Merge remote-tracking branch 'origin/master' into feat/todo-multi-in-progress
# Conflicts: # packages/client/ui-conversation/README.i18n.yaml
This commit is contained in:
Chinesezjc
@@ -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/2026-07-24-web-session-model-selector.md
|
||||
2026-07-24-web-session-model-selector.md: 4bd52bba8d3ba16c01fc59d2b561516f6fd0bb87
|
||||
2026-07-24-web-session-model-selector.zh.md: 35f3525b80e7d27fb573e854aefc317ed354df53
|
||||
@@ -0,0 +1,43 @@
|
||||
# Agent Note: Session model selection in the Web composer
|
||||
|
||||
Status: implemented
|
||||
|
||||
English | [中文](2026-07-24-web-session-model-selector.zh.md)
|
||||
|
||||
## Problem
|
||||
|
||||
The Web conversation displayed and sent through the Host's fixed provider/model route without exposing that route or letting a user change it. The TUI already had a session-local route target, but copying its presentation or hardcoding DeepSeek models in the browser would split model discovery and step-boundary semantics across front doors. A switch made while a response is running also needs one atomic boundary: prompt variables and request routing cannot observe different targets.
|
||||
|
||||
## Decision
|
||||
|
||||
The Web Host reuses `installAgentLlmTarget` for every created or resumed agent. The provider/model/reasoning target starts from the latest `request/header` when the session has used a model, otherwise from the Host default route. `session.selectModel` changes the session-local mutable target, and prompt assembly captures it with request routing; a switch during a running step therefore applies to the next assembled step. The next consumed target persists through the existing full `request/header` snapshot, while a choice that has not reached a request remains process-local.
|
||||
|
||||
The session RPC domain exposes a `session.models` directory and `session.selectModel`. The directory is built dynamically from the LLM registry and grouped by provider; each listed model's exact metadata adds adapter-owned reasoning effort ids, names, descriptions, and optional default. Provider catalogs and exact metadata load concurrently by provider and fail independently, so successful groups remain usable alongside retryable failure records. Catalog membership stays advisory: the current model is inserted as an unlisted row when its registered provider omits it, while exact resolution decides whether a route and explicit effort are available. Selection uses `resolveCallConfig` to reject unsupported effort ids and materialize an adapter-configured default before updating the target.
|
||||
|
||||
The browser `ModelService` owns one `ModelDirectory` per live session. Its snapshot contains the current complete target, grouped catalog, provider failures, operation error, and `idle`/`loading`/`ready`/`selecting`/`error` state. Mounting primes the trigger label and each menu open refreshes the directory. Directory and selection calls share an operation generation so older responses cannot replace a newer result; connection reset discards the process-local projection before restoring the Host target. Failures retain the previous current target and usable groups.
|
||||
|
||||
`@deepseek-ai/dsh-client-ui-conversation` declares the session-scoped single slot `conversation.input.model` as a child of its composer-bar entry. InputBar renders the seat in its trailing controls immediately before the pending indicator and primary button; the seat receives the bar's `locked` owner prop and session scope. `@deepseek-ai/dsh-client-ui-model` occupies that seat and also contributes `/model` over the same directory. Its compact trigger displays the catalog model name and effective reasoning label, falling back to ids when metadata is absent. The upward menu first offers Model and, when the current exact model supports it, Effort; Model drills into provider groups, while Effort drills into the adapter-ordered levels. The provider-default row appears only when the adapter does not configure a model default.
|
||||
|
||||
The production browser roster is the flat config tree in `apps/cli/cordis.yml`; the model feature is one `dshClient` row rather than a package hardcoded in Web boot code. Its package manifest orders it after the runtime and command feature, while Cordis service injection waits for the conversation slot before registering the composer occupant.
|
||||
|
||||
## Alternatives considered
|
||||
|
||||
**Use separate provider and model dropdowns.** The model list depends on the provider and repeats a two-stage interaction for every change. One grouped menu keeps the provider visible as organization without lengthening the trigger or each row.
|
||||
|
||||
**Hardcode the current DeepSeek catalog in the Web client.** This would drift from registered adapters and exclude deployment-owned providers. The LLM registry remains the source of provider and model metadata, including partial lookup failures.
