8 +22,6 @@ Workspace and Session lists are separate reconnect baselines. `workspace.create`
`host.openPath` opens a filesystem path with the operating system's default application (`open` on macOS, `Invoke-Item` on Windows, `xdg-open` on Linux). The opener is injectable for tests. The browser carrier applies the same loopback, same-origin restriction as `host.pickDirectory`.
`session.history` pages on message boundaries, and its tail page (no `beforeSeq`) carries two session-level extras the page window cannot supply: the in-flight partial's chunk events, and `todos` — the latest `todo/write` whole-list projection over the full log. Older pages omit `todos` because the projection is session-level, not per-page; a tail response that omits it means the whole log holds no `todo/write`, so clients read the absent field as the empty plan rather than as unchanged state.
The `command.*` and `skill.*` domains expose the host command registry and skill catalog to clients. Every method addresses one session's agent by `sessionId` (a served session always has an Agent; `command.*` resumes cold sessions through the same path as `session.*`, while `skill.list` resolves the project root from the session header without touching the Agent registry). `command.execute` runs a slash-command line host-side with pure admission semantics: the response reports whether the line resolved to a handler plus the minted lifecycle `commandId` when it did (correlating the acknowledgment with the flow node), while the outcome rides the durably logged `command/run`/`command/done` lifecycle pair broadcast on the mux stream; the carrier's request signal cancels the running handler. `host/commands-changed` is the catalog invalidation frame: clients refetch `command.list` instead of diffing.
## Carrier layer (`/client` + root)

2
packages/host/apiproxy/README.zh.md
View File

@@ -22,8 +22,6 @@ Workspace 列表与 Session 列表是相互独立的重连基线。`workspace.cr
`host.openPath` 会用操作系统的默认应用打开一个文件系统路径（macOS 为 `open`，Windows 为 `Invoke-Item`，Linux 为 `xdg-open`）。打开器可在测试中注入。浏览器载体对其施加与 `host.pickDirectory` 相同的回环、同源限制。
`session.history` 按消息边界分页，其尾页（不带 `beforeSeq`）额外携带两项页窗口本身无法提供的会话级数据：进行中局部消息的 chunk 事件，以及 `todos`——整份日志上最后一次 `todo/write` 的整表投影。较早的页面不带 `todos`，因为该投影是会话级而非分页级的；尾页响应缺少该字段意味着整份日志中没有任何 `todo/write`，因此客户端要把缺失字段读作空计划，而不是读作「状态未变」。
`command.*` 与 `skill.*` 领域向客户端暴露宿主命令注册表和技能目录。每个方法都通过 `sessionId` 寻址一个会话的 Agent（被服务的会话必有 Agent；`command.*` 经由与 `session.*` 相同的路径恢复冷会话，而 `skill.list` 从会话头解析项目根目录，不触碰 Agent 注册表）。`command.execute` 在宿主侧运行一条斜杠命令行，语义为纯准入：响应报告该行是否解析到处理器，并在解析到时回带铸造的生命周期 `commandId`（将本次确认与流节点关联）；结局经由持久落账并在 mux 流广播的 `command/run`/`command/done` 生命周期事件对承载；载体的请求信号可取消正在运行的处理器。`host/commands-changed` 是目录失效帧：客户端重新拉取 `command.list` 而不是做差分。
## 载体层（`/client` + 根路径）

1
packages/host/apiproxy/package.json
View File

@@ -47,6 +47,7 @@
"@deepseek-ai/dsh-session": "workspace:^",
"@deepseek-ai/dsh-session-persistence": "workspace:^",
"@deepseek-ai/dsh-session-projection": "workspace:^",
"@deepseek-ai/dsh-session-projection-cache": "workspace:^",
"@deepseek-ai/dsh-skill": "workspace:^",
"@deepseek-ai/dsh-tools": "workspace:^",
"@deepseek-ai/dsh-user-approval": "workspace:^",

40
packages/host/apiproxy/src/api-proxy.ts
View File

@@ -30,6 +30,8 @@ import type {
} from './api/index.ts'
// Type-only: resolves `ctx.get('sessionProjections')` to the projection registry.
import type {} from '@deepseek-ai/dsh-session-projection'
// Type-only: resolves `ctx.get('sessionProjectionCache')` (the cold listing column).
import type {} from '@deepseek-ai/dsh-session-projection-cache'
// Type-only edges: resolve `ctx.get('commands')`, the `commands/change` event, and `ctx.get('skills')`.
import type {} from '@deepseek-ai/dsh-commands'
import type {} from '@deepseek-ai/dsh-skill'
@@ -297,6 +299,28 @@ function projectionsFor(ctx: Context, agent: Agent): SessionProjectionsBlock | u
return registry.snapshot(agent.session)
}
/**
* The projection baseline of one session.list row, fail-soft: attached
* sessions cut the registry's live watermark cache; cold sessions view the
* persisted projection cache's identity-checked stored rows (zero log loads
* either way — the listing use case the cache exists for). The block shape
* (values + asOfSeq) matches the history tail's, so a client seeds its
* value store under the same higher-seq-wins rule. Any failure — and an
* empty value set — yields an absent block: a listing without projections
* is degraded, never broken.
*/
function listProjectionsFor(ctx: Context, meta: SessionHeader, session: Session | undefined): SessionProjectionsBlock | undefined {
try {
const block = session !== undefined
? ctx.get('sessionProjections')?.snapshot(session)
: ctx.get('sessionProjectionCache')?.cachedSnapshot(meta)
return block !== undefined && Object.keys(block.values).length > 0 ? block : undefined
} catch (error) {
ctx.logger.warn(`session.list: projection column for "${meta.id}" failed (serving the row without it): ${String(error)}`)
return undefined
}
}
/**
* Thrown by the cold-resume path when the id names no servable session
* (absent from the store, or a pre-project legacy log without a cwd).
@@ -654,13 +678,25 @@ export function createApiProxy(ctx: Context, defaults: ApiProxyDefaults): ApiPro
async list(request) {
const items = ctx.sessions.list().map((session) => {
const agent = ctx.agents.get(session.id)
return summarize(session, agent?.status === 'running')
const projections = listProjectionsFor(ctx, session.header, session)
return {
...summarize(session, agent?.status === 'running'),
...projections === undefined ? {} : { projections },
}
})
const attached = new Set(items.map(item => item.sessionId))
const persistence = ctx.get('sessionPersistence')
if (persistence !== undefined) {
const cold = (await persistence.list()).filter(meta => !attached.has(meta.id) && meta.cwd !== undefined)
items.push(...await Promise.all(cold.map(meta => summarizeCold(persistence, meta))))
items.push(...await Promise.all(cold.map(async (meta) => {
// Cold rows read the persisted projection cache only — never a
// log load; a session without a cache row simply has no column.
const projections = listProjectionsFor(ctx, meta, undefined)
return {
...await summarizeCold(persistence, meta),
...projections === undefined ? {} : { projections },
}
})))
}
items.sort((a, b) => b.updatedAt - a.updatedAt)
return ok(request, { items })

9
packages/host/apiproxy/src/api/sessions.schema.ts
View File

@@ -37,7 +37,7 @@ export const sessionEventSchema = z.object({
surfaceOp: z.unknown().optional(),
}) as unknown as z.ZodType<SessionEvent>
/** SessionSummary row of session.list. */
/** SessionSummary row of session.list (`projections` reuses the history block's shape and schema). */
export const sessionSummarySchema = z.object({
sessionId: sessionIdSchema,
updatedAt: z.number(),
@@ -45,7 +45,8 @@ export const sessionSummarySchema = z.object({
blank: z.boolean(),
parentSessionId: sessionIdSchema.optional(),
cwd: z.string().optional(),
}) satisfies z.ZodType<Wire<SessionSummary>>
projections: z.lazy(() => sessionProjectionsBlockSchema).optional(),
}) as unknown as z.ZodType<Wire<SessionSummary>>
/** session.list request payload (cursor is a reserved seat, unimplemented in v1). */
export const sessionListRequestSchema = z.object({
@@ -53,9 +54,9 @@ export const sessionListRequestSchema = z.object({
}) satisfies z.ZodType<Wire<RequestPayload<'session.list'>>>
/** session.list response value. */
export const sessionListValueSchema = z.object({
export const sessionListValueSchema: z.ZodType<Wire<ResponseValue<'session.list'>>> = z.object({
items: z.array(sessionSummarySchema),
}) satisfies z.ZodType<Wire<ResponseValue<'session.list'>>>
})
/** session.create request payload (at most one of workspaceId / cwd). */
export const sessionCreateRequestSchema = z.object({

12
packages/host/apiproxy/src/api/sessions.ts
View File

@@ -143,6 +143,18 @@ export interface SessionSummary {
parentSessionId?: SessionId
/** Session working directory (header.cwd passthrough); absent when unrecorded. */
cwd?: string
/**
* Projection baseline for this row, with zero log loads: attached sessions
* read the registry's live watermark cut; cold sessions read the persisted
* projection cache's stored rows — as stale as that session's last durable
* checkpoint (`asOfSeq` says exactly how stale), never wrong, and directly
* seedable into the client's per-session value store under its
* higher-seq-wins rule (a list baseline can never overwrite a newer push
* frame). Absent when no value is available (no registry, no cache row for
* a cold session, or a fail-soft cache read miss); a listing client treats
* absence as "no title yet", exactly like a blank session.
*/
projections?: SessionProjectionsBlock
}
/** Session-domain unary methods (the map keys session.* of RpcMethodMap). */

78
packages/host/apiproxy/tests/api-proxy-projections.spec.ts
View File

@@ -13,7 +13,7 @@ import { z } from 'zod'
import AgentRegistry from '@deepseek-ai/dsh-agent'
import type { Agent } from '@deepseek-ai/dsh-agent'
import { createUserMessage } from '@deepseek-ai/dsh-llm'
import SessionStore from '@deepseek-ai/dsh-session'
import SessionStore, { SessionId } from '@deepseek-ai/dsh-session'
import type { Session } from '@deepseek-ai/dsh-session'
import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection'
import type { ProjectionDefinition } from '@deepseek-ai/dsh-session-projection'
@@ -125,6 +125,82 @@ describe('session.history projections block', () => {
})
})
describe('session.list projections column', () => {
it('serves attached rows from the live registry cut, watermarked for client seeding', async () => {
const { ctx, session } = await harness(true)
ctx.sessionProjections.register(lastUserUnit())
seedMessages(session, 1)
const response = await api(ctx).sessions.list(request({}))
if (!response.result.ok) throw new Error('unreachable')
const row = response.result.value.items.find(item => item.sessionId === session.id)
expect(row?.projections?.values['test/last-user']).toEqual({ text: 'm0' })
expect(row?.projections?.asOfSeq).toBe(session.seq - 1)
})
it('omits the column entirely when no registry is mounted', async () => {
const { ctx, session } = await harness(false)
seedMessages(session, 1)
const response = await api(ctx).sessions.list(request({}))
if (!response.result.ok) throw new Error('unreachable')
const row = response.result.value.items.find(item => item.sessionId === session.id)
expect(row).toBeDefined()
expect(row !== undefined && 'projections' in row).toBe(false)
})
it('serves cold rows from the persisted projection cache with zero log loads', async () => {
const { ctx } = await harness(true)
const coldId = SessionId('session-cold-listing')
const load = () => { throw new Error('list must not load event logs') }
ctx.provide('sessionPersistence', {
list: async () => [{ version: 0, id: coldId, createdAt: 5, cwd: '/tmp' }],
locate: () => undefined,
load,
inspect: load,
readFrom: load,
} as never)
ctx.provide('sessionProjectionCache', {
// The carrier hands the listed header through as the identity witness.
cachedSnapshot: (meta: { id: unknown; createdAt: number }) =>
(meta.id === coldId && meta.createdAt === 5
? { asOfSeq: 7, values: { 'test/last-user': { text: 'cached' } } }
: undefined),
} as never)
const response = await api(ctx).sessions.list(request({}))
if (!response.result.ok) throw new Error('unreachable')
const row = response.result.value.items.find(item => item.sessionId === coldId)
expect(row?.running).toBe(false)
expect(row?.projections).toEqual({ asOfSeq: 7, values: { 'test/last-user': { text: 'cached' } } })
})
it('cold rows without a cache plugin (or without a stored row) just lack the column', async () => {
const { ctx } = await harness(true)
const coldId = SessionId('session-cold-uncached')
ctx.provide('sessionPersistence', {
list: async () => [{ version: 0, id: coldId, createdAt: 5, cwd: '/tmp' }],
locate: () => undefined,
} as never)
const response = await api(ctx).sessions.list(request({}))
if (!response.result.ok) throw new Error('unreachable')
const row = response.result.value.items.find(item => item.sessionId === coldId)
expect(row).toBeDefined()
expect(row !== undefined && 'projections' in row).toBe(false)
})
it('a throwing column read degrades that row, never the listing', async () => {
const { ctx, session } = await harness(true)
ctx.sessionProjections.register({
...lastUserUnit(),
view: () => { throw new Error('unit exploded') },
})
seedMessages(session, 1)
const response = await api(ctx).sessions.list(request({}))
if (!response.result.ok) throw new Error('unreachable')
const row = response.result.value.items.find(item => item.sessionId === session.id)
expect(row).toBeDefined()
expect(row !== undefined && 'projections' in row).toBe(false)
})
})
describe('session/projection push frame', () => {
/** Drain frames until `count` session/projection frames arrived. */
async function collect(iterable: AsyncIterable<RpcRequest<MuxFrame>>, count: number, abort: AbortController): Promise<MuxFrame[]> {

3
packages/host/apiproxy/tsconfig.json
View File

@@ -35,6 +35,9 @@
{
"path": "../../session-projection/session-projection"
},
{
"path": "../../session-projection/session-projection-cache"
},
{
"path": "../../skill/skill"
},

3
packages/session-persistence/session-checkpoint-policy/tests/session-checkpoint-policy.spec.ts
View File

@@ -22,6 +22,9 @@ class TestPersistence extends SessionPersistence {
inspect(_id: SessionId): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
return Promise.reject(new Error('not used'))
}
readFrom(_id: SessionId, _fromSeq: number): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
return Promise.reject(new Error('not used'))
}
list(): Promise<SessionHeader[]> { return Promise.resolve([]) }
listSnapshots(): Promise<never[]> { return Promise.resolve([]) }
}

6
packages/session-persistence/session-persistence-jsonl/src/index.ts
View File

@@ -134,6 +134,12 @@ export class SessionPersistenceJsonl extends SessionPersistence implements Persi
return this.coordinator.inspect(id, signal)
}
// JSONL is sequential media: no loadStoredFrom hook, so the coordinator
// parses the stored prefix (both encodings) and skips forward to fromSeq.
readFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
return this.coordinator.readFrom(id, fromSeq, signal)
}
// One method serves both public `list` and the backend hook; delegating it to
// the coordinator would call this hook recursively.

26
packages/session-persistence/session-persistence-sqlite/src/index.ts
View File

@@ -16,7 +16,7 @@ import { dirname, resolve } from 'node:path'
import {
SessionPersistence, SessionPersistenceRevision, PersistenceCoordinator,
type PersistenceBackend, type SessionLocation, type SessionPersistenceSnapshot,
type StoredPrefix,
type StoredPrefix, type StoredSuffix,
} from '@deepseek-ai/dsh-session-persistence'
import type { SessionEvent, SurfaceEventType, SessionId, SessionHeader } from '@deepseek-ai/dsh-session'
import {
@@ -161,6 +161,10 @@ export class SessionPersistenceSqlite extends SessionPersistence implements Pers
return this.coordinator.inspect(id, signal)
}
readFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
return this.coordinator.readFrom(id, fromSeq, signal)
}
// One method serves both public `list` and the backend hook; delegating it to
// the coordinator would call this hook recursively.
@@ -171,6 +175,26 @@ export class SessionPersistenceSqlite extends SessionPersistence implements Pers
return this.readPrefix(id, signal)
}
/**
* Seek-capable suffix read: SQL selects `seq >= fromSeq` directly, so the
* read scales with the suffix, not the log. Torn rows past the preserved
* region are dropped, never repaired (non-mutating read).
*/
async loadStoredFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<StoredSuffix | undefined> {
signal?.throwIfAborted()
await this.ready
signal?.throwIfAborted()
const row = this.rowFor(id)
if (row === undefined) return undefined
const meta = rowToMeta(row)
const eventRows = this.db
.prepare('SELECT seq, type, time, data, source_event_seqs, surface_op FROM events WHERE session_id = ? AND seq >= ? ORDER BY seq')
.all(id, fromSeq) as unknown as EventRow[]
signal?.throwIfAborted()
const { preserved } = scanRows(eventRows, fromSeq)
return { meta, events: preserved }
}
/**
* Read a session's row + ordered events into a {@link StoredPrefix}. The
* torn-tail marker is the seq from which a never-committed tail must be deleted

10
packages/session-persistence/session-persistence-sqlite/src/schema.ts
View File

@@ -213,10 +213,12 @@ export function rowToEvent(row: EventRow): SessionEvent {
* the committed region rejects.
*
* @param rows - one session's event rows, ordered by seq ascending.
* @param base - the seq the first row is expected to carry; `0` for a whole
* log, the requested `fromSeq` for a suffix read (`loadStoredFrom`).
* @returns the preserved event prefix, plus `tornFrom` — the seq the physical
* delete starts at — when a torn tail exists.
*/
export function scanRows(rows: readonly EventRow[]): { preserved: SessionEvent[]; tornFrom?: number } {
export function scanRows(rows: readonly EventRow[], base = 0): { preserved: SessionEvent[]; tornFrom?: number } {
// Pass 1: parse each row's data; a row whose data is not valid JSON is a hole.
// (The seq/type COLUMNS are always present even when `data` is corrupt.)
interface Parsed { ok: boolean; event?: SessionEvent }
@@ -244,8 +246,8 @@ export function scanRows(rows: readonly EventRow[]): { preserved: SessionEvent[]
if (i <= lastTurnEnd) throw new Error(`corrupt session log: unparsable committed event at seq ${rows[i]?.seq}`)
break // torn tail fragment after the last turn/end — stop, tolerate
}
if (p.event.seq !== i) {
if (i <= lastTurnEnd) throw new Error(`corrupt session log: seq gap in committed region (expected ${i}, got ${p.event.seq})`)
if (p.event.seq !== base + i) {
if (i <= lastTurnEnd) throw new Error(`corrupt session log: seq gap in committed region (expected ${base + i}, got ${p.event.seq})`)
break // gap after the last turn/end — torn tail, stop
}
preserved.push(p.event)
@@ -253,5 +255,5 @@ export function scanRows(rows: readonly EventRow[]): { preserved: SessionEvent[]
// Any rows past the preserved prefix are a never-committed torn tail; their
// first seq is the deletion point for load's physical repair.
return preserved.length < rows.length ? { preserved, tornFrom: preserved.length } : { preserved }
return preserved.length < rows.length ? { preserved, tornFrom: base + preserved.length } : { preserved }
}

4
packages/session-persistence/session-persistence/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/session-persistence/session-persistence/README.md
README.md: 08d8adac8040747a6dac01dbc41525073f17060c
README.zh.md: 7676f27a1aa934eb3472e1b32b9ecd55d460fb63
README.md: a8a4f14c8613a7e51bcf467e816b7f7bdb7ea80b
README.zh.md: 369e8a01b8ac411ed9acfbac86b9b34db8037c1f

