fix(goal): simplify blockers into one durable phase

This commit is contained in:
Tianyi Cui
2026-07-20 16:39:40 +08:00
parent 6e876e497a
commit 1468773dcd
18 changed files with 219 additions and 286 deletions

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
2026-07-19-persisted-same-session-goal-domain.md: 1484a919851c18ce978c93c068f6096bbf3b733f
2026-07-19-persisted-same-session-goal-domain.zh.md: 8bcedc6cab6b9219a33a655e8fcbc53762b7413f
2026-07-19-persisted-same-session-goal-domain.md: 75355aa8d94789e0cc227d393c50139708a73f6a
2026-07-19-persisted-same-session-goal-domain.zh.md: fd7e31f4b6a5acec1d2b44fb3088e301b36abd5c

View File

@@ -12,13 +12,13 @@ Durable lifecycle and permission to continue are different facts. A session may
## Decision
`@deepseek-ai/dsh-goal` in `packages/goal/goal/` owns one current same-session goal through `ctx.goals`. A goal has a branded id, objective, durable phase, compare-and-set revision, and `maxGoalRounds`. `defaultMaxGoalRounds` is a validated deployment setting with default `256`; `resolveCreate()` materializes it before mutation.
`@deepseek-ai/dsh-goal` in `packages/goal/goal/` owns one current same-session goal through `ctx.goals`. A goal has a branded id, objective, durable phase, compare-and-set revision, and `maxGoalRounds`. `defaultMaxGoalRounds` is a validated deployment setting with default `256`; `create()` materializes it internally before mutation rather than exposing resolution as another service verb.
The durable phases are `active`, `paused`, `blocked`, `usage-limited`, `budget-limited`, and `complete`. A separate live activation is `armed` or `disarmed`. Creation and explicit resume arm activation; pause, completion, blocking, limit transitions, and clear disarm it. Edits preserve activation. Activation is never part of the persisted snapshot.
The durable phases are `active`, `paused`, `blocked`, and `complete`. A blocked snapshot includes a policy-owned lower-kebab-case code and a normalized free-form message, so usage limits, round caps, execution failures, and human-input dependencies share one lifecycle state without losing their cause. A separate live activation is `armed` or `disarmed`. Creation and explicit resume arm activation; pause, completion, blocking, and clear disarm it. Edits preserve activation and any blocker reason; resume and completion clear that reason. Activation is never part of the persisted snapshot.
### Durable record and replay
Every non-clear mutation uses `Agent.inject()` to append a raw, model-visible `context/message` containing a versioned full snapshot. Clear appends a revisioned tombstone. The context source is `{ kind: 'goal', goalId, revision, round: 0 }`; metadata and rendered `<goal_state>...</goal_state>` content must agree exactly. The session log is the only durable source of truth, so persistence and fork inherit goal records without another database or header field.
Every non-clear mutation uses `Agent.inject()` to append a raw, model-visible `context/message` containing a versioned full snapshot. Clear appends a revisioned tombstone. The context source is `{ kind: 'goal', goalId, revision, round: 0 }`; metadata and rendered `<goal_state>...</goal_state>` content must agree exactly. This descriptive delimiter follows the repository's existing `<workspace_context>` convention and [Anthropic's published guidance to structure mixed prompt content with consistent descriptive XML tags](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices#structure-prompts-with-xml-tags). That is public model-experience prior art, not evidence about any provider's proprietary training corpus. The session log is the only durable source of truth, so persistence and fork inherit goal records without another database or header field.
The replay fold validates JSON shape, source attribution, rendered content, fresh ids, revision continuity, lifecycle transitions, counters, and monotonic per-goal timestamps. Goal rounds are positive sequential `user/message` source numbers for the current active revision and cannot exceed `maxGoalRounds`; ordinary session turns do not affect the counter. A malformed current-format record fails replay rather than being ignored or repaired.