|
||||
|
||||
**Keep `High`/`Max` as client-local UI state.** Static DeepSeek labels cannot represent `off`, pi-ai provider vocabularies, adapter defaults, validation, resume, or the next provider request. Exact-model metadata owns the selectable vocabulary, and the session target owns the selected id.
|
||||
|
||||
**Make the selection a global default.** A global mutation would unexpectedly redirect other open conversations. The target belongs to one live session, while Host configuration remains the default for sessions without a logged request.
|
||||
|
||||
**Reject changes while an agent is running.** The shared atomic target already separates the assembled step from the next selection. Keeping the selector available lets the user prepare the following step without altering the in-flight request.
|
||||
|
||||
**Persist every click as a new session event.** A choice is not model-visible until prompt assembly consumes it. Persisting unused UI intent would add a durable event that does not reconstruct a model request; the existing `request/header` records the first request that actually uses the route.
|
||||
|
||||
## Consequences
|
||||
|
||||
Any Host-backed Web conversation, including a blank session, can switch among dynamically discovered provider groups and adapter-owned reasoning levels without displaying duplicated `provider/model` labels. The current consumed target survives resume and reconnect; catalog names remain presentation-only, while selection and persistence use provider/model/effort ids. A provider catalog or exact-metadata outage degrades only that group. Route changes can reduce provider-side cache reuse, but the selector adds no prompt content and does not disturb the in-flight step. A model without reasoning metadata has no Effort row.
|
||||
|
||||
## Testing
|
||||
|
||||
Host tests pin grouped discovery, catalog and exact-metadata failure isolation, logged effort restoration, unlisted current targets, unsupported effort rejection, default materialization, and next-assembly switching. Client tests pin the shared directory, reconnect restoration, and complete-target submission. Component tests pin dynamic effort labels, descriptions, provider-default exposure, and effort submission. The keyless built-app fixture loads the production model plugin, selects OpenAI's GPT-5 and its Max effort, sends a turn, and verifies that the next generated response reports both ids.
|
||||
@@ -0,0 +1,43 @@
|
||||
# Agent Note: Web 对话输入区的会话模型选择
|
||||
|
||||
Status: implemented
|
||||
|
||||
[English](2026-07-24-web-session-model-selector.md) | 中文
|
||||
|
||||
## 问题
|
||||
|
||||
Web 对话原本通过 Host 固定的提供方与模型路由显示并发送消息，既不呈现该路由，也不允许用户更改。TUI 已经具备会话级的路由目标，但如果照搬其呈现方式，或在浏览器中硬编码 DeepSeek 模型，就会让模型发现逻辑和步骤边界语义分散到不同前门中。响应运行期间发生的切换还需要一个原子边界：提示词变量与请求路由不能观测到不同的目标。
|
||||
|
||||
## 决策
|
||||
|
||||