8
packages/session-persistence/session-persistence/README.md
View File

@@ -13,8 +13,9 @@ The persisted unit IS the existing `SessionEvent` (event-sourced model — the l
| `locate(meta): SessionLocation \| undefined` | Resolve an absolute per-session artifact target without I/O or materialization. Backends without an independent local artifact return `undefined`. |
| `create(meta): Promise<void>` | Register a new session's metadata. MAY defer the physical write until the first `append` (lazy materialization). |
| `append(id, events): Promise<void>` | Durably persist a batch. Append-only; first event `seq` == stored next-seq after any repair; rejects non-JSON-serializable data naming the offending type. |
| `load(id): Promise<{ meta; events }>` | Return a stored header plus a balanced contiguous log whose events are detached and validated and whose identified messages are deeply frozen. A live load first flushes its snapshot and rejects while its turn is open; a cold load preserves an interrupted final turn and closes it with synthetic `tool/result`/`step/end?`/`turn/end {interrupted}` events. Only a torn tail fragment is dropped; committed corruption, malformed messages, and unknown `version` reject. |
| `inspect(id, signal?): Promise<{ meta; events }>` | Return a detached valid stored prefix with validated, deeply frozen identified messages, without truncating a torn tail, synthesizing recovery closers, or publishing coordinator state. Serialized with same-id writes; the optional signal promptly rejects a queued caller, prevents that queued backend read from starting, and cancels active backend read work. Intended for read models and other observers that must never recover a log. |
| `load(id): Promise<{ meta; events }>` | Return a stored header plus a balanced contiguous log whose events are detached and validated and whose identified messages are deeply frozen. The coordinator upgrades the four pre-identity message event shapes into current wrappers in the returned snapshot; all other obsolete or malformed shapes still reject. A live load first flushes its snapshot and rejects while its turn is open; a cold load preserves an interrupted final turn and closes it with synthetic `tool/result`/`step/end?`/`turn/end {interrupted}` events. Only a torn tail fragment is dropped; committed corruption and unknown `version` reject. |
| `inspect(id, signal?): Promise<{ meta; events }>` | Return a detached valid stored prefix with upgraded, validated, deeply frozen identified messages, without truncating a torn tail, synthesizing recovery closers, or publishing coordinator state. Serialized with same-id writes; the optional signal promptly rejects a queued caller, prevents that queued backend read from starting, and cancels active backend read work. Intended for read models and other observers that must never recover a log. |
| `readFrom(id, fromSeq, signal?): Promise<{ meta; events }>` | The read-from-seq primitive: return the header plus the valid stored events with `seq >= fromSeq`, detached and non-mutating like `inspect` (no truncation, no closers, no coordinator state). A `fromSeq` at or past the stored end returns an empty event list; a negative or non-safe-integer `fromSeq` rejects. Seek-capable backends (SQLite) read only the suffix; sequential backends (JSONL) still parse the whole artifact and skip forward — the primitive bounds what is returned and refolded, not every backend's physical read. Intended for checkpoint consumers (e.g. the persisted projection cache) that fold only the tail past a watermark. |
| `list(signal?): Promise<SessionHeader[]>` | Lightweight listing from metadata, no full-log parse. The optional signal cancels backend listing work. A zero-event lazily-materialized session is absent from `list`. |
| `listSnapshots(signal?): Promise<SessionPersistenceSnapshot[]>` | Lightweight metadata plus an opaque branded per-log revision, without loading event logs. A revision stays equal while that log and its backing store are unchanged, changes after append or mutating load repair, and cannot collide solely because two stores use the same local counter. The optional signal requests cancellation of backend discovery work; first-party backends settle any started listing work before rejecting so an awaited call is quiescent. |
@@ -33,6 +34,8 @@ Each `session/event` copies its event into the session controller and starts an
Crash repair is cold-only. For a live id, `load(id)` snapshots the authoritative in-memory log, waits for that snapshot to become durable, and returns it with the coordinator's stored header only when balanced; an open live turn rejects instead of receiving synthetic interruption closers. A cold load reserves its id across backend reads and repair writes, so concurrent publication of a same-id live `Session` rejects and rolls back. HMR adoption reads through `loadStored`, applies the coordinator's cwd check, and never closes the active turn.
Backend reads normalize pre-identity `user/message`, `assistant/message`, `tool/result`, and `steering/message` payloads before current-shape validation. Each imported message receives the deterministic id `legacy-message:<session-id>:<event-seq>`; a tool-result content replacement inherits its target's imported id. The coordinator uses the same normalized view for `load`, `inspect`, ownerless-state claims, and HMR prefix adoption, so resumed sessions can append current events without a false prefix collision. Storage remains append-only: the read does not rewrite old records, and every later append uses the current shape. This is the narrow import exception from the [pre-identity message recovery decision](../../../.agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.md), not a general v0 migration promise.
When a live session emits `session/disposed`, the coordinator waits for its controller, serializes a final drain, then releases state owned by that exact `Session` object. Failed retirement leaves the controller in the live-session map, so backend teardown can retry it. Backend teardown stops event admission first, flushes every remaining controller, awaits per-id operations, and only then closes the storage handle.
The side-effect-free `locate` and lightweight `listSnapshots` queries remain backend-owned because they describe storage topology and revision identity rather than write orchestration. `listSnapshots(signal?)` passes the caller's exact signal into backend discovery so observers can cancel that work without detaching it.
@@ -43,6 +46,7 @@ The `PersistenceBackend<TornMarker>` hooks (the only seam between the coordinato
|---|---|
| `name` | Backend label for the dispose-failure `AggregateError`. |
| `loadStored(id, signal?)` | Read a stored prefix by id across every storage scope. Used by resume/load, non-mutating inspect, live adoption, and the create-collision probe. The optional signal belongs to observation-only reads. Returned metadata identifies `id`; an opaque `tornMarker` is present iff a torn tail must be truncated. |
| `loadStoredFrom?(id, fromSeq, signal?)` | Optional seek-capable suffix read behind the service's `readFrom`: the header plus stored events with `seq >= fromSeq`, non-mutating, no torn marker. SQLite implements it (`WHERE seq >= ?`); a backend that omits it gets the coordinator's fallback — `loadStored` plus a forward skip. |
| `appendBatch(meta, events, isMaterialized)` | Durably append a contiguous batch, lazily materializing ATOMICALLY when not yet materialized. |
| `commitRepair(meta, tornMarker, closers)` | Make a crash repair durable: truncate the torn tail (iff `tornMarker !== undefined` — a marker may be falsy, e.g. seq/offset `0`) and append `closers`. NOT required to be atomic. Used by load (truncate + closers) and live-adoption (truncate only). |
| `list(signal?)` | List all stored metadata, observing optional cancellation. |

8
packages/session-persistence/session-persistence/README.zh.md
View File

@@ -13,8 +13,9 @@
| `locate(meta): SessionLocation \| undefined` | 在不执行 I/O 或实体化的情况下解析绝对的每会话产物目标。没有独立本地产物的后端返回 `undefined`。 |
| `create(meta): Promise<void>` | 注册新会话元数据。可以将物理写入延迟到第一次 `append`（延迟实体化）。 |
| `append(id, events): Promise<void>` | 持久保存一个批次。仅追加；任何修复后，第一个事件 `seq` == 已存储 next-seq；非 JSON 可序列化数据会被拒绝，并命名违规类型。 |
| `load(id): Promise<{ meta; events }>` | 返回已存储 header 和平衡、连续的日志，其中事件已脱离并验证，带标识的消息已深度冻结。实时 load 先 flush 其快照，并在轮次开放时拒绝；冷 load 保留中断的最终轮次，并用合成 `tool/result`/`step/end?`/`turn/end {interrupted}` 事件关闭它。只丢弃撕裂尾部碎片；已提交损坏、格式错误的消息和未知 `version` 会被拒绝。 |
| `inspect(id, signal?): Promise<{ meta; events }>` | 返回脱离的有效已存储前缀，其中带标识的消息已经验证并深度冻结；不截断撕裂尾部、合成恢复 closer 或发布协调器状态。它与同 id 写入串行化；可选信号会迅速拒绝已排队调用方，阻止该后端读取启动，并取消活动后端读取工作。用于绝不应恢复日志的读模型和其他观察者。 |
| `load(id): Promise<{ meta; events }>` | 返回已存储 header 和平衡、连续的日志，其中事件已脱离并验证，带标识的消息已深度冻结。协调器会在返回快照中，将消息标识机制引入前的四种消息事件形状升级为当前包装层；其余过时或格式错误的形状仍会被拒绝。实时 load 先 flush 其快照，并在轮次开放时拒绝；冷 load 保留中断的最终轮次，并用合成 `tool/result`/`step/end?`/`turn/end {interrupted}` 事件关闭它。只丢弃撕裂尾部碎片；已提交损坏和未知 `version` 会被拒绝。 |
| `inspect(id, signal?): Promise<{ meta; events }>` | 返回脱离的有效已存储前缀，其中带标识的消息已经升级、验证并深度冻结；不截断撕裂尾部、合成恢复 closer 或发布协调器状态。它与同 id 写入串行化；可选信号会迅速拒绝已排队调用方，阻止该后端读取启动，并取消活动后端读取工作。用于绝不应恢复日志的读模型和其他观察者。 |
| `readFrom(id, fromSeq, signal?): Promise<{ meta; events }>` | read-from-seq 原语：返回 header 和 `seq >= fromSeq` 的有效已存储事件，与 `inspect` 同样脱离且非变更（不截断、不合成 closer、不发布协调器状态）。`fromSeq` 达到或超过已存储末尾时返回空事件列表；负数或非安全整数 `fromSeq` 会被拒绝。可寻址后端（SQLite）只读后缀；顺序后端（JSONL）仍解析整个产物并向前跳过——原语约束的是返回和重折叠的量，不是每个后端的物理读取。用于从水位续折尾部的 checkpoint 消费者（例如持久投影缓存）。 |
| `list(signal?): Promise<SessionHeader[]>` | 从元数据轻量列出，不解析完整日志。可选信号取消后端列表工作。零事件延迟实体化会话不在 `list` 中。 |
| `listSnapshots(signal?): Promise<SessionPersistenceSnapshot[]>` | 返回轻量元数据和不透明品牌化每日志修订，不加载事件日志。日志及其后端存储不变时，修订保持相等；append 或变更性 load 修复后会改变；不会仅因两个存储使用相同本地计数器而冲突。可选信号请求取消后端发现工作；第一方后端在拒绝前结算已启动列表工作，使已等待调用完全停稳。 |
@@ -33,6 +34,8 @@
崩溃修复只适用于冷状态。对于实时 id，`load(id)` 为权威内存日志制作快照，等待该快照持久，并只在平衡时将其与协调器已存储 header 一起返回；开放实时轮次会被拒绝，而不会收到合成中断 closer。冷 load 在后端读取和修复写入期间保留 id，因此同 id 实时 `Session` 的并发发布会拒绝并回滚。HMR 接管通过 `loadStored` 读取，应用协调器 cwd 检查，并绝不关闭活动轮次。
后端读取会在当前形状验证前，规范化消息标识机制引入前的 `user/message`、`assistant/message`、`tool/result` 以及 steering（中途引导）对应的 `steering/message` 载荷。每条导入消息都会获得确定性的 id `legacy-message:<session-id>:<event-seq>`；工具结果的内容替换会继承其目标导入后的 id。协调器对 `load`、`inspect`、无 owner 状态的认领和 HMR 前缀接管使用同一份规范化视图，因此恢复后的会话可以追加当前事件，不会被误判为发生前缀冲突。存储仍然仅追加：读取不会重写旧记录，此后追加的每个事件都使用当前形状。这是[消息标识机制引入前的消息恢复决策](../../../.agents/notes/implemented/bug-fix/2026-07-28-load-pre-identity-session-messages.md)所规定的范围受限的导入例外，并不构成通用的 v0 迁移承诺。
实时会话发出 `session/disposed` 时，协调器等待其 controller，串行化最终 drain，然后释放该精确 `Session` 对象拥有的状态。失败退役会将 controller 保留在实时会话 map 中，使后端拆卸可重试。后端拆卸先停止事件接纳，flush 每个剩余 controller，等待每 id 操作，最后才关闭存储句柄。
无副作用 `locate` 和轻量 `listSnapshots` 查询仍由后端负责，因为它们描述存储拓扑和修订身份，而非写入编排。`listSnapshots(signal?)` 将调用方的精确信号传入后端发现，使观察者可在不脱离该工作的情况下取消。
@@ -43,6 +46,7 @@
|---|---|
| `name` | dispose 失败 `AggregateError` 的后端标签。 |
| `loadStored(id, signal?)` | 在全部存储范围中按 id 读取已存储前缀。用于 resume/load、非变更 inspect、实时接管和 create 冲突探测。可选信号属于仅观察读取。返回元数据标识 `id`；当且仅当必须截断撕裂尾部时才存在不透明 `tornMarker`。 |
| `loadStoredFrom?(id, fromSeq, signal?)` | 服务 `readFrom` 背后的可选可寻址后缀读取：返回 header 和 `seq >= fromSeq` 的已存储事件，非变更、无撕裂标记。SQLite 实现它（`WHERE seq >= ?`）；不实现的后端使用协调器回退——`loadStored` 加向前跳过。 |
| `appendBatch(meta, events, isMaterialized)` | 持久追加连续批次；尚未实体化时以原子方式延迟实体化。 |
| `commitRepair(meta, tornMarker, closers)` | 使崩溃修复持久：截断撕裂尾部（当且仅当 `tornMarker !== undefined`；标记可为 falsy，例如 seq/offset `0`），并追加 `closers`。不要求原子性。由 load（截断 + closer）和实时接管（仅截断）使用。 |
| `list(signal?)` | 列出全部已存储元数据，观察可选取消。 |

219
packages/session-persistence/session-persistence/src/coordinator.ts
View File

@@ -25,6 +25,17 @@ export interface StoredPrefix<TornMarker = unknown> {
tornMarker?: TornMarker
}
/**
* A stored session's header plus the events at or past a requested seq — the
* return shape of the optional seek-capable
* {@link PersistenceBackend.loadStoredFrom} hook. Non-mutating reads carry no
* torn marker: there is nothing to repair.
*/
export interface StoredSuffix {
meta: SessionHeader
events: SessionEvent[]
}
/**
* The storage seam between {@link PersistenceCoordinator} and a concrete
* backend: the minimal set of durable primitives the orchestration calls. A
@@ -50,6 +61,22 @@ export interface PersistenceBackend<TornMarker = unknown> {
*/
loadStored(id: SessionId, signal?: AbortSignal): Promise<StoredPrefix<TornMarker> | undefined>
/**
* Optional seek-capable suffix read behind the service's `readFrom`: return
* the header plus the stored events with `seq >= fromSeq` without reading
* the whole log. A backend whose medium can address events by seq (SQLite)
* implements this so `readFrom` scales with the suffix; sequential backends
* omit it and the coordinator falls back to {@link loadStored} plus a
* forward skip. Non-mutating (no truncation, no closers). Validation of the
* region strictly below `fromSeq` is limited to seq contiguity — the
* service contract scopes this read to the suffix.
* @param id - persisted session id to resolve.
* @param fromSeq - first event seq to include (non-negative safe integer,
* validated by the coordinator before this hook runs).
* @param signal - optional cancellation for backend read work.
*/
loadStoredFrom?(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<StoredSuffix | undefined>
/**
* Durably append a CONTIGUOUS batch, lazily materializing the session first
* when `!isMaterialized`. The materialize-write and the first event batch MUST
@@ -146,10 +173,142 @@ function assertSupportedEvents(events: readonly SessionEvent[], id: SessionId):
}
}
/** Materialize stored events as validated snapshots with immutable messages. */
/** Return an object record without widening arrays into message payloads. */
function asRecord(value: unknown): Record<string, unknown> | undefined {
return typeof value === 'object' && value !== null && !Array.isArray(value)
? value as Record<string, unknown>
: undefined
}
type PersistedMessageId = SessionEvent<'user/message'>['data']['id']
/** Mint the stable import identity for a message persisted before identities existed. */
function legacyMessageId(id: SessionId, seq: number): PersistedMessageId {
return `legacy-message:${id}:${seq}` as PersistedMessageId
}
/** Read a replacement target while leaving malformed surface metadata to the session validator. */
function replacementStart(event: SessionEvent): number | undefined {
const op = asRecord((event as SessionEvent & { surfaceOp?: unknown }).surfaceOp)
return op?.['op'] === 'replace' && typeof op['start'] === 'number'
? op['start']
: undefined
}
/**
* Upgrade one pre-identity message event into the current wrapper shape.
* Current-looking malformed events remain untouched so validation rejects them
* instead of disguising corruption as legacy data.
*/
function migrateLegacyMessageEvent(
event: SessionEvent,
id: SessionId,
messageIds: ReadonlyMap<number, PersistedMessageId>,
): SessionEvent {
const data = asRecord(event.data)
if (data === undefined) return event
switch (event.type) {
case 'user/message': {
if (Object.hasOwn(data, 'id') || Object.hasOwn(data, 'role')
|| Object.hasOwn(data, 'message')
|| !Object.hasOwn(data, 'content') || !Object.hasOwn(data, 'source')) return event
return {
...event,
data: {
...data,
id: legacyMessageId(id, event.seq),
role: 'user',
},
} as SessionEvent
}
case 'assistant/message': {
if (Object.hasOwn(data, 'message')
|| !Object.hasOwn(data, 'content') || !Object.hasOwn(data, 'provenance')) return event
const { content, provenance, ...eventData } = data
return {
...event,
data: {
...eventData,
message: {
id: legacyMessageId(id, event.seq),
role: 'assistant',
content,
source: {
...asRecord(provenance),
kind: 'model',
},
},
},
} as SessionEvent
}
case 'tool/result': {
if (Object.hasOwn(data, 'message')
|| !Object.hasOwn(data, 'callId') || !Object.hasOwn(data, 'content')
|| !Object.hasOwn(data, 'isError')) return event
const { callId, content, isError, ...eventData } = data
const inheritedId = replacementStart(event)
return {
...event,
data: {
...eventData,
message: {
id: inheritedId === undefined
? legacyMessageId(id, event.seq)
: messageIds.get(inheritedId),
role: 'user',
content: [{
type: 'tool-result',
toolCallId: callId,
content,
isError,
}],
source: {
kind: 'tool',
callId,
},
},
},
} as SessionEvent
}
case 'steering/message': {
if (Object.hasOwn(data, 'message')
|| !Object.hasOwn(data, 'content') || !Object.hasOwn(data, 'source')) return event
const { content, source, ...eventData } = data
return {
...event,
data: {
...eventData,
message: {
id: legacyMessageId(id, event.seq),
role: 'user',
content,
source,
},
},
} as SessionEvent
}
default:
return event
}
}
/** Read the identified message carried by one validated current event. */
function eventMessageId(event: SessionEvent): PersistedMessageId | undefined {
const data = asRecord(event.data)
const message = event.type === 'user/message' ? data : asRecord(data?.['message'])
return typeof message?.['id'] === 'string' ? message['id'] as PersistedMessageId : undefined
}
/** Materialize stored events as upgraded, validated snapshots with immutable messages. */
function snapshotStoredEvents(events: readonly SessionEvent[], id: SessionId): SessionEvent[] {
assertSupportedEvents(events, id)
return events.map(snapshotSessionEvent)
const messageIds = new Map<number, PersistedMessageId>()
return events.map((event) => {
const snapshot = snapshotSessionEvent(migrateLegacyMessageEvent(event, id, messageIds))
const messageId = eventMessageId(snapshot)
if (messageId !== undefined) messageIds.set(snapshot.seq, messageId)
return snapshot
})
}
/**
@@ -325,6 +484,52 @@ export class PersistenceCoordinator<TornMarker = unknown> {
}
}
/**
* Read the stored events from `fromSeq` onward, detached and non-mutating
* (the read-from-seq primitive behind the service's `readFrom`). Runs on
* the same per-id chain as writes; a backend with the seek-capable
* {@link PersistenceBackend.loadStoredFrom} hook reads only the suffix,
* every other backend reads its stored prefix and skips forward here.
* @param id - persisted session to read.
* @param fromSeq - first event seq to include; a non-negative safe integer.
* @param signal - optional cancellation for queued and backend read work.
* @returns stored header and the valid stored events with `seq >= fromSeq`.
*/
readFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
if (!Number.isSafeInteger(fromSeq) || fromSeq < 0) {
return Promise.reject(new TypeError(`readFrom fromSeq must be a non-negative safe integer, got ${String(fromSeq)}`))
}
const retired = Promise.resolve(this.retirements.get(id))
const waited = signal === undefined ? retired : observeQueuedAbort(retired, signal, () => false)
return waited.then(() => this.serialize(id, () => this.readFromCore(id, fromSeq, signal), signal))
}
private async readFromCore(
id: SessionId,
fromSeq: number,
signal?: AbortSignal,
): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
signal?.throwIfAborted()
if (this.backend.loadStoredFrom !== undefined) {
let suffix: StoredSuffix | undefined
try {
suffix = await this.backend.loadStoredFrom(id, fromSeq, signal)
} catch (error: unknown) {
if (signal?.aborted) signal.throwIfAborted()
throw error
}
signal?.throwIfAborted()
if (suffix === undefined) throw new Error(`session "${id}" not found`)
this.assertStoredId(id, suffix.meta)
this.assertVersion(suffix.meta)
assertSupportedEvents(suffix.events, id)
return { meta: structuredClone(suffix.meta), events: structuredClone(suffix.events) }
}
const whole = await this.inspectCore(id, signal)
// Sequential fallback: contiguous seqs from 0 make the suffix an index slice.
return { meta: whole.meta, events: whole.events.slice(fromSeq) }
}
private async loadCore(id: SessionId): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
const stored = await this.backend.loadStored(id)
if (stored === undefined) throw new Error(`session "${id}" not found`)
@@ -526,7 +731,7 @@ export class PersistenceCoordinator<TornMarker = unknown> {
/* v8 ignore next -- a cursor > 0 means the session was materialized, so it exists */
if (stored === undefined) return false
this.assertStoredId(id, stored.meta)
return seedCoversPrefix(seed, stored.events.slice(0, cursor))
return seedCoversPrefix(seed, snapshotStoredEvents(stored.events, id).slice(0, cursor))
}
/**
@@ -614,19 +819,19 @@ export class PersistenceCoordinator<TornMarker = unknown> {
throw new Error(`session "${session.header.id}" is already persisted at a different cwd (persisted: ${String(meta.cwd)}, live: ${String(session.header.cwd)}) (id collision)`)
}
this.assertVersion(meta)
assertSupportedEvents(events, session.header.id)
if (!seedCoversPrefix(seed, events)) {
const storedEvents = snapshotStoredEvents(events, session.header.id)
if (!seedCoversPrefix(seed, storedEvents)) {
throw new Error(`session "${session.header.id}" already has a persisted log on disk that does not match this live session (id collision)`)
}
// Truncate-only repair (no closers): the open turn is NOT closed here.
if (tornMarker !== undefined) await this.backend.commitRepair(meta, tornMarker, [])
this.states.set(session.header.id, {
meta: { ...meta },
cursor: events.length,
cursor: storedEvents.length,
materialized: true,
owner: session,
})
const suffix = seed.slice(events.length)
const suffix = seed.slice(storedEvents.length)
if (suffix.length > 0) await this.appendCore(session.header.id, suffix)
}