@@ -26,7 +26,7 @@ When `Agent.inject()` defers a mutation inside an active tool batch, the service
### Lifecycle and live activation
At most one goal is current. Create requires no current non-complete goal and always generates a revision-one id not used earlier in the session; a completed goal may be replaced. Every other mutation carries the expected `GoalRef`, and stale ids or revisions reject. Resume accepts a stopped phase or a disarmed active goal only when the round cap has remaining capacity; budget limiting requires the admitted count to have reached the cap.
At most one goal is current. Create requires no current non-complete goal and always generates a revision-one id not used earlier in the session; a completed goal may be replaced. Every other mutation carries the expected `GoalRef`, and stale ids or revisions reject. Resume accepts a paused or blocked phase, or a disarmed active goal, only when the round cap has remaining capacity. The domain validates blocker reason shape but deliberately leaves reason codes and the decision to block to policy consumers.
A cache built from any seed starts disarmed, and every `agent/session-start` edge disarms it again. Resume and fork therefore preserve the durable objective and history but never initiate work on their own. A later human prompt can be interpreted by the model, whose policy surface may explicitly call resume and arm the goal.
@@ -36,7 +36,7 @@ The service accepts only the exact live `Agent` object registered under its id.
## Testing
Unit coverage pins creation defaults, exact-live-agent checks, compare-and-set rejection, every lifecycle transition, cap enforcement, clear/replacement, seeded replay and `SessionStore.fork()` inheritance, session-start disarming and active-goal rearming, FIFO deferred mutation reconciliation, reentrant append observation, rejected-injection rollback, stable corrupt-event replay, service/listener disposal, listener containment, backward-clock clamping, strict record decoding, lifecycle continuity, source/content agreement, and sequential round attribution. A keyless Loader/stdio process test mounts the service and a lifecycle consumer through test-only `cordis.yml`, then reads the persisted JSONL externally to verify the model-visible snapshot and absence of an unrequested goal round. The package source is held to the repository's per-file 100% coverage gate.
Unit coverage pins creation defaults, exact-live-agent checks, compare-and-set rejection, every lifecycle transition, blocker reason validation and retention, cap enforcement on resume, clear/replacement, seeded replay and `SessionStore.fork()` inheritance, session-start disarming and active-goal rearming, FIFO deferred mutation reconciliation, reentrant append observation, rejected-injection rollback, stable corrupt-event replay, service/listener disposal, listener containment, backward-clock clamping, strict record decoding, lifecycle continuity, source/content agreement, and sequential round attribution. A keyless Loader/stdio process test mounts the service and a lifecycle consumer through test-only `cordis.yml`, then reads the persisted JSONL externally to verify the model-visible snapshot and absence of an unrequested goal round. The package source is held to the repository's per-file 100% coverage gate.
## Alternatives considered
@@ -52,7 +52,7 @@ Unit coverage pins creation defaults, exact-live-agent checks, compare-and-set r
- Resume and fork expose the same durable phase while remaining operationally inert until an explicit resume mutation arms activation.
- Full snapshots simplify inspection and strict replay but repeat the objective and state fields in model history until compaction shadows them.
- Revision and lifecycle validation reject tampered, partially written, or producer-inconsistent goal records early.
- Round caps bound continuation count only; token, currency, time, and provider limits remain separate policy concerns.
- Round caps bound continuation count only; policy consumers map round, token, currency, time, and provider limits to blocked reasons when they stop work.
## Known limitations and deferred work