Web Host 为每个新建或恢复的 agent（智能体）复用 `installAgentLlmTarget`。如果会话已经使用过模型，提供方／模型／推理（reasoning）目标从最新的 `request/header` 开始；否则采用 Host 默认路由。`session.selectModel` 会更改会话级可变目标，提示词组装则将该目标与请求路由一并捕获，因此运行中步骤发生的切换会应用于下一个组装步骤。下一条实际采用的目标通过现有的完整 `request/header` 快照持久化；尚未进入请求的选择则仅保存在当前进程中。
|
||||
|
||||
会话 RPC 领域公开 `session.models` 模型目录与 `session.selectModel`。该目录从 LLM（大语言模型）注册表动态构建，并按提供方分组；每个已列出模型的精确元数据还会加入由适配器持有的推理强度 ID、名称、说明和可选默认值。各提供方的目录与精确元数据会按提供方并发加载，且彼此独立失败，因此成功加载的分组仍可与可重试的失败记录一同使用。模型是否位于目录仅供参考：如果当前模型的已注册提供方没有列出该模型，系统会将其作为未列出行插入；精确解析则决定路由与显式推理强度是否可用。选择操作通过 `resolveCallConfig` 拒绝不支持的推理强度 ID，并在更新目标前具体化适配器配置的默认值。
|
||||
|
||||
浏览器中的 `ModelService` 为每个实时会话持有一个 `ModelDirectory`。其快照包含当前完整目标、分组目录、提供方失败记录、操作错误，以及 `idle`、`loading`、`ready`、`selecting`、`error` 状态。挂载时会预先填充触发器标签，此后每次打开菜单都会刷新目录。目录与选择调用共用操作代次，防止较早响应覆盖较新结果；连接重置会先丢弃当前进程中的投影，再恢复 Host 目标。失败时保留先前的当前目标和可用分组。
|
||||
|
||||
`@deepseek-ai/dsh-client-ui-conversation` 将会话作用域的单实例 slot `conversation.input.model` 声明为其输入栏 entry 的子 slot。InputBar 在尾部控件区将该 seat 渲染于 pending 指示器与主按钮之前；该 seat 接收输入栏的 `locked` owner prop 与会话作用域。`@deepseek-ai/dsh-client-ui-model` 占用该 seat，并在同一目录上提供 `/model`。其紧凑型触发器显示目录中的模型名称与生效的推理强度标签；元数据缺失时则回退到相应 ID。向上展开的菜单首先提供 Model，并在当前精确模型支持时提供 Effort；Model 可深入提供方分组，Effort 可深入适配器排序的级别。仅当适配器没有配置模型默认值时，才显示提供方默认值行。
|
||||
|
||||
生产环境的浏览器名册是 `apps/cli/cordis.yml` 中的平铺 config tree；模型功能对应其中一行 `dshClient` 配置项，而不是 Web boot 代码中硬编码的包。其包 manifest（元数据清单）将加载顺序置于运行时与命令功能之后；Cordis 服务注入则等待 conversation slot 可用，再注册 composer 占用方。
|
||||
|
||||
## 考虑过的替代方案
|
||||
|
||||
**分别使用提供方与模型下拉框。** 模型列表依赖提供方，每次更改都需要经过两阶段交互。单个分组菜单仍以提供方组织模型，同时不会增加触发器或各行的显示长度。
|
||||
|
||||
**在 Web 客户端中硬编码当前 DeepSeek 目录。** 该目录会与已注册适配器发生偏离，也会排除部署自有的提供方。LLM 注册表继续作为提供方与模型元数据的真源，也负责呈现部分查询失败。
|
||||
|
||||
**将 `High`／`Max` 保留为客户端本地 UI 状态。** 静态 DeepSeek 标签无法覆盖 `off`、pi-ai 的提供方词汇、适配器默认值与校验，也不能参与恢复或下一次提供方请求。精确模型元数据拥有可选词汇，会话目标则拥有已选择的 ID。
|
||||
|
||||
**将选择设为全局默认值。** 全局变更会意外改道其他已打开的对话。目标仅属于一个实时会话；对于没有已记录请求的会话，Host 配置仍是默认值。
|
||||
|
||||
**agent 运行期间拒绝更改。** 共享原子目标已经将当前组装步骤与下一次选择分离。保持选择器可用，可以让用户为下一个步骤预先选择模型，而不会改变正在执行的请求。
|
||||
|
||||
**将每次点击作为新的会话事件持久化。** 只有提示词组装采用某项选择后，该选择才对模型可见。持久化尚未使用的 UI 意图，会增加一个无法重建模型请求的持久事件；现有 `request/header` 会记录首次实际使用该路由的请求。
|
||||
|
||||
## 影响
|
||||
|
||||
任何由 Host 支撑的 Web 对话（包括空白会话）都可以在动态发现的提供方分组和适配器持有的推理级别之间切换，而无需显示重复的 `provider/model` 标签。当前实际采用的目标会在恢复和重连后保留；目录名称仅用于呈现，而选择和持久化使用提供方／模型／推理强度 ID。某个提供方的目录或精确元数据不可用时，只有相应分组会降级。路由变更可能降低提供方侧的缓存复用率，但选择器不会添加任何提示词内容，也不会干扰正在执行的步骤。没有推理元数据的模型不显示 Effort 行。
|
||||
|
||||
## 测试
|
||||
|
||||
Host 测试固定分组发现、目录与精确元数据失败隔离、已记录推理强度恢复、当前未列出目标、不支持的推理强度拒绝、默认值具体化，以及切换仅影响下一次组装。客户端测试固定共享目录、重连恢复与完整目标提交。组件测试固定动态推理强度标签、说明、提供方默认值展示与推理强度提交。无密钥 built-app fixture（测试前置数据）加载生产模型插件，选择 OpenAI 的 GPT-5 及其 Max 推理强度，发起一个轮次，并验证下一条生成的响应会报告两个 ID。
|
||||
@@ -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/2026-07-27-typescript-sdk-and-sdk-subagent-backend.md
|
||||
2026-07-27-typescript-sdk-and-sdk-subagent-backend.md: a1481bb9e2c3abfc3111dce8a1c38835436e3e13