31
packages/session-persistence/session-persistence/src/index.ts
View File

@@ -23,7 +23,7 @@ export interface SessionPersistenceSnapshot {
// The backend-agnostic write-path orchestration first-party backends compose.
export { PersistenceCoordinator } from './coordinator.ts'
export type { PersistenceBackend, StoredPrefix } from './coordinator.ts'
export type { PersistenceBackend, StoredPrefix, StoredSuffix } from './coordinator.ts'
declare module 'cordis' {
interface Context {
@@ -93,7 +93,9 @@ export abstract class SessionPersistence extends Service {
* A coordinator-backed cold load reserves the identity across storage awaits,
* so concurrent publication of a same-id live Session rejects.
* Returned events are detached, and every identified message is deeply
* frozen; malformed identified messages reject before any stored event is returned.
* frozen. Coordinator-backed implementations upgrade supported pre-identity
* message events before validation; other malformed messages reject before
* any stored event is returned.
* @param id - the persisted session to reload.
* @returns the header and a log ending on a balanced `turn/end`.
*/
@@ -103,14 +105,35 @@ export abstract class SessionPersistence extends Service {
* Inspect a header and its valid contiguous stored prefix without repairing
* a torn tail, closing an interrupted turn, or publishing coordinator state.
* This read is serialized with writes for the same id and returns detached
* values with deeply frozen identified messages, so observers cannot mutate message
* identity/content or backend-owned state. Malformed identified messages reject.
* values with upgraded, deeply frozen identified messages, so observers
* cannot mutate message identity/content or backend-owned state. Other
* malformed messages reject.
* @param id - the persisted session to inspect.
* @param signal - optional cancellation for queued and backend read work.
* @returns the header and valid stored event prefix exactly as observed.
*/
abstract inspect(id: SessionId, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }>
/**
* Read the stored events from `fromSeq` onward — the read-from-seq
* primitive for read models that resume from a watermark (e.g. a persisted
* projection cache folding only the tail past its checkpoint). Like
* {@link inspect} it is non-mutating and detached: no torn-tail truncation,
* no synthetic closers, no coordinator-state publication; only events from
* the valid contiguous stored prefix are returned, so a torn fragment never
* reaches the caller. `fromSeq` at or beyond the stored prefix returns an
* empty event list (never an error). Backends whose medium can seek by seq
* (SQLite) read only the suffix; sequential media (JSONL, both encodings)
* still parse the whole artifact and skip forward — the primitive bounds
* what is RETURNED and refolded, not every backend's physical read.
* @param id - the persisted session to read.
* @param fromSeq - first event seq to include; a non-negative safe integer.
* @param signal - optional cancellation for queued and backend read work.
* @returns the header and the stored events with `seq >= fromSeq`.
*/
abstract readFrom(id: SessionId, fromSeq: number, signal?: AbortSignal):
Promise<{ meta: SessionHeader; events: SessionEvent[] }>
/**
* Lightweight listing from metadata, without a full-log parse.
* @param signal - optional cancellation for backend listing work.

37
packages/session-persistence/session-persistence/tests/contract.ts
View File

@@ -289,6 +289,43 @@ export function runPersistenceContract(name: string, make: () => Promise<Contrac
await expect(persistence.listSnapshots(controller.signal)).rejects.toBe(reason)
await expect(persistence.inspect(SessionId('cancelled-inspect'), controller.signal))
.rejects.toBe(reason)
await expect(persistence.readFrom(SessionId('cancelled-read-from'), 0, controller.signal))
.rejects.toBe(reason)
} finally {
await dispose()
}
})
it('readFrom returns exactly the stored suffix from the requested seq, without mutating the log', async () => {
const { persistence, dispose } = await make()
try {
const m = meta('read-from', '/work')
const log = oneTurnLog()
await persistence.create(m)
await persistence.append(m.id, log)
const whole = await persistence.readFrom(m.id, 0)
expect(whole.meta).toMatchObject({ id: m.id, cwd: '/work' })
expect(whole.events).toEqual(log)
const suffix = await persistence.readFrom(m.id, 3)
expect(suffix.events).toEqual(log.slice(3))
expect(suffix.events[0]?.seq).toBe(3)
// At/past the stored end: an empty tail, never an error.
await expect(persistence.readFrom(m.id, log.length)).resolves.toMatchObject({ events: [] })
await expect(persistence.readFrom(m.id, log.length + 100)).resolves.toMatchObject({ events: [] })
// Non-mutating: an interrupted-turn log is served as stored, no closers.
await persistence.append(m.id, [
{ type: 'turn/start', seq: 6, time: 7, data: { turn: 2, trigger: { kind: 'message', source: { kind: 'user' } } } },
])
const tail = await persistence.readFrom(m.id, 6)
expect(tail.events.map(event => event.type)).toEqual(['turn/start'])
await expect(persistence.readFrom(SessionId('absent-read-from'), 0)).rejects.toThrow('not found')
await expect(persistence.readFrom(m.id, -1)).rejects.toThrow('non-negative safe integer')
await expect(persistence.readFrom(m.id, 1.5)).rejects.toThrow('non-negative safe integer')
} finally {
await dispose()
}

145
packages/session-persistence/session-persistence/tests/coordinator-contract.ts
View File

@@ -13,8 +13,8 @@ import { createUserMessage } from '@deepseek-ai/dsh-llm'
import { describe, expect, it, vi } from 'vitest'
import { Context, type Fiber } from 'cordis'
import { scopeTarget } from '@deepseek-ai/dsh-scope'
import SessionStore, { SESSION_FORMAT_VERSION, SessionId } from '@deepseek-ai/dsh-session'
import type { Session, SessionEvent } from '@deepseek-ai/dsh-session'
import SessionStore, { SESSION_FORMAT_VERSION, Session, SessionId } from '@deepseek-ai/dsh-session'
import type { SessionEvent } from '@deepseek-ai/dsh-session'
import { meta, oneTurnLog, appendLog } from './contract.ts'
/**
@@ -45,6 +45,80 @@ function send(session: Session, events: readonly SessionEvent[]): void {
appendLog(session, events)
}
/** A valid persisted log from immediately before messages gained wrappers and identities. */
function legacyMessageLog(): SessionEvent[] {
return [
{ type: 'turn/start', seq: 0, time: 1, data: { turn: 1, trigger: { kind: 'message', source: { kind: 'user' } } } },
{
type: 'user/message',
seq: 1,
time: 2,
data: { content: [{ type: 'text', text: 'hi' }], source: { kind: 'user' } },
surfaceOp: 'append',
},
{ type: 'step/start', seq: 2, time: 3, data: { turn: 1, step: 1 } },
{
type: 'assistant/message',
seq: 3,
time: 4,
data: {
turn: 1,
step: 1,
content: [{ type: 'tool-call', id: 'call-1', name: 'read', arguments: '{}' }],
provenance: { provider: 'mock', model: 'mock' },
},
surfaceOp: 'append',
},
{
type: 'tool/call',
seq: 4,
time: 5,
data: { turn: 1, step: 1, callId: 'call-1', name: 'read', arguments: '{}' },
},
{
type: 'tool/result',
seq: 5,
time: 6,
data: {
turn: 1,
step: 1,
callId: 'call-1',
content: [{ type: 'text', text: 'full result' }],
isError: false,
},
sourceEventSeqs: [4],
surfaceOp: 'append',
},
{
type: 'steering/message',
seq: 6,
time: 7,
data: {
turn: 1,
content: [{ type: 'text', text: 'continue' }],
source: { kind: 'plugin', plugin: 'test' },
},
surfaceOp: 'append',
},
{
type: 'tool/result',
seq: 7,
time: 8,
data: {
turn: 1,
step: 1,
callId: 'call-1',
content: [{ type: 'text', text: 'pruned' }],
isError: false,
},
sourceEventSeqs: [5],
surfaceOp: { op: 'replace', start: 5, end: 5 },
},
{ type: 'step/end', seq: 8, time: 9, data: { turn: 1, step: 1 } },
{ type: 'turn/end', seq: 9, time: 10, data: { turn: 1, reason: { kind: 'completed' } } },
] as unknown as SessionEvent[]
}
/** A live session created inside its OWN fiber, so it survives a backend reload. */
async function liveSessionInFiber(
ctx: Context, id: string, cwd: string | undefined,
@@ -269,6 +343,48 @@ export function runCoordinatorContract(name: string, makeFixture: () => Promise<
}
})
it('loads pre-identity message logs into resumable current sessions', async () => {
const fix = await makeFixture()
const { ctx, fiber } = await freshCtx(fix)
try {
const id = SessionId('legacy-message-load')
await ctx.sessionPersistence.create(meta(id, WORK))
await ctx.sessionPersistence.append(id, legacyMessageLog())
for (const snapshot of [
await ctx.sessionPersistence.inspect(id),
await ctx.sessionPersistence.load(id),
]) {
const messages = snapshot.events.flatMap((event) => {
if (event.type === 'user/message') return [event.data]
if (event.type === 'assistant/message'
|| event.type === 'tool/result'
|| event.type === 'steering/message') return [event.data.message]
return []
})
expect(messages.map(message => message.id)).toEqual([
`legacy-message:${id}:1`,
`legacy-message:${id}:3`,
`legacy-message:${id}:5`,
`legacy-message:${id}:6`,
`legacy-message:${id}:5`,
])
expect(messages.every(message => Object.isFrozen(message))).toBe(true)
const resumed = new Session(id, snapshot.events, snapshot.meta)
expect(resumed.deriveMessages().map(message => message.id)).toEqual([
`legacy-message:${id}:1`,
`legacy-message:${id}:3`,
`legacy-message:${id}:5`,
`legacy-message:${id}:6`,
])
}
} finally {
await fiber.dispose()
await fix.cleanup()
}
})
it('rejects malformed persisted message events before returning them', async () => {
const fix = await makeFixture()
const { ctx, fiber } = await freshCtx(fix)
@@ -292,6 +408,31 @@ export function runCoordinatorContract(name: string, makeFixture: () => Promise<
.rejects.toThrow('message must have role "user"')
await expect(ctx.sessionPersistence.load(id))
.rejects.toThrow('message must have role "user"')
for (const type of ['tool/result', 'steering/message'] as const) {
const malformedId = SessionId(`invalid-${type}`)
await ctx.sessionPersistence.create(meta(malformedId, WORK))
await ctx.sessionPersistence.append(malformedId, [{
type,
seq: 0,
time: 1,
surfaceOp: 'append',
data: { message: null },
} as unknown as SessionEvent])
await expect(ctx.sessionPersistence.inspect(malformedId))
.rejects.toThrow('lacks an identified message')
}
const pluginId = SessionId('non-object-plugin-event')
await ctx.sessionPersistence.create(meta(pluginId, WORK))
await ctx.sessionPersistence.append(pluginId, [{
type: 'plugin/test',
seq: 0,
time: 1,
data: null,
} as unknown as SessionEvent])
await expect(ctx.sessionPersistence.inspect(pluginId))
.resolves.toMatchObject({ events: [{ type: 'plugin/test', data: null }] })
} finally {
await fiber.dispose()
await fix.cleanup()

123
packages/session-persistence/session-persistence/tests/persistence.spec.ts
View File

@@ -99,6 +99,10 @@ class MemoryPersistence extends SessionPersistence implements PersistenceBackend
return this.coordinator.inspect(id, signal)
}
readFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
return this.coordinator.readFrom(id, fromSeq, signal)
}
// --- PersistenceBackend hooks (the Map storage primitives) ---
// A Map-backed store has no torn tails, so `tornMarker` is never set.
@@ -157,6 +161,13 @@ class ControlledBackend implements PersistenceBackend<never> {
repairAttempts = 0
beforeAppend?: (attempt: number) => Promise<void>
beforeLoadStored?: (attempt: number, signal?: AbortSignal) => Promise<void>
/** When set, the declared seek hook delegates here so readFrom exercises it; unset throws (tests set it first). */
seekHook?: (id: SessionId, fromSeq: number, signal?: AbortSignal) => Promise<StoredPrefix<never> | undefined>
loadStoredFrom(id: SessionId, fromSeq: number, signal?: AbortSignal): Promise<StoredPrefix<never> | undefined> {
if (this.seekHook === undefined) throw new Error('seekHook not configured for this test')
return this.seekHook(id, fromSeq, signal)
}
async loadStored(id: SessionId, signal?: AbortSignal): Promise<StoredPrefix<never> | undefined> {
await this.beforeLoadStored?.(++this.loadAttempts, signal)
@@ -453,6 +464,58 @@ describe('PersistenceCoordinator observation cancellation', () => {
}
})
it('readFrom via the seek hook: serves the suffix, maps undefined to not-found, and relays hook failures by abort state', async () => {
const ctx = new Context()
await ctx.plugin(SessionStore)
const backend = new ControlledBackend()
const id = SessionId('seek-read-from')
const log = oneTurnLog()
backend.store.set(id, { meta: meta(id), events: log })
let coordinator!: PersistenceCoordinator<never>
const fiber = await ctx.plugin(Object.assign((inner: Context) => {
coordinator = new PersistenceCoordinator(inner, backend)
}, { inject: ['sessions'] }))
try {
// Happy path through the hook: only the suffix comes back, detached.
backend.seekHook = async (hookId, fromSeq) => {
const entry = backend.store.get(hookId)
if (entry === undefined) return undefined
return { meta: structuredClone(entry.meta), events: entry.events.filter(e => e.seq >= fromSeq) }
}
const suffix = await coordinator.readFrom(id, 3)
expect(suffix.events).toEqual(log.slice(3))
// The hook's undefined is the seam's not-found.
await expect(coordinator.readFrom(SessionId('missing-seek'), 0)).rejects.toThrow('not found')
// A hook failure with no cancellation in play propagates as-is.
const hookFailure = new Error('seek backend exploded')
backend.seekHook = () => Promise.reject(hookFailure)
await expect(coordinator.readFrom(id, 0)).rejects.toBe(hookFailure)
// A hook failure after cancellation surfaces the caller's abort reason,
// not the backend's internal teardown error. The abort fires only once
// the hook is provably entered, so the failure exercises the catch (not
// the pre-invocation throwIfAborted).
const controller = new AbortController()
const reason = new Error('read-from cancelled mid-hook')
let hookEntered = false
backend.seekHook = async (_hookId, _fromSeq, signal) => {
hookEntered = true
await new Promise<void>((resolve) => { signal?.addEventListener('abort', () => { resolve() }, { once: true }) })
throw new Error('backend teardown after abort')
}
const pending = coordinator.readFrom(id, 0, controller.signal)
const observed = pending.catch((error: unknown) => error)
await vi.waitFor(() => { expect(hookEntered).toBe(true) })
controller.abort(reason)
expect(await observed).toBe(reason)
} finally {
await fiber.dispose()
await ctx.fiber.dispose()
}
})
it('rejects a cancelled inspect while an in-flight retirement drain is still pending', async () => {
const ctx = new Context()
await ctx.plugin(SessionStore)
@@ -537,6 +600,66 @@ describe('PersistenceCoordinator retirement', () => {
}
})
it('a superseded retirement leaves the successor lifecycle\'s pending drain in place', async () => {
const ctx = new Context()
await ctx.plugin(SessionStore)
const backend = new ControlledBackend()
let coordinator!: PersistenceCoordinator<never>
const backendFiber = await ctx.plugin(Object.assign((inner: Context) => {
coordinator = new PersistenceCoordinator(inner, backend)
}, { inject: ['sessions'] }))
const internals = coordinator as unknown as CoordinatorInternals
const readGate = Promise.withResolvers<boolean>()
try {
const id = SessionId('superseded-retirement')
// First lifecycle: unmaterialized (zero events), so a same-id successor
// may legally reclaim the abandoned id later.
let first!: Session
const firstFiber = await ctx.plugin(Object.assign((inner: Context) => {
first = inner.sessions.create(id)
}, { inject: ['sessions'] }))
await ctx.sessions.flush(first)
// Occupy the per-id serialize chain with a gated read: everything the
// two retirements queue stays pending behind it. (Attempt counting
// starts here — an absent beforeLoadStored short-circuits the optional
// call without evaluating its ++ argument.)
backend.beforeLoadStored = async (attempt) => {
if (attempt === 1) await readGate.promise
}
const parked = coordinator.inspect(id).catch((error: unknown) => error)
await vi.waitFor(() => { expect(backend.loadAttempts).toBe(1) })
// First retirement queues behind the gate and stays pending.
await firstFiber.dispose()
await vi.waitFor(() => { expect(internals.retirements.has(id)).toBe(true) })
const firstRetirement = internals.retirements.get(id)
// Successor lifecycle retires while the first drain is still in flight:
// retire() replaces the map entry synchronously.
const secondFiber = await ctx.plugin(Object.assign((inner: Context) => {
inner.sessions.create(id)
}, { inject: ['sessions'] }))
await secondFiber.dispose()
await vi.waitFor(() => {
expect(internals.retirements.get(id)).not.toBe(firstRetirement)
})
// Release the chain: the first drain settles and its forget() must not
// delete the successor's entry (exact-entry guard); the successor's own
// forget() then clears the map.
readGate.resolve(true)
expect(await parked).toBeInstanceOf(Error) // the parked inspect (not found) is observed
await firstRetirement
await vi.waitFor(() => { expect(internals.retirements.has(id)).toBe(false) })
} finally {
readGate.resolve(true)
await backendFiber.dispose()
await ctx.fiber.dispose()
}
})
it('a replacement queued before retirement cleanup still collides with the live owner', async () => {
const ctx = new Context()
await ctx.plugin(SessionStore)