|
||||
2026-07-27-typescript-sdk-and-sdk-subagent-backend.zh.md: 64a55f6aa0cb4b9a4a5efc625ece6a0800217f52
|
||||
@@ -0,0 +1,47 @@
|
||||
# Agent Note: TypeScript SDK client and the SDK subagent backend
|
||||
|
||||
Status: implemented
|
||||
|
||||
English | [中文](2026-07-27-typescript-sdk-and-sdk-subagent-backend.zh.md)
|
||||
|
||||
## Problem
|
||||
|
||||
The stdio JSON-RPC serving surface (`@deepseek-ai/dsh-jsonrpc`, the [single-exe Agent Note](../architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.md)) had exactly one client: the Python SDK. TypeScript consumers wanting the same drive-a-harness-as-a-subprocess capability — repo tests, automation, and above all a subagent backend whose child is a *complete harness runtime* rather than a generic ACP agent — had nothing to import: the request/notification payload shapes existed only as anonymous object literals inside the server, and the transport class lived inside the server plugin package.
|
||||
|
||||
## Decision
|
||||
|
||||
Three packages, layered exactly like the existing Python stack, plus one seam registration:
|
||||
|
||||
- **`@deepseek-ai/dsh-sdk-protocol`** (`packages/sdk/sdk-protocol/`) — the wire made shared and nominal. `JsonRpcLineTransport` moves here verbatim from `dsh-jsonrpc` (which now imports it), and `types.ts` names every payload the server speaks: `InitializeParams/Result`, `SessionPromptParams/Result`, the four notification payloads, and the `HarnessSdkRequestMap`/`HarnessSdkNotificationMap` indexes. The server's `notify()` call sites are typed against these named payloads, so server drift breaks compilation, not clients. One behavioral change: an error response now rejects with `JsonRpcResponseError` carrying the wire `code`/`data` (the Python client already preserved these; the old transport threw a bare `Error` with only the message).
|
||||
- **`@deepseek-ai/dsh-sdk-client`** (`packages/sdk/sdk-client/`) — the TypeScript twin of `python/sdk`: `HarnessClient` (spawn, frame, fan out notifications, typed error surfaces, close-to-quiescence via the shared dispose ladder) under `DeepSeekHarness`/`HarnessSession` (lazy start, memoized `initialize`, `run()` pairing one `session/prompt` with its `session.finished`). Session-tree scoping from `subagent.started` lineage edges is client-side, mirroring `client.py`. Deliberate asymmetries with Python: the launch spec is explicit `command`/`args` (no bundled-runtime resolution — that is a distribution concern with no TS consumer yet); `env` replaces rather than merges (callers own credential policy; `scrubbedParentEnv` from the subprocess seam is one import away); `TurnResult` carries the structured `reason` (Python exposes only `status`); teardown walks a private stdin-EOF → SIGTERM → SIGKILL ladder to actual exit (the client runs outside any harness context, so it cannot ride `ctx.subprocess`).
|
||||
- **`@deepseek-ai/dsh-subagent-dsh-sdk`** (`packages/subagent/subagent-dsh-sdk/`) — the second out-of-process `SubagentProvider`, structured as `subagent-acp`'s sibling: same all-false capabilities and `inheritsParentContext: false`, same publish-after-handshake ownership transaction, same result-never-rejects flattening through an `onError` sink, same parent-namespace run id. The child answer is read from streamed `session.event`s — the last complete `assistant/message`, else accumulated `text-delta` chunks, so partial answers survive cancellation. Stop reasons map from the child's structured `TurnEndReason` (`completed`/`max-tokens`/`aborted` pass through; everything else, including a settled-without-turn child, is `error`). Its `provider`/`model` config feeds the child's `initialize`; `env` is where deployments pass the child's own key and `DSH_CORDIS_CONFIG`.
|
||||
- **The subagent seam grows `out-of-process.ts`**: the provider-side vocabulary both out-of-process backends share — `NO_START_CAPABILITIES`, timing-bound validation, child cwd resolution (config override, else the delegating parent session's workspace), the never-reject `settleRunResult`, and the `subprocessRunHandle` publication. Process mechanics (spawn, env scrub, tree-scoped teardown) live in the `dsh-subprocess` seam; `subagent-acp` spawns through `ctx.subprocess`, while this backend spawns through the SDK client (the subprocess README's documented exception for SDK-managed transports) and applies the seam's `scrubbedParentEnv()` itself.
|
||||
|
||||
`dsh-jsonrpc` keeps serving unchanged (the wire is byte-identical); `dsh-jsonrpc-agent-pkg` (the Python runtime closure) gains the `dsh-sdk-protocol` dependency line.
|
||||
|
||||
## Testing
|
||||
|
||||
Four tiers, per [testing policy](../../../../docs/testing.md):
|
||||
|
||||
- **Keyless unit** — `sdk-client` drives a scripted fake runtime (`tests/fake-runtime.ts`, env-scripted, protocol-only — the Python `test_client.py` pattern) over real stdio; `subagent-dsh-sdk` drives the same fake through the real provider. 100% per-file coverage on all three packages.
|
||||
- **Keyless Loader composition** — `subagent-dsh-sdk/tests/loader-composition.e2e.ts` boots a test-only cordis.yml (`examples/jsonrpc-agent/tests/fixtures/subagent/subagent-dsh-sdk/`) where the child is a REAL second harness runtime with its own cordis.yml; asserts the parent tool result and the child's own persisted transcript both carry the parent session's cwd. The child launch resolves through `resolveExampleLaunch`, so src/lib modes both hold.
|
||||
- **Keyless snapshot** — `examples/jsonrpc-agent/tests/sdk.snapshot.ts` is the jsonrpc example's first snapshot suite: the real `dsh-jsonrpc-agent` runtime driven through the real `dsh-sdk-client`, replaying recorded fixtures via `llm-replay` behind the new `cordis.snapshot.yml` overlay (passed explicitly through `DSH_CORDIS_CONFIG`; the jsonrpc bin performs no snapshot config swap of its own). Three scenarios — text turn, bash tool, spawn subagent — each pinning the normalized notification stream, the SDK turn result, and the persisted parent+child logs. This also closes the protocol-tier gap the single-exe note's Python-side snapshot left on the vitest side.
|
||||
- **With-key e2e** — the snapshot suite's `DSH_SNAPSHOT=record` mode is the live-API path (it produced the committed fixtures); the composition e2e needs no key by design.
|
||||
|
||||
## Alternatives considered
|
||||
|
||||
**Import wire types from `dsh-jsonrpc` instead of extracting a protocol package.** Makes every SDK consumer (including `subagent-dsh-sdk`, which must not serve JSON-RPC) depend on the server plugin and its `dsh-agent`/`dsh-llm-deepseek` peer set, and leaves the notification payloads anonymous. The capability-seam rule (interface/implementation/consumer as separate packages) already names this shape; the transport is genuinely two-sided.
|
||||
|
||||
**Have `subagent-dsh-sdk` speak raw JSON-RPC without the client SDK.** Duplicates the request/notification pairing, subscription fan-out, timeout, and teardown logic the SDK exists to own; the user's ask was explicitly a backend that *uses* the SDK, and the layering earns its keep by making the backend ~200 lines of policy over a reusable client.