4
packages/session-projection/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/session-projection/README.md
README.md: 81c67d56e136ba4853e86d889b485d4df80ac1fe
README.zh.md: 72e23b78a48a989f355f9be3d34d81a440ca1d04
README.md: ae80a905705d205adb4a1ee66c72fa28d0d8b6d6
README.zh.md: 97e25dd16caeeb444f5f5309eed3341d422fa1b1

1
packages/session-projection/README.md
View File

@@ -7,3 +7,4 @@ Session-projection capability family: the seam through which domain host plugins
| Package | ctx key | Role |
|---|---|---|
| [`session-projection`](session-projection/README.md) | `sessionProjections` | The interface package: the merge-extensible `SessionProjectionMap` type table, the `ProjectionDefinition` unit contract, and the eagerly driven registry carriers read synchronously |
| [`session-projection-cache`](session-projection-cache/README.md) | `sessionProjectionCache` | Persisted projection cache: durable per-session unit checkpoints over the domain data form, throttled write-behind with mandatory turn/end + detach points, and the cold-read ladder (cache row + persistence tail replay) |

1
packages/session-projection/README.zh.md
View File

@@ -7,3 +7,4 @@
| 包 | ctx 键 | 职责 |
|---|---|---|
| [`session-projection`](session-projection/README.md) | `sessionProjections` | 接口包（package）：merge-extensible 的 `SessionProjectionMap` 类型表、`ProjectionDefinition` 单元契约，以及供载体同步读取的正向驱动注册表 |
| [`session-projection-cache`](session-projection-cache/README.md) | `sessionProjectionCache` | 持久投影缓存：基于域数据形态的按会话单元 checkpoint 持久化、带 turn/end + detach 两个必写点的节流后写，以及冷读阶梯（缓存行 + 持久化尾部重放） |

6
packages/session-projection/session-projection-cache/README.i18n.yaml Normal file
View File

@@ -0,0 +1,6 @@
# Bilingual-pair consistency record (docs/i18n/README.md): the git blob hash of each
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/session-projection/session-projection-cache/README.md
README.md: 5d4ad07fab6648acdb40c6aa86d32cc78b4c016e
README.zh.md: ab4076df28cfe2b5a8b41d609967039dcacb7ef4

62
packages/session-projection/session-projection-cache/README.md Normal file
View File

@@ -0,0 +1,62 @@
# @deepseek-ai/dsh-session-projection-cache
English | [中文](README.zh.md)
The persisted projection cache (`ctx.sessionProjectionCache`): durable checkpoints of every registered projection unit's state, one record per session on the domain data form (`session_projcache` domain — the shipped json backend lands it beside `workspace.json` under the configured storage root). Design authority: the [session-projection RFC](../../../.agents/notes/proposed/architecture/2026-07-27-session-projection-and-command-log.md) (persisted projection cache section).
A stored row `(key → {ver, seq, val})` is a fold shortcut, never an authority: possibly stale (`seq` says exactly how stale) but never wrong. Consequences the implementation commits to:
- **Every background write is fail-soft.** A failed durable write logs a warning and keeps the cache stale; the next write or cold read self-heals. A crash between writes costs a longer tail replay, never a wrong value.
- **A `ver` mismatch against the live unit's `stateVersion` discards, never migrates.** A unit bump invalidates its rows at read time; the key refolds from the log.
- **Whole-record writes.** Each write replaces the session's full checkpoint (the registry cut is always complete), snapshotted through the lossless-JSON boundary — a unit state violating the plain-JSON contract fails loud.
- **Records are bound to a log lifecycle, not just an id.** Each record stores the header identity (`createdAt`, `cwd`) it was folded from; every read validates it (the live or stored header is the witness) before accepting a row, so a deleted-then-recreated id or a persistence store swapped under a surviving cache discards the unrelated record instead of seeding phantom values.
- **The log leads, the cache follows.** A live checkpoint flushes the session's buffered events durably BEFORE the cache row lands, so a crash can leave the cache behind the log (a longer tail replay) but never ahead of it.
## Write policy
Two mandatory points, throttled in between:
| Trigger | Nature |
|---|---|
| `turn/end` | Mandatory — the turn-final value is what cold reads want. |
| Session disposal (detach) | Mandatory — the live-to-cold moment; after it the cold ladder serves this session. |
| `writeEveryEvents` committed events | Config throttle (count). |
| `writeIntervalMs` since the first dirty event | Config throttle (interval). |
Both `Config` fields are required (no defaults): flush cadence is a deployment choice with no universally correct value, stated in cordis.yml.
## Listing read (`cachedSnapshot(meta)`)
The zero-I/O rung: whole values viewed straight from the identity-matching stored record (version-matching keys only), returned as a `{asOfSeq, values}` cut — `asOfSeq` is the lowest served-row watermark, so a client seeding its per-session value store under higher-seq-wins can never let a stale list block overwrite a newer push frame. `undefined` when no usable record exists (unknown id, unrelated lifecycle, or no version-matching rows); the api-proxy list carrier turns that into an absent column.
## Cold read (`coldSnapshot(id, signal?)`)
The read ladder, zero full-log load on the happy path: cached rows → `sessionProjections.restoreFloor` (anchored one event below the lowest usable watermark) → persistence `readFrom(id, floor)` → `sessionProjections.restore` → fail-soft write-back of the refreshed rows. The anchor makes a shrunk log (crash-repair truncation) provable: an overreaching row triggers exactly one full re-read from seq 0 instead of serving a ghost value. No registered units serve `{asOfSeq: -1, values: {}}` without touching persistence; a session with no persisted log rejects with the seam's `not found`.
`write(session)` is the synchronous-cut checkpoint both mandatory points use; carriers may call it directly (not fail-soft — the fail-soft wrappers own containment).
## Composition
```yaml
- id: session-projection-cache
name: '@deepseek-ai/dsh-session-projection-cache'
config:
writeEveryEvents: 200
writeIntervalMs: 5000
```
Injects `storageDomain`, `sessionProjections`, `sessionPersistence`, `sessions`. Without this row the projection system runs live-only (watermark cache; cold reads fall back to full log loads wherever a carrier implements them).
## Model Experience
None, as the cache only persists and restores host-side read models of already-logged session state and touches no prompt, message, schema, stream, or tool result.
#### KV Cache effect
None; the cache never assembles or sends provider requests.
## Known Limitations and Deferred Work
- **No eviction or retention surface** — records accumulate per session; pruning stored checkpoints is out-of-band maintenance, same stance as session persistence itself.
- **Interval throttle is per-session coarse** — the timer arms at the first dirty event after a clean write; a steady sub-threshold trickle writes once per interval, not a sliding window.
- **`coldSnapshot` reads are not deduplicated** — two concurrent cold reads of one session each run the ladder; last write-back wins (rows are equivalent), acceptable for listing-scale call rates.

62
packages/session-projection/session-projection-cache/README.zh.md Normal file
View File

@@ -0,0 +1,62 @@
# @deepseek-ai/dsh-session-projection-cache
[English](README.md) | 中文
持久投影缓存（`ctx.sessionProjectionCache`）：把每个已注册投影单元的状态持久化为检查点（checkpoint），基于域数据形态（domain data form）每会话一条记录（`session_projcache` 域——出厂 json 后端将其落在配置的存储根目录下、`workspace.json` 旁边）。设计权威：[session-projection RFC](../../../.agents/notes/proposed/architecture/2026-07-27-session-projection-and-command-log.md)（persisted projection cache 一节）。
一条存储行 `(key → {ver, seq, val})` 是折叠捷径，绝不是权威：可能陈旧（`seq` 精确说明陈旧到哪），但绝不会错。实现据此承诺：
- **每次后台写入都 fail-soft。** 持久写失败只记一条警告并保持缓存陈旧；下一次写入或冷读自愈。两次写之间崩溃的代价是更长的尾部重放，绝不是错误的值。
- **`ver` 与活单元 `stateVersion` 不匹配即丢弃，绝不迁移。** 单元递增版本会在读取时使其行失效；该 key 从日志重新折叠。
- **整记录写入。** 每次写入替换该会话的完整检查点（注册表切面始终是完整的），并经无损 JSON 边界快照——违反纯 JSON 契约的单元状态会大声失败。
- **记录绑定到日志生命周期，而不只是 id。** 每条记录存储其折叠来源的 header 身份（`createdAt`、`cwd`）；每次读取先以活 header 或存储 header 为证验证它,再接受任何行——被删后重建的 id、或缓存幸存而持久化存储被换掉时,无关记录被整体丢弃,绝不播种幻影值。
- **日志领先，缓存跟随。** 活会话检查点先把缓冲事件持久 flush,缓存行才落地,因此崩溃只会让缓存落后于日志（更长的尾部重放）,绝不领先于它。
## 写策略
两个必写点，其间节流：
| 触发 | 性质 |
|---|---|
| `turn/end` | 必写——冷读要的正是轮次终值。 |
| 会话销毁（detach） | 必写——live 转 cold 的时刻；此后冷读阶梯接管该会话。 |
| 累计 `writeEveryEvents` 个已提交事件 | 配置节流（条数）。 |
| 距首个脏事件 `writeIntervalMs` 毫秒 | 配置节流（间隔）。 |
两个 `Config` 字段均必填（无默认值）：写入节奏是部署选择，没有普适正确值，由 cordis.yml 明示。
## 列表读（`cachedSnapshot(meta)`）
零 I/O 一档：从身份匹配的存储记录直接 view 全量值（仅版本匹配的 key）,以 `{asOfSeq, values}` 切面返回——`asOfSeq` 取所服务行的最低水位,客户端在 higher-seq-wins 规则下播种值仓时,陈旧列表块永远压不过更新的推送帧。无可用记录（未知 id、无关生命周期、无版本匹配行）时返回 `undefined`；api-proxy 列表载体将其转为列缺席。
## 冷读（`coldSnapshot(id, signal?)`）
读取阶梯，快乐路径零全量日志加载：缓存行 → `sessionProjections.restoreFloor`（锚在最低可用水位下一格）→ 持久化 `readFrom(id, floor)` → `sessionProjections.restore` → 刷新行的 fail-soft 写回。这个锚使缩短的日志（崩溃修复截断）可被证明：越界的行恰好触发一次从 seq 0 的全量重读，而不是把幽灵值当现值服务。无已注册单元时直接服务 `{asOfSeq: -1, values: {}}`，不触碰持久化；无持久日志的会话以 seam 的 `not found` 拒绝。
`write(session)` 是两个必写点共用的同步切面检查点；载体可以直接调用（非 fail-soft——由 fail-soft 包装层负责遏制）。
## 组合
```yaml
- id: session-projection-cache
name: '@deepseek-ai/dsh-session-projection-cache'
config:
writeEveryEvents: 200
writeIntervalMs: 5000
```
注入 `storageDomain`、`sessionProjections`、`sessionPersistence`、`sessions`。没有这一行时，投影系统只跑 live（水位缓存；冷读在实现了它的载体处退回全量日志加载）。
## 模型体验
无，因为缓存只持久化并恢复 host 侧的、由已入日志会话状态派生的读模型，不触碰任何提示词、消息、schema、流或工具结果。
#### KV 缓存影响
无；缓存从不组装或发送提供方请求。
## 已知局限与延后工作
- **没有淘汰或保留面**——记录按会话累积；清理存储的检查点是带外维护，与会话持久化本身同一立场。
- **间隔节流按会话粗粒度**——计时器在一次干净写入后的首个脏事件时武装；持续的低于阈值的涓流每个间隔写一次，不是滑动窗口。
- **`coldSnapshot` 读取不去重**——同一会话的两个并发冷读各跑一遍阶梯；写回最后者胜（行等价），对列表级调用频率可接受。

50
packages/session-projection/session-projection-cache/package.json Normal file
View File

@@ -0,0 +1,50 @@
{
"name": "@deepseek-ai/dsh-session-projection-cache",
"description": "Persisted projection cache (ctx.sessionProjectionCache): durable per-session projection checkpoints over the domain data form, throttled write-behind, and the cold-read ladder (cache row + persistence tail replay)",
"version": "0.0.1",
"private": true,
"type": "module",
"main": "lib/index.js",
"types": "lib/types/index.d.ts",
"exports": {
".": {
"types": "./lib/types/index.d.ts",
"default": "./lib/index.js"
},
"./invariant": {
"types": "./lib/types/invariant.d.ts",
"default": "./lib/invariant.js"
},
"./src/*": "./src/*",
"./package.json": "./package.json"
},
"files": [
"lib/index.js",
"lib/invariant.js",
"lib/types/**/*.d.ts",
"lib/types/**/*.d.ts.map",
"src"
],
"license": "BSD-3-Clause",
"dependencies": {
"schemastery": "^3.18.0",
"zod": "^4.4.3"
},
"peerDependencies": {
"@deepseek-ai/dsh-invariants": "^0.0.1",
"@deepseek-ai/dsh-session": "^0.0.1",
"@deepseek-ai/dsh-session-persistence": "^0.0.1",
"@deepseek-ai/dsh-session-projection": "^0.0.1",
"@deepseek-ai/dsh-storage-domain": "^0.0.1",
"cordis": "^4.0.0-rc.7"
},
"devDependencies": {
"@deepseek-ai/dsh-invariants": "workspace:^",
"@deepseek-ai/dsh-session": "workspace:^",
"@deepseek-ai/dsh-session-persistence": "workspace:^",
"@deepseek-ai/dsh-session-projection": "workspace:^",
"@deepseek-ai/dsh-storage": "workspace:^",
"@deepseek-ai/dsh-storage-domain": "workspace:^",
"cordis": "^4.0.0-rc.7"
}
}

300
packages/session-projection/session-projection-cache/src/index.ts Normal file
View File