|
||||
|
||||
**Fold the SDK backend into `subagent-acp` with a transport switch.** The two backends share the subprocess lifecycle but nothing about the wire (ACP SDK connection vs harness JSON-RPC), the child contract (any ACP agent vs a harness runtime), or the result extraction (`agent_message_chunk` accumulation vs session-event reading). A config discriminant would bury two protocols in one package; the genuinely shared provider-side parts moved into the subagent seam's `out-of-process.ts`, and the process mechanics live in the `dsh-subprocess` seam.
|
||||
|
||||
**Give the TS SDK bundled-runtime resolution parity with Python.** Python's carrier resolution exists to ship wheels to users without Node. A TypeScript consumer definitionally has Node and (in-repo) the workspace; inventing a distribution story with no consumer violates the require-current-need rule. Deferred until a real npm-distribution consumer appears.
|
||||
|
||||
**Reuse `dsh-acp-snapshot`'s `runScenario` for the SDK snapshots.** That harness speaks ACP (`ClientSideConnection`, `InputStep` scripts). The SDK suite's whole point is to drive the *SDK client* as the entry surface; it reuses the normalize/refresh library layer (`normalizeSessionLog`, `refreshFixtureReplacements`, …) and leaves the ACP driver alone.
|
||||
|
||||
## Consequences
|
||||
|
||||
**Bought**: the SDK runtime protocol now has named, compiler-checked types shared by its server and both client SDKs; TypeScript consumers get the same subprocess-driving capability Python has, with typed errors and structured turn reasons; the subagent seam gains a harness-native out-of-process backend whose children are full peers (own config, persistence, tools) — the recursive-composition story the seam note anticipated; the jsonrpc example finally has snapshot coverage, through the SDK path itself.
|
||||
|
||||
**Paid**: a third package in the `sdk/` group and a fourth subagent backend to keep current; the SDK backend boots a complete plugin tree per child (heavier per-run than an ACP child; pooling remains future work, same as ACP); the wire still has no cancel method, so both the SDK's `RequestTimeoutError` and the backend's dispose settle locally while the server-side turn runs on until process teardown; fixtures for the snapshot suite were recorded against `deepseek-v4-flash` and re-record on model-behavior drift like every other recorded corpus.
|
||||
@@ -0,0 +1,47 @@
|
||||
# Agent Note: TypeScript SDK 客户端与 SDK subagent 后端
|
||||
|
||||
Status: implemented
|
||||
|
||||