@@ -0,0 +1,300 @@
/**
* Persisted projection cache (`ctx.sessionProjectionCache`): durable
* checkpoints of every registered projection unit's state, one record per
* session on the domain data form (`session_projcache` domain — the shipped
* json backend lands it beside `workspace.json`). The cache is a fold
* shortcut, never an authority: a row is possibly stale (its `seq`
* says how stale) but never wrong, so every write path is fail-soft (a lost
* write costs a longer tail replay on the next cold read) and a
* `ver` mismatch discards the row instead of migrating it. Design
* authority: the session-projection RFC
* (.agents/notes/proposed/architecture/2026-07-27-session-projection-and-command-log.md).
* @module @deepseek-ai/dsh-session-projection-cache
*/
import { Context, Service } from 'cordis'
import z from 'schemastery'
import { snapshotJsonValue } from '@deepseek-ai/dsh-session'
import type { Session, SessionEvent, SessionHeader, SessionId } from '@deepseek-ai/dsh-session'
// Empty type import: applies the package's cordis Context merge
// (`ctx.sessionPersistence`), which this service reads on the cold path.
import type {} from '@deepseek-ai/dsh-session-persistence'
import type { ProjectionCheckpoint, ProjectionSnapshot } from '@deepseek-ai/dsh-session-projection'
import type { KvTable } from '@deepseek-ai/dsh-storage-domain'
import { projectionCacheDomainSpec } from './spec.ts'
import type { CheckpointIdentity, CheckpointRecord } from './spec.ts'
export { checkpointIdentity, checkpointRecord, checkpointRow, projectionCacheDomainSpec } from './spec.ts'
export type { CheckpointIdentity, CheckpointRecord } from './spec.ts'
declare module 'cordis' {
interface Context {
sessionProjectionCache: SessionProjectionCache
}
}
/**
* Plugin config. Both throttle triggers are deployment choices with no
* universally correct value, so the composition states them explicitly
* (cordis.yml); the two mandatory write points (`turn/end` and session
* disposal) are policy, not tunables, and always fire.
*/
export interface Config {
/** Committed events per session that force a durable checkpoint write between mandatory points. */
writeEveryEvents: number
/** Longest time (milliseconds) a dirty checkpoint may stay unwritten between mandatory points. */
writeIntervalMs: number
}
export const Config: z<Config> = z.object({
writeEveryEvents: z.natural().min(1).required(),
writeIntervalMs: z.natural().min(1).required(),
})
/** Per-session write-behind bookkeeping (live sessions only; dropped at retire). */
interface DirtyState {
/** Committed events since the last durable write. */
pending: number
/** Interval trigger armed at the first dirty event after a clean write. */
timer: ReturnType<typeof setTimeout> | undefined
}
/**
* The persisted projection cache service. Opens the `session_projcache`
* domain at init, checkpoints live sessions on a throttled write-behind
* (count/interval triggers from {@link Config}) plus two mandatory points —
* `turn/end` and session disposal (the live-to-cold moment) — and serves the
* cold-read ladder: cached row, persistence `readFrom` tail, registry
* `restore`, durable write-back. Every durable write is fail-soft: failures
* log a warning and the cache self-heals on the next write or cold read.
*/
export class SessionProjectionCache extends Service {
static inject = ['storageDomain', 'sessionProjections', 'sessionPersistence', 'sessions']
static Config: z<Config> = Config
private table?: KvTable<SessionId, CheckpointRecord>
private readonly dirty = new Map<Session, DirtyState>()
constructor(ctx: Context, public config: Config) {
super(ctx, 'sessionProjectionCache')
}
/** Open the domain and install the write-behind listeners. */
protected async [Service.init](): Promise<void> {
const domain = await this.ctx.storageDomain.open(projectionCacheDomainSpec)
this.ctx.effect(() => () => domain.close(), 'sessionProjectionCache.domainClose')
this.table = domain.table('sessions')
this.installWritePath()
}
/**
* The stored record for one session, accepted only when its bound log
* identity matches `expected`. A session id names a slot, not a lifecycle:
* a recreated id or a persistence store swapped under a surviving cache
* must not let an old record seed state folded from an unrelated log.
* Synchronous from the domain's in-memory state.
* @param id - the session whose record is read.
* @param expected - the log identity the caller holds (live or stored header).
* @returns the identity-matching record, or `undefined` (absent or unrelated).
*/
private recordFor(id: SessionId, expected: CheckpointIdentity): CheckpointRecord | undefined {
const record = this.requireTable().get(id)
if (record === undefined) return undefined
return identityMatches(record.identity, expected) ? record : undefined
}
/**
* The zero-I/O listing read: whole values viewed straight from the stored
* rows (version-matching keys only), each cut carried with its watermark
* so a client value store can seed under its higher-seq-wins rule — as
* stale as the last durable checkpoint but never wrong, and never from an
* unrelated log (the caller's header is the identity witness). Fresher
* paths (the history tail baseline, {@link coldSnapshot}) supersede these
* values whenever a session is actually opened.
* @param meta - the listed session's header (identity witness; no log read).
* @returns the cut (`asOfSeq` = lowest served-row watermark), or
* `undefined` when no usable row exists for this lifecycle.
*/
cachedSnapshot(meta: SessionHeader): ProjectionSnapshot | undefined {
const record = this.recordFor(meta.id, identityOf(meta))
if (record === undefined) return undefined
const values = this.ctx.sessionProjections.viewCheckpoint(record.rows)
const keys = Object.keys(values)
if (keys.length === 0) return undefined
// The block carries ONE cut: the lowest served watermark is the seq every
// value is at least current as of (under-claiming is safe under
// higher-seq-wins; over-claiming would let a stale value outrank pushes).
const asOfSeq = Math.min(...keys.map(key => (record.rows[key] as { seq: number }).seq))
return { asOfSeq, values }
}
/**
* Durably checkpoint one live session NOW (both mandatory points call
* this; tests and carriers may too). The registry cut is snapshotted at
* this boundary (states are live references), then the whole record is
* replaced. NOT fail-soft — callers on the fail-soft paths contain it.
* @param session - the live session to checkpoint.
* @returns resolution after durability and event emission.
*/
async write(session: Session): Promise<void> {
const rows = this.ctx.sessionProjections.checkpoint(session)
this.markClean(session)
// Durability barrier: the checkpoint cut was taken above, so flushing
// AFTER it guarantees every event inside the cut is durably logged
// before the cache row lands — a crash can leave the cache behind the
// log (longer tail replay) but never ahead of it (phantom values folded
// from events no stored log contains). At detach the store entry is
// already gone; persistence's own retirement drain covers that path and
// any residual overreach is caught by the cold read's anchored floor.
if (this.ctx.sessions.get(session.id) === session) await this.ctx.sessions.flush(session)
await this.put(session.id, identityOf(session.header), rows)
}
/**
* Cold-read one persisted session's projections with zero full-log load:
* cached rows + a persistence `readFrom` tail from the registry's restore
* floor, refolded by the registry and written back (fail-soft) so the next
* cold read starts closer. A cache row invalidated by a shrunk log
* (crash-repair truncation) triggers one full re-read from seq 0 — the
* ladder's slow rung, still no crash. Rejects when the session has no
* persisted log (`not found` from the persistence seam).
* @param id - the persisted session to read.
* @param signal - optional cancellation for the persistence reads.
* @returns the snapshot cut at the stored log end.
*/
async coldSnapshot(id: SessionId, signal?: AbortSignal): Promise<ProjectionSnapshot> {
const record = this.requireTable().get(id)
const cached = record?.rows ?? {}
const floor = this.ctx.sessionProjections.restoreFloor(cached)
const persistence = this.ctx.sessionPersistence
if (floor === undefined) {
// No unit registered: nothing to fold, but the not-found contract must
// hold in this topology too — the probe read rejects for an absent log
// and dates the empty cut for a present one.
const probe = await persistence.readFrom(id, 0, signal)
return { asOfSeq: probe.events.at(-1)?.seq ?? -1, values: {} }
}
let restored: { snapshot: ProjectionSnapshot; checkpoint: ProjectionCheckpoint }
const tail = await persistence.readFrom(id, floor, signal)
// The tail's stored header is the identity witness: a record bound to a
// different lifecycle (recreated id, swapped store) is discarded whole
// before any of its rows can seed a fold.
const related = record === undefined || identityMatches(record.identity, identityOf(tail.meta))
try {
if (!related) throw new Error('unrelated log identity')
restored = this.ctx.sessionProjections.restore(cached, tail.events, floor)
} catch {
// The recoverable restore failures: an unrelated record, or a row
// overreaching the stored log end (or predating the floor). Both imply
// floor > 0 (baseSeq-0 restores never throw and an unrelated record
// still carried a usable watermark), so the full log is a fresh read.
const whole = await persistence.readFrom(id, 0, signal)
restored = this.ctx.sessionProjections.restore({}, whole.events, 0)
}
await this.putSoft(id, identityOf(tail.meta), restored.checkpoint, 'cold-read write-back')
return restored.snapshot
}
// --- write-behind (throttle + mandatory points) ---
private installWritePath(): void {
// Every committed event advances the dirty counter; turn/end is a
// mandatory point (the durable value most reads want is the turn-final
// one), count/interval throttle the in-turn stream.
this.ctx.on('session/event', (session: Session, event: SessionEvent) => {
if (event.type === 'turn/end') {
void this.flushSoft(session, 'turn/end')
return
}
const state = this.dirty.get(session) ?? { pending: 0, timer: undefined }
this.dirty.set(session, state)
state.pending += 1
if (state.pending >= this.config.writeEveryEvents) {
void this.flushSoft(session, 'count threshold')
return
}
state.timer ??= setTimeout(() => {
void this.flushSoft(session, 'interval')
}, this.config.writeIntervalMs)
})
// Detach (the live-to-cold moment): the second mandatory point. After
// this write the cold-read ladder serves the session from the cache.
// flushSoft's synchronous prefix reads and resets the dirty state, so
// dropping it (timer already cleared by markClean) right after is safe.
this.ctx.on('session/disposed', (session: Session) => {
void this.flushSoft(session, 'detach')
this.markClean(session)
this.dirty.delete(session)
})
// Clear pending timers with the plugin (their sessions outlive the cache).
this.ctx.effect(() => () => {
for (const state of this.dirty.values()) {
if (state.timer !== undefined) clearTimeout(state.timer)
}
this.dirty.clear()
}, 'sessionProjectionCache.timers')
}
/**
* One fail-soft durable checkpoint. Every caller has work by construction:
* the throttle triggers only fire dirty (markClean clears the timer with
* the counter) and the two mandatory points write unconditionally.
*/
private async flushSoft(session: Session, trigger: string): Promise<void> {
try {
await this.write(session)
} catch (error) {
this.ctx.logger.warn(`session projection cache: ${trigger} write for "${session.id}" failed (cache stays stale): ${String(error)}`)
}
}
/** Reset one session's dirty bookkeeping (its checkpoint is being written). */
private markClean(session: Session): void {
const state = this.dirty.get(session)
if (state === undefined) return
state.pending = 0
if (state.timer !== undefined) {
clearTimeout(state.timer)
state.timer = undefined
}
}
/** Replace one session's stored record with its log identity and a detached snapshot of `rows`. */
private async put(id: SessionId, identity: CheckpointIdentity, rows: ProjectionCheckpoint): Promise<void> {
const detached = snapshotJsonValue(rows)
if (detached === undefined) {
throw new TypeError('projection checkpoint is not losslessly JSON-serializable (a unit state violates the plain-JSON contract)')
}
await this.requireTable().put(id, { identity, rows: detached as CheckpointRecord['rows'] })
}
/** Fail-soft {@link put}: cache writes must never fail their caller's read or event path. */
private async putSoft(id: SessionId, identity: CheckpointIdentity, rows: ProjectionCheckpoint, what: string): Promise<void> {
try {
await this.put(id, identity, rows)
} catch (error) {
this.ctx.logger.warn(`session projection cache: ${what} for "${id}" failed (cache stays stale): ${String(error)}`)
}
}
private requireTable(): KvTable<SessionId, CheckpointRecord> {
/* v8 ignore next -- Service.init assigns the table before the service becomes injectable */
if (this.table === undefined) throw new Error('session projection cache is not initialized')
return this.table
}
}
/** Project a header onto the identity fields a record is bound to. */
function identityOf(header: SessionHeader): CheckpointIdentity {
return { createdAt: header.createdAt, ...header.cwd === undefined ? {} : { cwd: header.cwd } }
}
/** Whether a stored record's bound identity names the caller's lifecycle. */
function identityMatches(stored: CheckpointIdentity, expected: CheckpointIdentity): boolean {
return stored.createdAt === expected.createdAt && stored.cwd === expected.cwd
}
export default SessionProjectionCache

35
packages/session-projection/session-projection-cache/src/invariant.ts Normal file
View File

@@ -0,0 +1,35 @@
/**
* Package-owned invariant companion for `@deepseek-ai/dsh-session-projection-cache`.
* @module @deepseek-ai/dsh-session-projection-cache/invariant
*/
/* jscpd:ignore-start */
import type { Context } from 'cordis'
import type { InvariantInstaller } from '@deepseek-ai/dsh-invariants'
const PACKAGE_NAME = '@deepseek-ai/dsh-session-projection-cache'
/** Cordis companion plugin name. */
export const name = 'session-projection-cache-invariant'
/** Service required before the companion can reserve package ownership. */
export const inject = ['invariants']
/**
* No runtime invariant: the cache's correctness relation (a stored row equals
* the registry fold at its `seq` watermark) is only checkable by re-running the
* fold over the persisted log — duplicating the implementation rather than
* detecting drift — and its staleness is by design (fail-soft writes). The
* durable boundary is already schema-validated by the storage-domain layer
* on every reopen, and the read ladder's version/watermark guards are proven
* by the package spec.
*/
const install: InvariantInstaller = () => {}
/**
* Register this package's invariant companion.
* @param ctx - Cordis context carrying the invariant service.
* @returns the installed registration's disposer after setup succeeds.
*/
export const apply = (ctx: Context): Promise<() => void> =>
Promise.resolve(ctx.invariants.register(PACKAGE_NAME, install))
/* jscpd:ignore-end */

71
packages/session-projection/session-projection-cache/src/spec.ts Normal file
View File

@@ -0,0 +1,71 @@
/**
* The session-projcache domain declaration: one `sessions` table keyed by
* {@link SessionId}, each record the full projection checkpoint for one
* session (`key → {ver, seq, val}` rows). The spec object
* is the single source of the domain's identity, version, and record schema;
* the storage-domain routing decides the medium (the shipped composition's
* json backend lands it at `<root>/session_projcache.json`, beside
* `workspace.json`).
* @module @deepseek-ai/dsh-session-projection-cache/src/spec
*/
import { z } from 'zod'
import { SessionId } from '@deepseek-ai/dsh-session'
import { defineDomain, domainTable } from '@deepseek-ai/dsh-storage-domain'
/**
* One persisted checkpoint row (the RFC's `(sessionId, key, ver, seq, val)`
* minus the two record keys). `val` is the unit's internal state — plain
* JSON by the unit contract; `z.json()` enforces that at the durable
* boundary. A row is never wrong, only possibly stale: `seq` says exactly
* how stale, and a `ver` mismatch against the live unit's `stateVersion`
* discards it at read time (never a migration).
*/
export const checkpointRow = z.object({
ver: z.number().int().nonnegative(),
seq: z.number().int().gte(-1),
val: z.json(),
})
/**
* The stored-log identity a record is bound to: the immutable header fields
* that distinguish one session lifecycle from another under the same id. A
* session id names a slot, not a lifecycle — a deleted-then-recreated id, or
* a persistence root swapped under a surviving cache, would otherwise let an
* old row pass every watermark check and seed state folded from an unrelated
* log. Reads validate this against the live header (listing) or the stored
* header (cold read) before accepting any row.
*/
export const checkpointIdentity = z.object({
createdAt: z.number().int().nonnegative(),
cwd: z.string().optional(),
})
/** The identity fields a record is bound to, inferred from {@link checkpointIdentity}. */
export type CheckpointIdentity = z.infer<typeof checkpointIdentity>
/**
* One session's stored record: the log identity it was folded from plus its
* checkpoint rows keyed by projection key. The whole record is replaced on
* every write (whole-value discipline — the registry checkpoint is always
* the complete per-session cut).
*/
export const checkpointRecord = z.object({
identity: checkpointIdentity,
rows: z.record(z.string(), checkpointRow),
})
/** One stored per-session checkpoint record, inferred from {@link checkpointRecord}. */
export type CheckpointRecord = z.infer<typeof checkpointRecord>
/**
* The session-projcache domain spec. Version bumps discard the whole medium
* (cache semantics: a stale or unreadable cache costs a longer tail replay,
* never a wrong value). v2 added the record's log-identity binding; v3
* renamed the row fields to `ver`/`seq`/`val`.
*/
export const projectionCacheDomainSpec = defineDomain({
name: 'session_projcache',
version: 3,
tables: { sessions: domainTable<SessionId, CheckpointRecord>(checkpointRecord) },
})

387
packages/session-projection/session-projection-cache/tests/cache.spec.ts Normal file
View File

@@ -0,0 +1,387 @@
/**
* SessionProjectionCache behavior: mandatory-point writes (turn/end, detach),
* count/interval throttling between them, fail-soft durability (a failed
* write logs and stays stale, never throws into the event path), and the
* cold-read ladder (cached row + readFrom tail + registry restore +
* write-back; version bump and shrunk-log rows degrade to a full re-read).
*/
import { afterEach, describe, expect, it, vi } from 'vitest'
import { Context } from 'cordis'
import { z } from 'zod'
import Storage from '@deepseek-ai/dsh-storage'
import { DomainFacility } from '@deepseek-ai/dsh-storage-domain'
import SessionStore, { SessionId } from '@deepseek-ai/dsh-session'
import type { Session, SessionEvent } from '@deepseek-ai/dsh-session'
import SessionProjectionRegistry from '@deepseek-ai/dsh-session-projection'
import type { ProjectionDefinition } from '@deepseek-ai/dsh-session-projection'
import { MemoryMediaPool, MemoryStorageBackend } from '../../../storage/storage-domain/tests/helpers/memory-backend.ts'
import SessionProjectionCache from '../src/index.ts'
declare module '@deepseek-ai/dsh-session-projection/types' {
interface SessionProjectionMap {
'cache-test/marks': { marks: string[] }
}
}
declare module '@deepseek-ai/dsh-session' {
interface SessionEventMap {
'cache-test/mark': { marks: string[] }
}
interface OutOfBandSessionEventMap {
'cache-test/mark': true
}
}
type MarksState = { marks: string[] } | null
const marksUnit = (stateVersion = 1): ProjectionDefinition<'cache-test/marks', MarksState> => ({
key: 'cache-test/marks',
schema: z.object({ marks: z.array(z.string()) }),
init: () => null,
apply: (state, event) => (event.type === 'cache-test/mark' ? (event).data : state),
view: state => state ?? { marks: [] },
stateVersion,
})
/** A persistence double serving readFrom over a fixed per-id stored log (headers stamp createdAt 0). */
function fakePersistence(logs: Map<string, SessionEvent[]>) {
const readFrom = vi.fn(async (id: SessionId, fromSeq: number) => {
const events = logs.get(String(id))
if (events === undefined) throw new Error(`session "${id}" not found`)
return {
meta: { version: 0, id, createdAt: 0 },
events: events.filter(event => event.seq >= fromSeq),
}
})
return { readFrom }
}
/** Header shape for cachedSnapshot calls (fake logs stamp createdAt 0, no cwd). */
const headerOf = (id: SessionId, createdAt = 0, cwd?: string) =>
({ version: 0, id, createdAt, ...cwd === undefined ? {} : { cwd } })
interface HarnessOptions {
pool?: MemoryMediaPool
config?: { writeEveryEvents: number; writeIntervalMs: number }
stateVersion?: number
logs?: Map<string, SessionEvent[]>
}
const contexts: Context[] = []
async function harness(options: HarnessOptions = {}) {
const pool = options.pool ?? new MemoryMediaPool()
const logs = options.logs ?? new Map<string, SessionEvent[]>()
const ctx = new Context()
contexts.push(ctx)
await ctx.plugin(Storage)
ctx.storage.backend.register('memory', new MemoryStorageBackend(pool))
const facility = new DomainFacility(ctx, { backend: 'memory', routes: {} })
ctx.storage.mount('domain', facility)
ctx.provide('storageDomain', facility)
await ctx.plugin(SessionStore)
await ctx.plugin(SessionProjectionRegistry)
ctx.sessionProjections.register(marksUnit(options.stateVersion))
const persistence = fakePersistence(logs)
ctx.provide('sessionPersistence', persistence as never)
const fiber = await ctx.plugin(SessionProjectionCache, options.config ?? { writeEveryEvents: 100, writeIntervalMs: 60_000 })
return { ctx, pool, logs, fiber, persistence, cache: ctx.sessionProjectionCache }
}
const mark = (session: Session, marks: string[]): SessionEvent =>
session.append('cache-test/mark', { marks })
const endTurn = (session: Session): SessionEvent =>
session.append('turn/end', { turn: 1, reason: { kind: 'completed' } })
/** The stored medium record for one session id (undefined = never written). */
function storedRecord(pool: MemoryMediaPool, id: Session['id']) {
return pool.media.get('session_projcache')?.tables.get('sessions')?.get(String(id)) as
{
identity: { createdAt: number; cwd?: string }
rows: Record<string, { ver: number; seq: number; val: unknown }>
} | undefined
}
/** The stored medium rows for one session id (undefined = never written). */
function storedRows(pool: MemoryMediaPool, id: Session['id']) {
return storedRecord(pool, id)?.rows
}
/** Wait until queued fail-soft writes (event-listener fire-and-forget) drain. */
const settle = () => new Promise(resolve => setTimeout(resolve, 0))
afterEach(async () => {
vi.useRealTimers()
await Promise.all(contexts.splice(0).map(ctx => ctx.fiber.dispose()))
})
describe('SessionProjectionCache write policy', () => {
it('writes a durable checkpoint at turn/end (mandatory point)', async () => {
const { ctx, pool } = await harness()
const session = ctx.sessions.create(SessionId('turn-end'))
mark(session, ['a'])
expect(storedRows(pool, session.id)).toBeUndefined() // throttled: no write yet
const end = endTurn(session)
await settle()
const rows = storedRows(pool, session.id)
expect(rows?.['cache-test/marks']).toEqual({ ver: 1, seq: end.seq, val: { marks: ['a'] } })
})
it('writes at session disposal (detach, the live-to-cold moment)', async () => {
const { ctx, pool } = await harness()
// Sessions dispose with their owning fiber: create in a child plugin.
let session: Session | undefined
const owner = await ctx.plugin(Object.assign((inner: Context) => {
session = inner.sessions.create(SessionId('detach'))
}, { inject: ['sessions'] }))
if (session === undefined) throw new Error('session was not created')
mark(session, ['live'])
await owner.dispose()
await settle()
expect(storedRows(pool, session.id)?.['cache-test/marks']?.val).toEqual({ marks: ['live'] })
})
it('flushes when the in-turn event count reaches the configured threshold', async () => {
const { ctx, pool } = await harness({ config: { writeEveryEvents: 3, writeIntervalMs: 60_000 } })
const session = ctx.sessions.create(SessionId('count'))
mark(session, ['1'])
mark(session, ['2'])
await settle()
expect(storedRows(pool, session.id)).toBeUndefined()
mark(session, ['3'])
await settle()
expect(storedRows(pool, session.id)?.['cache-test/marks']?.val).toEqual({ marks: ['3'] })
})
it('flushes on the configured interval when the count threshold is not reached', async () => {
vi.useFakeTimers()
const { ctx, pool } = await harness({ config: { writeEveryEvents: 100, writeIntervalMs: 250 } })
const session = ctx.sessions.create(SessionId('interval'))
mark(session, ['slow'])
await vi.advanceTimersByTimeAsync(249)
expect(storedRows(pool, session.id)).toBeUndefined()
await vi.advanceTimersByTimeAsync(1)
await vi.advanceTimersByTimeAsync(0)
expect(storedRows(pool, session.id)?.['cache-test/marks']?.val).toEqual({ marks: ['slow'] })
})
it('write() on a never-dirty session checkpoints directly and rejects a non-JSON unit state', async () => {
const { ctx, pool } = await harness()
// Never dirtied: no events — write() still lands the init-derived cut.
const clean = ctx.sessions.create(SessionId('clean-write'))
await ctx.sessionProjectionCache.write(clean)
expect(storedRows(pool, clean.id)?.['cache-test/marks']).toEqual({ ver: 1, seq: -1, val: null })
// A unit whose state violates the plain-JSON contract fails the write loud.
ctx.sessionProjections.register({
key: 'cache-test/marks2' as never,
schema: { parse: (value: unknown) => value } as never,
init: () => new Map<string, string>(),
apply: (state: unknown) => state,
view: () => null as never,
stateVersion: 1,
})
await expect(ctx.sessionProjectionCache.write(clean)).rejects.toThrow('not losslessly JSON-serializable')
})
it('plugin disposal clears armed interval timers and leaves cleaned sessions alone', async () => {
vi.useFakeTimers()
const { ctx, pool, fiber } = await harness({ config: { writeEveryEvents: 100, writeIntervalMs: 5000 } })
const armed = ctx.sessions.create(SessionId('armed'))
const cleaned = ctx.sessions.create(SessionId('cleaned'))
mark(armed, ['pending']) // timer armed, no write yet
mark(cleaned, ['done'])
endTurn(cleaned) // mandatory write; markClean leaves {pending: 0, timer: undefined} in the map
await vi.advanceTimersByTimeAsync(0)
await fiber.dispose()
// The armed timer died with the plugin: advancing time writes nothing.
await vi.advanceTimersByTimeAsync(10_000)
expect(storedRows(pool, armed.id)).toBeUndefined()
})
it('contains a durable write failure: logs a warning, event path unharmed, next write self-heals', async () => {
const { ctx, pool } = await harness()
const warn = vi.spyOn(ctx.logger, 'warn').mockImplementation(() => {})
const session = ctx.sessions.create(SessionId('fail-soft'))
mark(session, ['x'])
pool.failNextWrites = 1
endTurn(session)
await settle()
expect(storedRows(pool, session.id)).toBeUndefined()
expect(warn).toHaveBeenCalledWith(expect.stringContaining('turn/end write for "fail-soft" failed'))
// Self-heal: the next mandatory point writes the current cut.
mark(session, ['y'])
endTurn(session)
await settle()
expect(storedRows(pool, session.id)?.['cache-test/marks']?.val).toEqual({ marks: ['y'] })
})
})
describe('SessionProjectionCache cold read', () => {
const storedLog = (marks: string[][]): SessionEvent[] => {
const events: SessionEvent[] = [
{ type: 'turn/start', seq: 0, time: 0, data: { turn: 1, trigger: { kind: 'message', source: { kind: 'user' } } } },
]
for (const m of marks) {
events.push({ type: 'cache-test/mark', seq: events.length, time: events.length, data: { marks: m } })
}
events.push({ type: 'turn/end', seq: events.length, time: events.length, data: { turn: 1, reason: { kind: 'completed' } } })
return events
}
/** Pre-seed the medium with one stored checkpoint record (before the domain opens). */
function seedRow(
pool: MemoryMediaPool,
id: string,
row: { ver: number; seq: number; val: unknown },
identity: { createdAt: number; cwd?: string } = { createdAt: 0 },
): void {
pool.versions.set('session_projcache', 3)
pool.media.set('session_projcache', {
tables: new Map([['sessions', new Map([[id, { identity, rows: { 'cache-test/marks': row } }]])]]),
global: null,
})
}
it('serves a cold session from the cache row plus a bounded tail read, and writes the refresh back', async () => {
const pool = new MemoryMediaPool()
const logs = new Map([['cold', storedLog([['a'], ['a', 'b']])]])
// A warm-era checkpoint at watermark 1 (only ['a'] folded).
seedRow(pool, 'cold', { ver: 1, seq: 1, val: { marks: ['a'] } })
const { cache, persistence, pool: samePool } = await harness({ pool, logs })
const id = SessionId('cold')
const snapshot = await cache.coldSnapshot(id)
expect(snapshot.values['cache-test/marks']).toEqual({ marks: ['a', 'b'] })
expect(snapshot.asOfSeq).toBe(3)
// The tail read was bounded by the anchored floor (watermark 1 -> floor 1), not 0.
expect(persistence.readFrom).toHaveBeenCalledWith(id, 1, undefined)
// Write-back: the stored row advanced to the served cut.
expect(storedRows(samePool, id)?.['cache-test/marks'])
.toEqual({ ver: 1, seq: 3, val: { marks: ['a', 'b'] } })
})
it('discards a version-mismatched row and refolds the full log', async () => {
const pool = new MemoryMediaPool()
const logs = new Map([['bumped', storedLog([['a']])]])
seedRow(pool, 'bumped', { ver: 1, seq: 2, val: { marks: ['stale'] } })
const { cache, persistence } = await harness({ pool, logs, stateVersion: 2 })
const snapshot = await cache.coldSnapshot(SessionId('bumped'))
expect(snapshot.values['cache-test/marks']).toEqual({ marks: ['a'] })
// Mismatch pulls the floor to 0: one full read, no second pass needed.
expect(persistence.readFrom).toHaveBeenCalledTimes(1)
expect(persistence.readFrom).toHaveBeenCalledWith(SessionId('bumped'), 0, undefined)
})
it('detects a log shrunk below the row watermark and degrades to one full re-read', async () => {
const pool = new MemoryMediaPool()
const logs = new Map([['shrunk', storedLog([['a']])]]) // seqs 0..2
seedRow(pool, 'shrunk', { ver: 1, seq: 9, val: { marks: ['ghost'] } })
const { cache, persistence } = await harness({ pool, logs })
const snapshot = await cache.coldSnapshot(SessionId('shrunk'))
expect(snapshot.values['cache-test/marks']).toEqual({ marks: ['a'] })
expect(snapshot.asOfSeq).toBe(2)
// Anchored tail read (floor 9) came back empty -> full re-read from 0.
expect(persistence.readFrom).toHaveBeenNthCalledWith(1, SessionId('shrunk'), 9, undefined)
expect(persistence.readFrom).toHaveBeenNthCalledWith(2, SessionId('shrunk'), 0, undefined)
})
it('write-back failure is contained: the snapshot is still served', async () => {
const pool = new MemoryMediaPool()
const logs = new Map([['soft', storedLog([['a']])]])
const { ctx, cache } = await harness({ pool, logs })
const warn = vi.spyOn(ctx.logger, 'warn').mockImplementation(() => {})
pool.failNextWrites = 1
const snapshot = await cache.coldSnapshot(SessionId('soft'))
expect(snapshot.values['cache-test/marks']).toEqual({ marks: ['a'] })
expect(warn).toHaveBeenCalledWith(expect.stringContaining('cold-read write-back for "soft" failed'))
})
it('rejects for a session with no persisted log', async () => {
const { cache } = await harness()
await expect(cache.coldSnapshot(SessionId('absent'))).rejects.toThrow('not found')
})
it('discards a record bound to a different log lifecycle and refolds from the actual log', async () => {
const pool = new MemoryMediaPool()
const logs = new Map([['reborn', storedLog([['real']])]]) // stored header stamps createdAt 0
// A checkpoint from a PRIOR lifecycle of the same id (different createdAt):
// its rows pass every watermark check, but the identity does not match.
seedRow(pool, 'reborn', { ver: 1, seq: 2, val: { marks: ['phantom'] } }, { createdAt: 999 })
const { cache, pool: samePool } = await harness({ pool, logs })
const snapshot = await cache.coldSnapshot(SessionId('reborn'))
expect(snapshot.values['cache-test/marks']).toEqual({ marks: ['real'] })
// The write-back rebinds the record to the actual log's identity.
expect(storedRecord(samePool, SessionId('reborn'))?.identity).toEqual({ createdAt: 0 })
})
it('cachedSnapshot returns undefined when every stored row is version-mismatched', async () => {
const pool = new MemoryMediaPool()
seedRow(pool, 'all-stale', { ver: 99, seq: 4, val: { marks: ['old'] } })
const { cache } = await harness({ pool })
expect(cache.cachedSnapshot(headerOf(SessionId('all-stale')))).toBeUndefined()
})
it('binds identity on cwd too: a matching cwd serves, a moved session does not', async () => {
const pool = new MemoryMediaPool()
seedRow(pool, 'homed', { ver: 1, seq: 2, val: { marks: ['w'] } }, { createdAt: 0, cwd: '/work' })
const { cache } = await harness({ pool })
const id = SessionId('homed')
expect(cache.cachedSnapshot(headerOf(id, 0, '/work'))?.values['cache-test/marks']).toEqual({ marks: ['w'] })
expect(cache.cachedSnapshot(headerOf(id, 0, '/elsewhere'))).toBeUndefined()
expect(cache.cachedSnapshot(headerOf(id, 0))).toBeUndefined()
})
it('dates an empty stored log at -1 in the zero-units topology', async () => {
const pool = new MemoryMediaPool()
const logs = new Map([['empty', [] as SessionEvent[]]])
const ctx = new Context()
contexts.push(ctx)
await ctx.plugin(Storage)
ctx.storage.backend.register('memory', new MemoryStorageBackend(pool))
const facility = new DomainFacility(ctx, { backend: 'memory', routes: {} })
ctx.storage.mount('domain', facility)
ctx.provide('storageDomain', facility)
await ctx.plugin(SessionStore)
await ctx.plugin(SessionProjectionRegistry)
ctx.provide('sessionPersistence', fakePersistence(logs) as never)
await ctx.plugin(SessionProjectionCache, { writeEveryEvents: 100, writeIntervalMs: 60_000 })
await expect(ctx.sessionProjectionCache.coldSnapshot(SessionId('empty')))
.resolves.toEqual({ asOfSeq: -1, values: {} })
})
it('cachedSnapshot serves identity-matching rows with the cut watermark and refuses unrelated ones', async () => {
const pool = new MemoryMediaPool()
seedRow(pool, 'listed', { ver: 1, seq: 4, val: { marks: ['t'] } })
const { cache } = await harness({ pool })
const id = SessionId('listed')
// Matching header: values plus the watermark the client seeds under.
expect(cache.cachedSnapshot(headerOf(id))).toEqual({ asOfSeq: 4, values: { 'cache-test/marks': { marks: ['t'] } } })
// A recreated id (different createdAt): the record is unrelated — no block.
expect(cache.cachedSnapshot(headerOf(id, 777))).toBeUndefined()
// Unknown id: no block.
expect(cache.cachedSnapshot(headerOf(SessionId('never-cached')))).toBeUndefined()
})
it('holds the not-found contract with zero registered units, and dates the empty cut for a present log', async () => {
// Same composition minus any registered unit: restoreFloor is undefined,
// yet coldSnapshot must still reject for an absent log (probe read) and
// serve an empty cut at the stored end for a present one.
const pool = new MemoryMediaPool()
const logs = new Map([['bare', storedLog([['a']])]]) // seqs 0..2
const ctx = new Context()
contexts.push(ctx)
await ctx.plugin(Storage)
ctx.storage.backend.register('memory', new MemoryStorageBackend(pool))
const facility = new DomainFacility(ctx, { backend: 'memory', routes: {} })
ctx.storage.mount('domain', facility)
ctx.provide('storageDomain', facility)
await ctx.plugin(SessionStore)
await ctx.plugin(SessionProjectionRegistry)
ctx.provide('sessionPersistence', fakePersistence(logs) as never)
await ctx.plugin(SessionProjectionCache, { writeEveryEvents: 100, writeIntervalMs: 60_000 })
await expect(ctx.sessionProjectionCache.coldSnapshot(SessionId('absent'))).rejects.toThrow('not found')
await expect(ctx.sessionProjectionCache.coldSnapshot(SessionId('bare')))
.resolves.toEqual({ asOfSeq: 2, values: {} })
})
})

39
packages/session-projection/session-projection-cache/tsconfig.json Normal file
View File

@@ -0,0 +1,39 @@
{
"extends": "../../../tsconfig.base.json",
"compilerOptions": {
"rootDir": "src",
"outDir": "lib/types"
},
"include": [
"src"
],
"references": [
{
"path": "../../../vendor/cosmokit"
},
{
"path": "../../../vendor/cordis"
},
{
"path": "../../../vendor/schemastery"
},
{
"path": "../../core/session"
},
{
"path": "../../session-persistence/session-persistence"
},
{
"path": "../session-projection"
},
{
"path": "../../storage/storage"
},
{
"path": "../../storage/storage-domain"
},
{
"path": "../../support/invariants"
}
]
}

4
packages/session-projection/session-projection/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/session-projection/session-projection/README.md
README.md: 2e026aab55933c96ba961481f9597bc18cbbe910
README.zh.md: a3e0b0f46466d19321b0950dc41d06473a54a1ce
README.md: f4898b8e567fa5998c18111c5f4e27a8a350a42e
README.zh.md: 385862868df495a5c857d91c32f6503c3ef72025

4
packages/session-projection/session-projection/README.md
View File

@@ -23,7 +23,7 @@ Session-projection seam. It owns `ctx.sessionProjections`, the registry that DRI
- **Same-reference means no work.** `apply` MUST return the same state reference for events that do not concern the unit; the drive gates the change feed on `Object.is`, so non-matching events cost one call and nothing downstream.
- **Whole-value event rule (load-bearing).** A state-carrying log event MUST carry the complete post-change state, never a bare delta — it keeps every transition trivially cheap and every served value self-describing (last-wins for consumers).
- **Synchronous unit discipline.** `init`/`apply`/`view` MUST be synchronous; carriers read `snapshot()` in the same tick as their page slice, which is what makes `asOfSeq` one consistent cut. An accidentally-async `view` returns a Promise, which fails the boundary `schema.parse` loudly.
- **State is plain JSON, `stateVersion` is its invalidation anchor.** The persisted projection cache (a later phase) stores `(sessionId, key, stateVersion, observedSeq, stateJson)` rows; bump `stateVersion` whenever the state shape or the fold semantics change so stale rows are discarded instead of forward-applied into garbage.
- **State is plain JSON, `stateVersion` is its invalidation anchor.** The persisted projection cache stores `(sessionId, key, ver, seq, val)` rows; bump `stateVersion` whenever the state shape or the fold semantics change so stale rows are discarded instead of forward-applied into garbage.
- **No wire vocabulary here.** The registry exposes only the change feed and the snapshot read face; carriers (api-proxy) mint their own frames (`session/projection`) and blocks from them.
- **Optional seam.** Domain plugins register under `ctx.inject(['sessionProjections'], …)` so headless assemblies without the registry stay unaffected; carriers use `ctx.get('sessionProjections')` and omit their block/frames entirely when the registry is absent.
@@ -43,5 +43,5 @@ None; projections never assemble or send provider requests.
- **Every tail page carries every registered key** — there is no per-key opt-out or lazy-key request shape yet; acceptable while values are UI-scale whole states (a todo list, a goal snapshot), revisit if a domain's value grows large.
- **Eager drive touches every unit per event** — cheap by construction (whole-value rule, same-reference gate), but a hot path would justify per-unit event-type prefilters, addable without contract change.
- **The persisted projection cache is a later phase** — cells live in memory only; a restart rebuilds by folding the in-memory log on first touch. The `stateVersion` field is the forward-declared invalidation anchor for that phase.
- **Registry cells live in memory only** — a restart rebuilds by folding the log on first touch; compositions that mount `dsh-session-projection-cache` seed that fold from persisted rows instead.
- **Synchronous unit discipline is only partially mechanical** — the boundary `schema.parse` rejects a Promise-returning `view`, but an `apply` that blocks or reads torn non-session state is a review concern; the invariant companion documents why no runtime check exists.

4
packages/session-projection/session-projection/README.zh.md
View File

@@ -23,7 +23,7 @@
- **同引用即无工作。** 对与单元无关的事件，`apply` 必须返回同一个状态引用；驱动以 `Object.is` 把守变更流，因此不匹配的事件只花一次调用，不产生任何下游工作。
- **全量值事件规则（承重）。** 携带状态的日志事件必须携带变更后的完整状态，绝不携带裸增量——这让每次状态转移始终足够廉价，也让每个被供给的值自描述（对消费方即 last-wins）。
- **单元的同步纪律。** `init`/`apply`/`view` 必须是同步的；载体在切出页面切片的同一 tick 内读取 `snapshot()`，`asOfSeq` 之所以是一个一致切面正系于此。误写成异步的 `view` 会返回 Promise，让边界的 `schema.parse` 当场大声失败。
- **状态是纯 JSON，`stateVersion` 是其失效锚点。** 持久投影缓存（persisted projection cache，后续阶段）存储 `(sessionId, key, stateVersion, observedSeq, stateJson)` 行；状态形状或折叠语义一旦变化就递增 `stateVersion`，使陈旧行被丢弃，而不是被正向 apply 成垃圾。
- **状态是纯 JSON，`stateVersion` 是其失效锚点。** 持久投影缓存（persisted projection cache）存储 `(sessionId, key, ver, seq, val)` 行；状态形状或折叠语义一旦变化就递增 `stateVersion`，使陈旧行被丢弃，而不是被正向 apply 成垃圾。
- **本层没有协议词汇。** 注册表只暴露变更流与快照读取面；载体（api-proxy）据此自铸各自的帧（`session/projection`）与块。
- **可选 seam。** 领域插件在 `ctx.inject(['sessionProjections'], …)` 下注册，因此不带注册表的 headless 组装完全不受影响；载体使用 `ctx.get('sessionProjections')`，注册表缺席时完全省略自己的块与帧。
@@ -43,5 +43,5 @@
- **每个尾页携带每个已注册的 key**——尚无逐 key 的 opt-out 或惰性 key 请求形状；在值都是 UI 量级的全量状态（一张 todo 清单、一份 goal 快照）时可以接受，若某领域的值变大再重议。
- **正向驱动（eager drive）逐事件触达每个单元**——按构造开销很低（全量值规则、同引用闸门），但若出现热点路径，可加按单元的事件类型预过滤，契约不变。
- **持久投影缓存属于后续阶段**——cell 目前只活在内存里；重启后首次触达时靠折叠内存日志重建。`stateVersion` 字段是为该阶段预先声明的失效锚点。
- **注册表 cell 只活在内存里**——重启后首次触达时靠折叠日志重建；挂载了 `dsh-session-projection-cache` 的组合改由持久行播种该折叠。
- **单元同步纪律只有部分可机械把关**——边界 `schema.parse` 能拒绝返回 Promise 的 `view`，但阻塞的 `apply`、或读取撕裂的非会话状态的 `apply`，只能靠评审把关；invariant 配套记载了为何不存在运行时检查。