[English](2026-07-27-typescript-sdk-and-sdk-subagent-backend.md) | 中文
|
||||
|
||||
## Problem
|
||||
|
||||
stdio JSON-RPC 服务表面（`@deepseek-ai/dsh-jsonrpc`，见[单文件可执行 Agent Note](../architecture/2026-07-10-single-file-executable-sdk-runtime-distribution.md)）当时只有一个客户端：Python SDK。想要同样"把 harness 作为子进程驱动"能力的 TypeScript 消费者——仓库测试、自动化，尤其是一个其子进程是*完整 harness 运行时*（而非通用 ACP 代理）的 subagent 后端——无物可导入：请求/通知载荷形状只以匿名对象字面量存在于服务器内部，传输类也躺在服务器插件包里。
|
||||
|
||||
## Decision
|
||||
|
||||
三个包，分层与既有 Python 栈完全一致，外加一个接缝注册：
|
||||
|
||||
- **`@deepseek-ai/dsh-sdk-protocol`**（`packages/sdk/sdk-protocol/`）—— 把线协议做成共享且具名。`JsonRpcLineTransport` 从 `dsh-jsonrpc` 原样移入（后者现在导入它），`types.ts` 为服务器所说的每个载荷命名：`InitializeParams/Result`、`SessionPromptParams/Result`、四个通知载荷，以及 `HarnessSdkRequestMap`/`HarnessSdkNotificationMap` 索引。服务器的 `notify()` 调用点以这些具名载荷标注类型，服务器漂移会先破坏编译而不是破坏客户端。一处行为变化：错误响应现在以携带线上 `code`/`data` 的 `JsonRpcResponseError` 拒绝（Python 客户端本就保留这些；旧传输只抛携带消息的裸 `Error`）。
|
||||
- **`@deepseek-ai/dsh-sdk-client`**（`packages/sdk/sdk-client/`）—— `python/sdk` 的 TypeScript 孪生：`HarnessClient`（生成、分帧、通知扇出、有类型的错误表面、经共享处置阶梯关闭至静止）之上是 `DeepSeekHarness`/`HarnessSession`（惰性启动、记忆化 `initialize`、`run()` 把一个 `session/prompt` 与其 `session.finished` 配对）。基于 `subagent.started` 血缘边的会话树范围限定在客户端完成，镜像 `client.py`。与 Python 的刻意不对称：启动规格是显式 `command`/`args`（无捆绑运行时解析——那是尚无 TS 消费者的发行问题）；`env` 整体替换而非合并（凭据策略归调用方；subprocess 接缝的 `scrubbedParentEnv` 一个 import 即得）；`TurnResult` 携带结构化 `reason`（Python 只暴露 `status`）；拆除走私有的 stdin-EOF → SIGTERM → SIGKILL 阶梯直到真正退出（客户端运行在任何 harness 上下文之外，无法搭乘 `ctx.subprocess`）。
|
||||
- **`@deepseek-ai/dsh-subagent-dsh-sdk`**（`packages/subagent/subagent-dsh-sdk/`）—— 第二个进程外 `SubagentProvider`，以 `subagent-acp` 的同胞结构组织：同样的全 false 能力与 `inheritsParentContext: false`，同样的握手后发布所有权事务，同样的经 `onError` 汇把结果压平为绝不拒绝，同样的父命名空间 run id。子答案从流式 `session.event` 读取——最后一条完整 `assistant/message`，否则累积的 `text-delta` 块，部分答案在取消时得以保留。停止原因由子进程的结构化 `TurnEndReason` 映射（`completed`/`max-tokens`/`aborted` 直通；其余一切、包括未跑回合就尘埃落定的子进程，都是 `error`）。其 `provider`/`model` 配置喂给子进程的 `initialize`；`env` 是部署传入子进程自有密钥与 `DSH_CORDIS_CONFIG` 的地方。
|
||||
- **subagent 接缝增长出 `out-of-process.ts`**：两个进程外后端共享的 provider 侧词汇——`NO_START_CAPABILITIES`、时限校验、子进程 cwd 解析（配置覆盖、否则发起委托的父会话工作区）、绝不拒绝的 `settleRunResult`、以及 `subprocessRunHandle` 发布。进程机制（spawn、环境擦除、进程树拆除）属于 `dsh-subprocess` 接缝；`subagent-acp` 经 `ctx.subprocess` 生成子进程，本后端则经 SDK 客户端生成（subprocess README 记载的 SDK 托管传输例外）并自行应用接缝的 `scrubbedParentEnv()`。
|
||||
|
||||
`dsh-jsonrpc` 的服务不变（线上字节完全一致）；`dsh-jsonrpc-agent-pkg`（Python 运行时闭包）增加 `dsh-sdk-protocol` 一行依赖。
|
||||
|
||||
## Testing
|
||||
|
||||