156
packages/session-projection/session-projection/src/index.ts
View File

@@ -66,9 +66,9 @@ export interface ProjectionDefinition<K extends keyof SessionProjectionMap, S> {
view(state: S): SessionProjectionMap[K]
/**
* Persisted-cache invalidation anchor: bump whenever the state shape or the
* fold semantics change, so persisted `(sessionId, key, stateVersion,
* observedSeq, state)` rows from an older unit are discarded instead of
* being forward-applied into garbage. Non-negative integer.
* fold semantics change, so persisted `(sessionId, key, ver, seq, val)`
* rows from an older unit are discarded instead of being forward-applied
* into garbage. Non-negative integer.
*/
stateVersion: number
}
@@ -97,6 +97,26 @@ export interface ProjectionSnapshot {
values: Partial<SessionProjectionMap>
}
/**
* One unit's checkpoint: its internal state (plain JSON by the unit
* contract), the seq of the last event folded into it, and the unit
* `stateVersion` that produced it — the persisted projection-cache row
* `(sessionId, key, ver, seq, val)` minus the two outer keys. A row is
* never authoritative, only a fold shortcut: `restore` discards it on a
* version mismatch or when it claims events past the stored log end.
*/
export interface ProjectionCheckpointRow {
/** The registering unit's `stateVersion` at fold time. */
ver: number
/** Seq of the last event folded into `val`; -1 for the empty log. */
seq: number
/** The unit's internal state — plain JSON per the unit contract. */
val: unknown
}
/** Checkpoint rows keyed by projection key (one session's persisted cache value). */
export type ProjectionCheckpoint = Record<string, ProjectionCheckpointRow>
/** Type-erased unit view the drive machinery works with (the register seam already proved the typed contract). */
interface ErasedDefinition {
key: string
@@ -206,6 +226,136 @@ export class SessionProjectionRegistry extends Service {
return { asOfSeq: session.seq - 1, values: values }
}
/**
* State-level checkpoint of every registered unit for one session, read
* from the watermark cache (missing cells fold lazily over the in-memory
* log). This is the write side of the persisted projection cache: the
* returned rows are the `(key → {ver, seq, val})` part of the durable
* `(sessionId, key, ver, seq, val)`
* rows. Every `val` is a DETACHED structured clone — never the live
* cell reference: the watermark cache is this registry's authoritative
* mutable state, and a caller reaching the live reference could corrupt
* every subsequent snapshot and frame through it (plain JSON by the unit
* contract, so the clone is total).
* @param session - the session whose unit states are checkpointed.
* @returns one row per registered key; empty when no unit is registered.
*/
checkpoint(session: Session): ProjectionCheckpoint {
const rows: ProjectionCheckpoint = {}
for (const registration of this.registrations.values()) {
const cell = this.cellFor(registration, session)
rows[registration.def.key] = {
ver: registration.def.stateVersion,
seq: cell.observedSeq,
val: structuredClone(cell.state),
}
}
return rows
}
/**
* The stored seq a {@link restore} tail read over `checkpoint` must start
* at: one event BELOW the lowest usable watermark (a row is usable when
* its `ver` matches the live unit's `stateVersion`; an absent or mismatched row
* pulls the floor to `0` — that key must refold the full log). The
* one-below anchor is load-bearing: the tail then proves how far the
* stored log still extends, so {@link restore} can detect a log that
* shrank below a row's watermark (crash-repair truncation) instead of
* serving the stale row as current — an empty tail read from the anchor
* yields an end below every watermark and the restore rejects for a full
* re-read.
* @param checkpoint - persisted rows for one session (possibly stale or empty).
* @returns the seq to hand the persistence `readFrom`, or `undefined`
* when no unit is registered (no read needed — {@link restore} would
* serve empty values regardless).
*/
restoreFloor(checkpoint: ProjectionCheckpoint): number | undefined {
let floor: number | undefined
for (const registration of this.registrations.values()) {
const row = checkpoint[registration.def.key]
const need = row !== undefined && row.ver === registration.def.stateVersion
? Math.max(row.seq + 1, 0)
: 0
floor = floor === undefined ? need : Math.min(floor, need)
}
return floor === undefined ? undefined : Math.max(floor - 1, 0)
}
/**
* View a checkpoint's rows without any log read: for every registered
* unit whose row's `ver` matches, serve the schema-validated
* `view` of the stored state; mismatched or absent rows leave their key
* absent (a cold or listing consumer treats it as not-yet-available and a
* fuller read path refolds it). The zero-I/O rung of the read ladder —
* values are as stale as their rows, never wrong.
* @param checkpoint - persisted rows for one session (possibly stale or empty).
* @returns whole values per key with a usable row; empty when none.
*/
viewCheckpoint(checkpoint: ProjectionCheckpoint): Partial<SessionProjectionMap> {
const values: Record<string, unknown> = {}
for (const registration of this.registrations.values()) {
const def = registration.def
const row = checkpoint[def.key]
if (row === undefined || row.ver !== def.stateVersion) continue
values[def.key] = def.schema.parse(def.view(row.val))
}
return values
}
/**
* Cold read: fold every registered unit over a stored log suffix, seeding
* each from its checkpoint row when usable — the one read recipe (cached
* state + forward tail replay + `view`) applied without a live `Session`.
* Call with the events returned by a persistence
* `readFrom(id, restoreFloor(checkpoint))` and that same floor as
* `baseSeq`; the floor's one-below anchor makes the supplied end honest,
* so a shrunk log is detected here. A row is usable iff its
* `ver` matches the live unit's `stateVersion`, it does not predate `baseSeq`
* (`seq >= baseSeq - 1`), and it does not claim events past the
* supplied end (`seq <= endSeq`); an unusable row is discarded
* and its key refolds from `init` — which is only sound over the full
* log, so a discarded row with `baseSeq > 0` throws (the caller re-reads
* from seq 0, e.g. after a crash-repair truncation shrank the log below
* a row's watermark).
* @param checkpoint - persisted rows for one session (possibly stale or empty).
* @param events - the stored events with `seq >= baseSeq`, in seq order.
* @param baseSeq - the seq `events` starts at (its first event's seq when non-empty).
* @returns the snapshot cut at the supplied log end (`asOfSeq` is the last
* supplied event's seq, `baseSeq - 1` for an empty tail) plus the
* refreshed checkpoint rows at that cut, ready for a durable write-back.
*/
restore(checkpoint: ProjectionCheckpoint, events: readonly SessionEvent[], baseSeq: number):
{ snapshot: ProjectionSnapshot; checkpoint: ProjectionCheckpoint } {
const endSeq = events.at(-1)?.seq ?? baseSeq - 1
const values: Record<string, unknown> = {}
const refreshed: ProjectionCheckpoint = {}
for (const registration of this.registrations.values()) {
const def = registration.def
const row = checkpoint[def.key]
const usable = row !== undefined
&& row.ver === def.stateVersion
&& row.seq >= baseSeq - 1
&& row.seq <= endSeq
if (!usable && baseSeq > 0) {
throw new Error(
`session projection ${JSON.stringify(def.key)} cannot restore from seq ${baseSeq}: `
+ 'its checkpoint row is missing, version-mismatched, or beyond the supplied log end; re-read from seq 0',
)
}
let state = usable ? row.val : def.init()
const from = usable ? row.seq : baseSeq - 1
for (const event of events) {
if (event.seq > from) state = def.apply(state, event)
}
values[def.key] = def.schema.parse(def.view(state))
refreshed[def.key] = { ver: def.stateVersion, seq: endSeq, val: state }
}
return {
snapshot: { asOfSeq: endSeq, values: values },
checkpoint: refreshed,
}
}
/** Fold one unit from init over `events`, producing a cell watermarked at the last folded event. */
private buildCell(def: ErasedDefinition, events: readonly SessionEvent[]): UnitCell {
let state = def.init()

148
packages/session-projection/session-projection/tests/registry.spec.ts
View File

@@ -169,6 +169,154 @@ describe('SessionProjectionRegistry drive', () => {
expect(ctx.sessionProjections.snapshot(session).values).toEqual({})
})
it('checkpoints every registered unit with its stateVersion and per-cell watermark', async () => {
const { ctx, session } = await harness()
ctx.sessionProjections.register(marksUnit())
ctx.sessionProjections.register({ ...countUnit(), stateVersion: 7 })
const markEvent = mark(session, ['a'])
const rows = ctx.sessionProjections.checkpoint(session)
expect(rows['test/marks']).toEqual({ ver: 1, seq: markEvent.seq, val: { marks: ['a'] } })
expect(rows['test/count']).toEqual({ ver: 7, seq: markEvent.seq, val: 1 })
// Empty log: init-derived state at watermark -1.
const fresh = ctx.sessions.create()
expect(ctx.sessionProjections.checkpoint(fresh)['test/marks']).toEqual({ ver: 1, seq: -1, val: null })
})
it('checkpoint states are detached clones — mutating them cannot corrupt the watermark cache', async () => {
const { ctx, session } = await harness()
ctx.sessionProjections.register(marksUnit())
mark(session, ['a'])
const rows = ctx.sessionProjections.checkpoint(session)
// Hostile (or merely careless) consumer mutates the handed-out state.
;(rows['test/marks']?.val as { marks: string[] }).marks.push('INJECTED')
// The registry's authoritative cell is untouched: snapshot and a fresh
// checkpoint both still serve the committed value.
expect(ctx.sessionProjections.snapshot(session).values['test/marks']).toEqual({ marks: ['a'] })
expect(ctx.sessionProjections.checkpoint(session)['test/marks']?.val).toEqual({ marks: ['a'] })
})
it('restoreFloor anchors one below the lowest usable watermark and at 0 for missing or mismatched rows', async () => {
const { ctx } = await harness()
expect(ctx.sessionProjections.restoreFloor({})).toBeUndefined() // no unit registered
ctx.sessionProjections.register(marksUnit())
ctx.sessionProjections.register(countUnit())
expect(ctx.sessionProjections.restoreFloor({})).toBe(0)
// Lowest usable watermark is count's 5 → the anchored tail starts AT 5
// (one below the first needed seq 6), so the read proves seq 5 still exists.
expect(ctx.sessionProjections.restoreFloor({
'test/marks': { ver: 1, seq: 10, val: { marks: [] } },
'test/count': { ver: 1, seq: 5, val: 6 },
})).toBe(5)
// A version-mismatched row forces that key back to a full refold.
expect(ctx.sessionProjections.restoreFloor({
'test/marks': { ver: 2, seq: 10, val: { marks: [] } },
'test/count': { ver: 1, seq: 5, val: 6 },
})).toBe(0)
// A fresh (-1) row still needs the whole tail from 0.
expect(ctx.sessionProjections.restoreFloor({
'test/marks': { ver: 1, seq: -1, val: null },
'test/count': { ver: 1, seq: -1, val: 0 },
})).toBe(0)
})
it('restore folds the tail past each usable row and refolds from init on version mismatch', async () => {
const { ctx } = await harness()
ctx.sessionProjections.register(marksUnit())
ctx.sessionProjections.register(countUnit())
const tail: SessionEvent[] = [
{ type: 'test/mark', seq: 3, time: 3, data: { marks: ['new'] } },
{ type: 'turn/end', seq: 4, time: 4, data: { turn: 1, reason: { kind: 'completed' } } },
]
// marks row usable (watermark 2, tail starts at 3); count row mismatched — but
// a mismatch with baseSeq > 0 cannot silently refold: it throws for a re-read.
expect(() => ctx.sessionProjections.restore({
'test/marks': { ver: 1, seq: 2, val: { marks: ['old'] } },
'test/count': { ver: 99, seq: 2, val: 3 },
}, tail, 3)).toThrow(/re-read from seq 0/)
// The full-log re-read (baseSeq 0) refolds the mismatched key from init.
const full: SessionEvent[] = [
{ type: 'turn/start', seq: 0, time: 0, data: { turn: 1, trigger: { kind: 'message', source: { kind: 'user' } } } },
{ type: 'test/mark', seq: 1, time: 1, data: { marks: ['old'] } },
{ type: 'test/mark', seq: 2, time: 2, data: { marks: ['old', '2'] } },
...tail,
]
const { snapshot, checkpoint } = ctx.sessionProjections.restore({
'test/marks': { ver: 1, seq: 2, val: { marks: ['old', '2'] } },
'test/count': { ver: 99, seq: 2, val: 3 },
}, full, 0)
expect(snapshot.asOfSeq).toBe(4)
expect(snapshot.values['test/marks']).toEqual({ marks: ['new'] })
expect(snapshot.values['test/count']).toBe(5) // refolded from init over all 5 events
// The refreshed rows sit at the served cut, ready for a durable write-back.
expect(checkpoint['test/marks']).toEqual({ ver: 1, seq: 4, val: { marks: ['new'] } })
expect(checkpoint['test/count']).toEqual({ ver: 1, seq: 4, val: 5 })
})
it('restore over a suffix folds only past each row watermark and serves an exact empty-tail cut', async () => {
const { ctx } = await harness()
ctx.sessionProjections.register(marksUnit())
ctx.sessionProjections.register(countUnit())
const rows = {
'test/marks': { ver: 1, seq: 4, val: { marks: ['done'] } },
'test/count': { ver: 1, seq: 2, val: 3 },
}
const tail: SessionEvent[] = [
{ type: 'turn/start', seq: 3, time: 3, data: { turn: 2, trigger: { kind: 'message', source: { kind: 'user' } } } },
{ type: 'turn/end', seq: 4, time: 4, data: { turn: 2, reason: { kind: 'completed' } } },
]
const { snapshot } = ctx.sessionProjections.restore(rows, tail, 3)
expect(snapshot.asOfSeq).toBe(4)
// marks already covers the tail (watermark 4): nothing re-applied.
expect(snapshot.values['test/marks']).toEqual({ marks: ['done'] })
// count folds exactly seqs 3 and 4 on top of its checkpoint.
expect(snapshot.values['test/count']).toBe(5)
// Empty tail (checkpoint is current): the cut sits at baseSeq - 1.
const { snapshot: current } = ctx.sessionProjections.restore({
'test/marks': { ver: 1, seq: 4, val: { marks: ['done'] } },
'test/count': { ver: 1, seq: 4, val: 5 },
}, [], 5)
expect(current.asOfSeq).toBe(4)
expect(current.values['test/count']).toBe(5)
})
it('viewCheckpoint serves version-matching rows without any log and skips mismatched keys', async () => {
const { ctx } = await harness()
ctx.sessionProjections.register(marksUnit())
ctx.sessionProjections.register(countUnit())
const values = ctx.sessionProjections.viewCheckpoint({
'test/marks': { ver: 1, seq: 4, val: { marks: ['stored'] } },
'test/count': { ver: 99, seq: 4, val: 5 }, // mismatched: absent
})
expect(values['test/marks']).toEqual({ marks: ['stored'] })
expect('test/count' in values).toBe(false)
expect(ctx.sessionProjections.viewCheckpoint({})).toEqual({})
})
it('restore rejects a row claiming events past the supplied log end (shrunk log ⇒ re-read)', async () => {
const { ctx } = await harness()
ctx.sessionProjections.register(countUnit())
const rows = { 'test/count': { ver: 1, seq: 9, val: 10 } }
// The anchored floor sits ON the watermark, so the tail read must return
// at least seq 9 from an intact log…
const floor = ctx.sessionProjections.restoreFloor(rows)
expect(floor).toBe(9)
// …an intact log serves the anchor event and the checkpoint stands as-is.
const anchor: SessionEvent = { type: 'turn/end', seq: 9, time: 9, data: { turn: 2, reason: { kind: 'completed' } } }
expect(ctx.sessionProjections.restore(rows, [anchor], 9).snapshot.values['test/count']).toBe(10)
// …while a log crash-repaired down to fewer events returns an empty tail:
// the row overreaches the proven end and a tail read cannot fix this key.
expect(() => ctx.sessionProjections.restore(rows, [], 9)).toThrow(/re-read from seq 0/)
// The full re-read discards the overreaching row and refolds from init.
const events: SessionEvent[] = [
{ type: 'turn/start', seq: 0, time: 0, data: { turn: 1, trigger: { kind: 'message', source: { kind: 'user' } } } },
{ type: 'turn/end', seq: 1, time: 1, data: { turn: 1, reason: { kind: 'completed' } } },
]
const { snapshot } = ctx.sessionProjections.restore(rows, events, 0)
expect(snapshot.asOfSeq).toBe(1)
expect(snapshot.values['test/count']).toBe(2)
})
it('fails loud when a unit view violates its own schema (async unit output is unrepresentable)', async () => {
const { ctx, session } = await harness()
ctx.sessionProjections.register({

5
packages/session-query/session-query-sqlite/tests/sqlite.spec.ts
View File

@@ -149,6 +149,11 @@ class TestPersistence extends SessionPersistence {
return structuredClone(entry)
}
async readFrom(id: SessionIdType, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
const whole = await this.inspect(id, signal)
return { meta: whole.meta, events: whole.events.filter(event => event.seq >= fromSeq) }
}
async list(): Promise<SessionHeader[]> {
TestPersistence.listStarted?.()
await TestPersistence.listGate

5
packages/session-query/session-query/tests/session-query.spec.ts
View File

@@ -96,6 +96,11 @@ class TestPersistence extends SessionPersistence {
return Promise.resolve(result)
}
async readFrom(id: SessionIdType, fromSeq: number, signal?: AbortSignal): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
const whole = await this.inspect(id, signal)
return { meta: whole.meta, events: whole.events.filter(event => event.seq >= fromSeq) }
}
list(signal?: AbortSignal): Promise<SessionHeader[]> {
TestPersistence.listCalls += 1
TestPersistence.listSignals.push(signal)

5
packages/session-query/session-query/tests/tracing.spec.ts
View File

@@ -76,6 +76,11 @@ class TracePersistence extends SessionPersistence {
return Promise.resolve(structuredClone(entry))
}
async readFrom(id: SessionIdType, fromSeq: number): Promise<{ meta: SessionHeader; events: SessionEvent[] }> {
const whole = await this.inspect(id)
return { meta: whole.meta, events: whole.events.filter(event => event.seq >= fromSeq) }
}
list(): Promise<SessionHeader[]> {
TracePersistence.listCalls += 1
if (TracePersistence.listFailure !== undefined) return Promise.reject(TracePersistence.listFailure)

4
packages/support/acp-snapshot/README.i18n.yaml
View File

@@ -2,5 +2,5 @@
# side as of the last confirmed-consistent state. Both languages carry equal authority;
# after editing either side, bring the other along and re-record with:
# pnpm run verify-translation-pairing --write packages/support/acp-snapshot/README.md
README.md: 93998c20bed7a2542c23932bd659a64aec63a585
README.zh.md: a355cf0f35b5e7bec41ab0d9063c932211a7200b
README.md: 371ede587b84ba96770d4a2b1ee89b029d92dd25
README.zh.md: f596f9021ae9b8c5973efafae7f7d695293b96e1

8
packages/support/acp-snapshot/README.md
View File

@@ -9,7 +9,7 @@ Four layers, importable separately:
- **`launchAcpTestAgent` (launcher)** — boots a source agent under tsx or a built `lib` agent under plain Node from a supplied cwd, connects the SDK client over a raw-byte stdout tee, collects session updates and stderr, surfaces asynchronous spawn failures through startup, fails closed on unhandled permission requests, and owns graceful or signalled shutdown. Shutdown waits for process exit, inherited stdio closure, and ACP parser exhaustion before resolving or propagating a child error, so captures are complete and callers can remove owned paths after either outcome. When Windows accepts forced termination but publishes its exit marker asynchronously, shutdown gives that marker a bounded grace before treating fallback refusal as a second failure. Snapshot and ordinary e2e suites share this process boundary; a test supplies only agent paths, cwd, environment overrides, and any permission policy.
- **`runScenario` (harness)** — drives ACP JSON-RPC stdio from a deterministic `input.json` script through the launcher, tees raw stdout for the expected-output and purity checks, and harvests every persisted raw JSONL session log (parent and subagent children, primary-first) after graceful stdin EOF. `AgentUnderTest` supplies absolute `binScript`, optional `libBinScript`, `configPath`, and `tsconfigPath` paths because the subprocess cwd is outside the repo; `workspaceParent` may move the generated child cwd from the platform temp directory when that grant is itself under test. Startup failures preserve captured agent stderr in the rejected diagnostic.
- **Normalizers** — pure functions turning the two captured surfaces into stable text: `normalizeStdout` (JSON-RPC ids → first-seen sequence; UUIDs and every native/JavaScript filesystem spelling of the generated cwd → tokens, longest-first; cwd-rooted separators selected as canonical `/` or host-native; doubles as the stdout-purity check), `normalizeSessionLog` (times zeroed, `seq` kept, the same cwd-path policy), `scrubSystemPrompts` (prompt text → `{{system}}`), `scrubToolSchemas` (schema bulk → `{{tools}}`), and `scrubRequestHeaders` (all header bulk → `{{system}}`/`{{tools}}`/`{{messagePrefix}}` outside each pin, structure kept — [pinned-header Agent Note](../../../.agents/notes/archived/testing/2026-07-06-pin-request-header-content-in-one-scenario.md)).
- **`defineAcpSnapshotSuite` (factory)** — registers the whole describe/it tree for a scenario table: per-scenario expected-output and re-persisted-log comparisons, record/refresh fixture write-back, rejection of structured `UNKNOWN_TOOL` results, the per-header-class pin (`system-prompt.expected.md` plus `tool-schemas.expected.json`) with its live uniformity guard, and the fixture guard block (no orphan scenario dirs, required files present, exactly one pin per class, every JSONL prompt/schema-scrubbed, non-pinning fixtures fully header-scrubbed). Refresh evaluates fresh leaves with the harvested run's ids, cwd, and every cwd alias, then reuses normalized-equivalent leaves only when the complete logical-record layout aligns and volatile string replacements form a bijection; ambiguous logs keep fresh strings, and fresh semantic values remain authoritative. It also expands packed timing envelopes before aligning event times, so switching between packed and unpacked layouts cannot shift later records. A newly inserted `session/title` receives its preceding event's time so feature-driven insertions do not churn the remainder of a fixture. Each scenario directory's `session.jsonl` plus contiguous `session.<n>.jsonl` siblings are the ordered primary/child inventory; the scenario table does not duplicate their count. Must be called at vitest collection time.
- **`defineAcpSnapshotSuite` (factory)** — registers the whole describe/it tree for a scenario table: per-scenario expected-output and re-persisted-log comparisons, record/refresh fixture write-back, rejection of structured `UNKNOWN_TOOL` results, a tokenized pin per header class composed with independently shared `system-prompt.expected.md` and `tool-schemas.expected.json` sidecars, and a live uniformity guard. Its fixture guards reject orphan scenario dirs, missing files, multiple pins for one class, duplicate sidecar content, unscrubbed JSONL headers, and malformed pinning headers. Refresh evaluates fresh leaves with the harvested run's ids, cwd, and every cwd alias, then reuses normalized-equivalent leaves only when the complete logical-record layout aligns and volatile string replacements form a bijection; ambiguous logs keep fresh strings, and fresh semantic values remain authoritative. It also expands packed timing envelopes before aligning event times, so switching between packed and unpacked layouts cannot shift later records. A newly inserted `session/title` receives its preceding event's time so feature-driven insertions do not churn the remainder of a fixture. Each scenario directory's `session.jsonl` plus contiguous `session.<n>.jsonl` siblings are the ordered primary/child inventory; the scenario table does not duplicate their count. Must be called at vitest collection time.
Committed session fixtures use canonical packed rows. An in-flight branch that merges this contract runs the [temporary repository migrator](../../../scripts/migrate-packed-session-fixtures.ts) with `pnpm run migrate:packed-session-fixtures`; its [removal proposal](../../../.agents/notes/proposed/process/2026-07-26-remove-packed-session-fixture-migrator.md) owns deletion after affected branches converge.
@@ -51,11 +51,13 @@ defineAcpSnapshotSuite({
})
```
A scenario booting a differently-composed tree sets its own `configPath` (an overlay whose basename still ends in `cordis.yml`, so the bin's replay swap finds the sibling `*cordis.snapshot.yml`) and, when that composition changes the request header, its own `headerClass` with its own pinning scenario — the acp-agent example's Code Mode and filesystem scenarios are templates. `workspaceParent` moves the generated cwd outside the platform temp area when temporary-directory grants are themselves under test; the harness still owns and removes only the generated child. A scenario's committed `workspace/` is copied into that child first, then `prepareWorkspace` runs against the generated cwd before the agent starts. Reserve this hook for fixtures Git cannot represent portably, keep ordinary seeds in `workspace/`, and pair it with `posixOnly` when the generated paths are invalid on Windows. Each pinning directory stores the normalized full prompt sequence in generated `system-prompt.expected.md` and the corresponding full tool-schema sequence in generated `tool-schemas.expected.json`; `session.jsonl` stores `"system":"{{system}}","tools":"{{tools}}"` while retaining config, reason, and any model-visible prefix. A pin with legitimate mid-run header changes declares `expectedHeaderChanges`, which fixes the length of both sidecar sequences.
A scenario booting a differently-composed tree sets its own `configPath` (an overlay whose basename still ends in `cordis.yml`, so the bin's replay swap finds the sibling `*cordis.snapshot.yml`) and, when that composition changes the request header, its own `headerClass` with its own pinning scenario — the acp-agent example's Code Mode and filesystem scenarios are templates. `workspaceParent` moves the generated cwd outside the platform temp area when temporary-directory grants are themselves under test; the harness still owns and removes only the generated child. A scenario's committed `workspace/` is copied into that child first, then `prepareWorkspace` runs against the generated cwd before the agent starts. Reserve this hook for fixtures Git cannot represent portably, keep ordinary seeds in `workspace/`, and pair it with `posixOnly` when the generated paths are invalid on Windows.
A pin owns its generated `system-prompt.expected.md` or `tool-schemas.expected.json` by default; `systemPromptSource` and `toolSchemasSource` name another pin when the complete corresponding sequence is identical, so each distinct version is committed once. The pin's `session.jsonl` stores `"system":"{{system}}","tools":"{{tools}}"` while retaining config, reason, and any model-visible prefix. A pin with legitimate mid-run header changes declares `expectedHeaderChanges`; a shared source must declare the same count, and record/refresh rejects claimants that generate different bytes.
Every scenario compares `stdout.expected.jsonl` with cwd-rooted separators canonicalized to `/`. On Windows, `pinsNativeWindowsStdout` additionally compares the complete `stdout.expected.windows.jsonl` after the shared expected output and requires that sidecar exactly when enabled. A scenario whose driven behavior needs POSIX process semantics (e.g. cancelling a live bash call kills a detached process group) declares `posixOnly`, which skips its run test on Windows while the fixture guards keep covering its committed files everywhere.
The example also ships a `cordis.snapshot.yml` replay overlay next to its `cordis.yml` (the bin swaps them under `DSH_SNAPSHOT=replay` — [single-source replay config Agent Note](../../../.agents/notes/archived/testing/2026-07-04-single-source-acp-replay-config.md)); replay fixtures are served by [`dsh-llm-replay`](../llm-replay/README.md), which this package points at via the `DSH_SNAPSHOT_*` env vars it sets on the child. `pnpm run test:snapshot:record` calls the live LLM and rewrites the recorded scenarios' model fixtures; `pnpm run test:snapshot:refresh` stays keyless, runs the replay overlay, and rewrites stdout, comparable session-log expected outputs, and each pin's prompt and tool-schema sidecars from the committed model scripts. Fixture roles, record/replay/refresh semantics, and scenario-table fields are documented on `Scenario` and in the [snapshot Agent Note](../../../.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md).
The example also ships a `cordis.snapshot.yml` replay overlay next to its `cordis.yml` (the bin swaps them under `DSH_SNAPSHOT=replay` — [single-source replay config Agent Note](../../../.agents/notes/archived/testing/2026-07-04-single-source-acp-replay-config.md)); replay fixtures are served by [`dsh-llm-replay`](../llm-replay/README.md), which this package points at via the `DSH_SNAPSHOT_*` env vars it sets on the child. `pnpm run test:snapshot:record` calls the live LLM and rewrites the recorded scenarios' model fixtures; `pnpm run test:snapshot:refresh` stays keyless, runs the replay overlay, and rewrites stdout, comparable session-log expected outputs, and owned prompt and tool-schema sidecars from the committed model scripts. Fixture roles, record/replay/refresh semantics, and scenario-table fields are documented on `Scenario` and in the [snapshot Agent Note](../../../.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md).
Constraints: `suite.ts` and `harness.ts` import vitest (the harness polls its durable-boundary waits through `vi.waitFor`), so the package entry is importable only inside a vitest run (the launcher and normalizers have no such dependency but ship from the same entry). The launcher and suite factory are ACP-specific by design — the launcher speaks the SDK's `ClientSideConnection` — while the normalizers are transport-neutral session-log/text helpers also consumed by the TUI snapshot suite and the web browser e2e lane. Input scripts cover initialization, fresh-session creation, text prompting, cancellation, expected RPC failures, and durable turn-boundary waits. Permission round-trips are a FIFO queue of option-kind selections (`allow_once`, `reject_once`, …) mapped to the agent-issued `optionId`; an absent or exhausted queue answers `cancelled`, and an unoffered kind rejects the run.

8
packages/support/acp-snapshot/README.zh.md
View File

@@ -9,7 +9,7 @@ ACP 快照套件工具包：无密钥快照层（`pnpm run test:snapshot`，见[
- **`launchAcpTestAgent`（启动器）**：从指定 cwd 在 tsx 下启动源 agent，或在普通 Node 下启动已构建 `lib` agent；通过原始字节 stdout tee 连接 SDK 客户端，收集会话更新和 stderr，在启动过程中公开异步 spawn 失败，对未处理权限请求快速失败，并负责优雅或带信号关闭。关闭会等待进程退出、继承 stdio 关闭和 ACP parser 耗尽，然后才解析或传播子级错误，使捕获内容完整，且调用方可在任一结果后移除自有路径。当 Windows 接受强制终止但异步发布退出标记时，关闭会给该标记有界宽限，然后才将回退拒绝视为第二次失败。快照和普通 e2e 套件共享该进程边界；测试只需提供 agent 路径、cwd、环境覆盖和任何权限策略。
- **`runScenario`（harness）**：通过启动器从确定性 `input.json` 脚本驱动 ACP JSON-RPC stdio，将原始 stdout tee 给预期输出和纯度检查，并在优雅 stdin EOF 后收集每个持久化原始 JSONL 会话日志（父级和 subagent 子级，主级优先）。`AgentUnderTest` 提供绝对 `binScript`、可选 `libBinScript`、`configPath` 和 `tsconfigPath` 路径，因为子进程 cwd 位于仓库外。当生成子级 cwd 自身位于待测授权中时，`workspaceParent` 可以将它从平台临时目录移出。启动失败会在拒绝诊断中保留已捕获 agent stderr。
- **规范化器**：将两个已捕获接口转换为稳定文本的纯函数：`normalizeStdout`（JSON-RPC id → 首次出现序列；UUID 以及生成 cwd 的每个原生/JavaScript 文件系统写法 → token，按最长优先；根据 cwd 的分隔符选择规范 `/` 或宿主原生形式；同时作为 stdout 纯度检查）、`normalizeSessionLog`（时间归零、保留 `seq`、使用同一 cwd 路径策略）、`scrubSystemPrompts`（提示词文本 → `{{system}}`）、`scrubToolSchemas`（schema bulk → `{{tools}}`）和 `scrubRequestHeaders`（每个 pin 之外的所有 header bulk → `{{system}}`/`{{tools}}`/`{{messagePrefix}}`，保留结构；见[header 固定 Agent Note](../../../.agents/notes/archived/testing/2026-07-06-pin-request-header-content-in-one-scenario.md)）。
- **`defineAcpSnapshotSuite`（工厂）**：为场景表注册完整 describe/it 树：每场景预期输出与重新持久化日志比较、录制/刷新 fixture 回写、拒绝结构化 `UNKNOWN_TOOL` 结果、每 header 类别 pin（`system-prompt.expected.md` 加 `tool-schemas.expected.json`）及其实时一致性保护，以及 fixture 保护块（无遗留场景目录、必需文件存在、每类别恰好一个 pin、每个 JSONL 的提示词/schema 已擦除、非 pin fixture 的 header 已完全擦除）。刷新会使用收集所得本次运行的 id、cwd 及全部 cwd 别名评估本次生成的叶值；只有完整逻辑记录布局对齐且易变字符串替换形成双射时，才会复用规范化后等价的叶值；有歧义的日志保留本次生成的字符串，而本次生成的语义值仍为权威数据。它还会在对齐事件时间前展开打包时序 envelope，因此切换打包/非打包布局无法移动后续记录。新插入的 `session/title` 使用前一个事件的时间，因此功能驱动的插入不会扰动 fixture 余下部分。每个场景目录的 `session.jsonl` 和连续 `session.<n>.jsonl` 同级文件是有序主级/子级清单；场景表不重复其数量。必须在 vitest 收集时调用。
- **`defineAcpSnapshotSuite`（工厂）**：为场景表注册完整 describe/it 树：每场景预期输出与重新持久化日志比较、录制/刷新 fixture 回写、拒绝结构化 `UNKNOWN_TOOL` 结果、每个 header 类别一个 token 化 pin（由可独立共享的 `system-prompt.expected.md` 和 `tool-schemas.expected.json` sidecar 组合而成），以及实时一致性保护。其 fixture 保护会拒绝遗留场景目录、缺失文件、一个类别包含多个 pin、重复的 sidecar 内容、未擦除的 JSONL header，以及格式错误的 pin header。刷新会使用收集所得本次运行的 id、cwd 及全部 cwd 别名评估本次生成的叶值；只有完整逻辑记录布局对齐且易变字符串替换形成双射时，才会复用规范化后等价的叶值；有歧义的日志保留本次生成的字符串，而本次生成的语义值仍为权威数据。它还会在对齐事件时间前展开打包时序 envelope，因此切换打包/非打包布局无法移动后续记录。新插入的 `session/title` 使用前一个事件的时间，因此功能驱动的插入不会扰动 fixture 余下部分。每个场景目录的 `session.jsonl` 和连续 `session.<n>.jsonl` 同级文件是有序主级/子级清单；场景表不重复其数量。必须在 vitest 收集时调用。
签入仓库的会话 fixture 使用规范打包行。合并此契约的在途分支通过 `pnpm run migrate:packed-session-fixtures` 运行[临时仓库迁移器](../../../scripts/migrate-packed-session-fixtures.ts)；待受影响分支收敛后，由其[移除提案](../../../.agents/notes/proposed/process/2026-07-26-remove-packed-session-fixture-migrator.md)负责删除该迁移器。
@@ -51,11 +51,13 @@ defineAcpSnapshotSuite({
})
```
启动不同组合树的场景会设置自己的 `configPath`（一个 basename 仍以 `cordis.yml` 结尾的 overlay，使 bin 的回放交换可找到同级 `*cordis.snapshot.yml`）；当该组合改变请求 header 时，还会设置自己的 `headerClass` 和 pin 场景，acp-agent 示例的 Code Mode 与文件系统场景是模板。当临时目录授权自身待测时，`workspaceParent` 将生成 cwd 移出平台临时区域；harness 仍只拥有并移除生成的子级。场景签入的 `workspace/` 会先复制到该子级，随后 `prepareWorkspace` 在 agent 启动前针对生成 cwd 运行。此 hook 仅用于 Git 无法跨平台表示的 fixture；普通种子应留在 `workspace/` 中，而生成路径在 Windows 上无效时还必须搭配 `posixOnly`。每个 pin 目录将规范化的完整提示词序列存入生成的 `system-prompt.expected.md`，将对应完整工具 schema 序列存入生成的 `tool-schemas.expected.json`；`session.jsonl` 存储 `"system":"{{system}}","tools":"{{tools}}"`，同时保留配置、原因和任何模型可见前缀。具有合法运行中 header 变更的 pin 声明 `expectedHeaderChanges`，用于固定两个 sidecar 序列的长度。
启动不同组合树的场景会设置自己的 `configPath`（一个 basename 仍以 `cordis.yml` 结尾的 overlay，使 bin 的回放交换可找到同级 `*cordis.snapshot.yml`）；当该组合改变请求 header 时，还会设置自己的 `headerClass` 和 pin 场景，acp-agent 示例的 Code Mode 与文件系统场景是模板。当临时目录授权自身待测时，`workspaceParent` 将生成 cwd 移出平台临时区域；harness 仍只拥有并移除生成的子级。场景签入的 `workspace/` 会先复制到该子级，随后 `prepareWorkspace` 在 agent 启动前针对生成 cwd 运行。此 hook 仅用于 Git 无法跨平台表示的 fixture；普通种子应留在 `workspace/` 中，而生成路径在 Windows 上无效时还必须搭配 `posixOnly`。
每个 pin 默认拥有其生成的 `system-prompt.expected.md` 或 `tool-schemas.expected.json`；当完整的对应序列相同时，`systemPromptSource` 和 `toolSchemasSource` 指定另一个 pin 作为来源，因此每个不同版本只提交一次。该 pin 的 `session.jsonl` 存储 `"system":"{{system}}","tools":"{{tools}}"`，同时保留配置、原因和任何模型可见前缀。具有合法运行中 header 变更的 pin 声明 `expectedHeaderChanges`；共享来源必须声明相同的 header 变更数量，录制/刷新会拒绝生成不同字节的共享引用方。
每个场景都比较 `stdout.expected.jsonl`，其中以 cwd 为根的分隔符规范化为 `/`。在 Windows 上，`pinsNativeWindowsStdout` 还会在共享预期输出之后比较完整 `stdout.expected.windows.jsonl`，并在启用时精确要求该 sidecar。驱动行为需要 POSIX 进程语义的场景（例如取消实时 bash 调用会终止脱离进程组）声明 `posixOnly`，在 Windows 上跳过运行测试，但 fixture 保护仍在所有平台覆盖其已提交文件。
示例还发布 `cordis.snapshot.yml` 回放 overlay，位于 `cordis.yml` 旁边（bin 在 `DSH_SNAPSHOT=replay` 下交换它们，见[单源回放配置 Agent Note](../../../.agents/notes/archived/testing/2026-07-04-single-source-acp-replay-config.md)）；回放 fixture 由 [`dsh-llm-replay`](../llm-replay/README.md) 提供，该包通过对子级设置的 `DSH_SNAPSHOT_*` env var 指向它。`pnpm run test:snapshot:record` 调用实时 LLM，并重写已记录场景的模型 fixture；`pnpm run test:snapshot:refresh` 保持无密钥，运行回放 overlay，并从已提交模型脚本重写 stdout、可比较会话日志预期输出，以及每个 pin 的提示词与工具 schema sidecar。Fixture 角色、录制/回放/刷新语义和场景表字段记录在 `Scenario` 以及[快照 Agent Note](../../../.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md) 中。
示例还发布 `cordis.snapshot.yml` 回放 overlay，位于 `cordis.yml` 旁边（bin 在 `DSH_SNAPSHOT=replay` 下交换它们，见[单源回放配置 Agent Note](../../../.agents/notes/archived/testing/2026-07-04-single-source-acp-replay-config.md)）；回放 fixture 由 [`dsh-llm-replay`](../llm-replay/README.md) 提供，该包通过对子级设置的 `DSH_SNAPSHOT_*` env var 指向它。`pnpm run test:snapshot:record` 调用实时 LLM，并重写已记录场景的模型 fixture；`pnpm run test:snapshot:refresh` 保持无密钥，运行回放 overlay，并从已提交模型脚本重写 stdout、可比较会话日志预期输出，以及各 pin 自有的提示词与工具 schema sidecar。Fixture 角色、录制/回放/刷新语义和场景表字段记录在 `Scenario` 以及[快照 Agent Note](../../../.agents/notes/implemented/testing/2026-06-19-acp-snapshot-tests.md) 中。
约束：`suite.ts` 与 `harness.ts` 导入 vitest（harness 通过 `vi.waitFor` 轮询其持久边界等待），因此包入口只能在 vitest 运行中导入（启动器和规范化器没有此依赖，但从同一入口发布）。启动器和套件工厂按设计专用于 ACP，启动器使用 SDK 的 `ClientSideConnection`；规范化器是与传输无关的会话日志/文本辅助工具，还由 TUI 快照套件和 web 浏览器 e2e lane 消费。输入脚本覆盖初始化、新建会话、文本提示、取消、预期 RPC 失败和持久轮次边界等待。权限往返是选项类别选择（`allow_once`、`reject_once`等）的 FIFO 队列，映射到 agent 发出的 `optionId`；缺少或耗尽的队列回答 `cancelled`，未提供类别会拒绝运行。

Some files were not shown because too many files have changed in this diff Show More