四层，依[测试政策](../../../../docs/testing.md)：
|
||||
|
||||
- **免密钥单元** —— `sdk-client` 通过真实 stdio 驱动脚本化伪运行时（`tests/fake-runtime.ts`，环境变量脚本化、纯协议——即 Python `test_client.py` 的模式）；`subagent-dsh-sdk` 经真实 provider 驱动同一伪运行时。三个包全部 100% 逐文件覆盖。
|
||||
- **免密钥 Loader 组合** —— `subagent-dsh-sdk/tests/loader-composition.e2e.ts` 启动仅测试用 cordis.yml（`examples/jsonrpc-agent/tests/fixtures/subagent/subagent-dsh-sdk/`），其中子进程是真实的第二个 harness 运行时、带自己的 cordis.yml；断言父工具结果与子进程自己持久化的转录都携带父会话 cwd。子启动经 `resolveExampleLaunch` 解析，src/lib 两种模式都成立。
|
||||
- **免密钥快照** —— `examples/jsonrpc-agent/tests/sdk.snapshot.ts` 是 jsonrpc 示例的第一个快照套件：真实 `dsh-jsonrpc-agent` 运行时经真实 `dsh-sdk-client` 驱动，在新的 `cordis.snapshot.yml` 覆盖层后经 `llm-replay` 回放已录制夹具（经 `DSH_CORDIS_CONFIG` 显式传入；jsonrpc bin 自身不做快照配置切换）。三个场景——文本回合、bash 工具、spawn 子代理——各自钉住规范化通知流、SDK 回合结果与持久化的父+子日志。这也补上了单文件可执行 Note 的 Python 侧快照在 vitest 侧留下的协议层缺口。
|
||||
- **带密钥 e2e** —— 快照套件的 `DSH_SNAPSHOT=record` 模式即真实 API 路径（已提交夹具由它产出）；组合 e2e 设计上无需密钥。
|
||||
|
||||
## Alternatives considered
|
||||
|
||||
**从 `dsh-jsonrpc` 导入线类型而不是提取协议包。** 会让每个 SDK 消费者（包括绝不能提供 JSON-RPC 服务的 `subagent-dsh-sdk`）依赖服务器插件及其 `dsh-agent`/`dsh-llm-deepseek` peer 集合，且通知载荷仍然匿名。能力接缝规则（接口/实现/消费者三包分立）已经点名了这种形态；这个传输是货真价实的双边物。
|
||||
|
||||
**让 `subagent-dsh-sdk` 直说裸 JSON-RPC、绕开客户端 SDK。** 会复制 SDK 存在意义所在的请求/通知配对、订阅扇出、超时与拆除逻辑；用户的要求明确是一个*使用* SDK 的后端，分层的回报是后端成为可复用客户端之上约 200 行的纯策略。
|
||||
|
||||
**把 SDK 后端折进 `subagent-acp`、用传输开关区分。** 两个后端共享子进程生命周期，但线协议（ACP SDK 连接 vs harness JSON-RPC）、子进程契约（任意 ACP 代理 vs harness 运行时）、结果提取（`agent_message_chunk` 累积 vs 会话事件读取）毫无共享。配置判别子会把两个协议埋进一个包；真正共享的 provider 侧部分移入 subagent 接缝的 `out-of-process.ts`，进程机制则住在 `dsh-subprocess` 接缝。
|
||||
|
||||
**给 TS SDK 与 Python 对等的捆绑运行时解析。** Python 的载体解析是为了给没有 Node 的用户发 wheel。TypeScript 消费者定义上就有 Node 且（仓库内）有工作区；为不存在的消费者发明发行故事违反"要求当前需求"规则。推迟到真实 npm 发行消费者出现。
|
||||
|
||||
**复用 `dsh-acp-snapshot` 的 `runScenario` 做 SDK 快照。** 那个 harness 说 ACP（`ClientSideConnection`、`InputStep` 脚本）。SDK 套件的全部意义就是以 *SDK 客户端*为入口表面；它复用 normalize/refresh 库层（`normalizeSessionLog`、`refreshFixtureReplacements`……），不动 ACP 驱动器。
|
||||
|
||||
## Consequences
|
||||
|
||||
**买到**：SDK 运行时协议现在拥有服务器与两个客户端 SDK 共享的、编译器校验的具名类型；TypeScript 消费者获得与 Python 相同的子进程驱动能力，且带类型化错误与结构化回合原因；subagent 接缝获得一个 harness 原生的进程外后端，其子进程是完整对等体（自有配置、持久化、工具）——正是接缝 Note 预期的递归组合故事；jsonrpc 示例终于有了快照覆盖，而且走的就是 SDK 路径本身。
|
||||
|
||||
**付出**：`sdk/` 组多了第三个包、subagent 多了第四个要保持最新的后端；SDK 后端每个子进程启动完整插件树（单次成本高于 ACP 子进程；池化与 ACP 一样留作未来工作）；线上仍无取消方法，SDK 的 `RequestTimeoutError` 与后端的 dispose 都只在本地定格、服务器侧回合继续跑到进程拆除为止；快照夹具录制于 `deepseek-v4-flash`，与其他录制语料一样随模型行为漂移而重录。
|
||||
Reference in New Issue
Block a user