View File

@@ -12,13 +12,13 @@ Status: implemented
## 决策
位于 `packages/goal/goal/``@deepseek-ai/dsh-goal` 通过 `ctx.goals` 管理一个当前的同会话目标。目标包含品牌化 id、目标描述、持久阶段、比较并交换修订号和 `maxGoalRounds``defaultMaxGoalRounds` 是经过校验的部署配置,默认值为 `256``resolveCreate()` 在变更前将其解析为完整值。
位于 `packages/goal/goal/``@deepseek-ai/dsh-goal` 通过 `ctx.goals` 管理一个当前的同会话目标。目标包含品牌化 id、目标描述、持久阶段、比较并交换修订号和 `maxGoalRounds``defaultMaxGoalRounds` 是经过校验的部署配置,默认值为 `256``create()` 在变更前于内部将其解析为完整值,而不会把解析过程暴露为额外的服务动词
持久阶段包括 `active``paused``blocked``usage-limited``budget-limited``complete`。独立的实时激活态为 `armed``disarmed`。创建与显式恢复会激活目标;暂停、完成、阻塞、达到限制和清除都会解除激活。编辑保留激活态。持久快照绝不包含激活态。
持久阶段包括 `active``paused``blocked``complete`。阻塞快照包含由策略提供的 kebab-case 小写代码和规范化自由文本消息,因此用量限制、回合上限、执行失败和等待人工输入可以共享一个生命周期状态而不丢失原因。独立的实时激活态为 `armed``disarmed`。创建与显式恢复会激活目标;暂停、完成、阻塞和清除都会解除激活。编辑保留激活态及阻塞原因;恢复和完成会清除该原因。持久快照绝不包含激活态。
### 持久记录与回放
每次非清除变更都通过 `Agent.inject()` 追加一条原始且模型可见的 `context/message`,其中包含带版本的完整快照。清除操作追加带修订号的墓碑。上下文来源为 `{ kind: 'goal', goalId, revision, round: 0 }`;元数据必须与渲染后的 `<goal_state>...</goal_state>` 内容完全一致。会话日志是唯一的持久事实来源,因此持久化和 fork 会继承目标记录,而无需另设数据库或头字段。
每次非清除变更都通过 `Agent.inject()` 追加一条原始且模型可见的 `context/message`,其中包含带版本的完整快照。清除操作追加带修订号的墓碑。上下文来源为 `{ kind: 'goal', goalId, revision, round: 0 }`;元数据必须与渲染后的 `<goal_state>...</goal_state>` 内容完全一致。这个描述性分隔符沿用了仓库已有的 `<workspace_context>` 约定,也符合 [Anthropic 关于用一致且描述明确的 XML 标签组织混合提示词内容的公开指南](https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/claude-prompting-best-practices#structure-prompts-with-xml-tags)。这是公开的模型体验先例,并非对任何提供方专有训练语料的推断。会话日志是唯一的持久事实来源,因此持久化和 fork 会继承目标记录,而无需另设数据库或头字段。
回放折叠会校验 JSON 形状、来源归属、渲染内容、新 id、修订连续性、生命周期转换、计数器以及单个目标内单调递增的时间戳。目标回合是当前活跃修订上带正数且连续编号的 `user/message` 来源,且不能超过 `maxGoalRounds`;普通会话轮次不会影响该计数器。当前格式的畸形记录会使回放失败,而不会被忽略或修复。
@@ -26,7 +26,7 @@ Status: implemented
### 生命周期与实时激活态
最多只有一个当前目标。创建要求不存在未完成的当前目标,并始终生成该会话此前未使用过、修订号为一的 id已完成目标可以被替换。其他每次变更都携带预期的 `GoalRef`,陈旧的 id 或修订号会被拒绝。仅当回合上限仍有余量时,停止阶段或已解除激活的活跃目标才能恢复;只有已接纳回合数达到上限后,才能标记预算受限
最多只有一个当前目标。创建要求不存在未完成的当前目标,并始终生成该会话此前未使用过、修订号为一的 id已完成目标可以被替换。其他每次变更都携带预期的 `GoalRef`,陈旧的 id 或修订号会被拒绝。仅当回合上限仍有余量时,暂停或阻塞阶段以及已解除激活的活跃目标才能恢复。领域层校验阻塞原因的形状,但会把原因代码和是否阻塞的决策留给策略消费者
从任何种子构建的缓存都以未激活状态开始,每次 `agent/session-start` 边沿也会再次解除激活。因此,恢复和 fork 会保留持久目标与历史,但绝不会自行启动工作。后续人类提示词可由模型解释,其策略表面可以显式调用恢复操作并激活目标。
@@ -36,7 +36,7 @@ Status: implemented
## 测试
单元测试固定创建默认值、精确实时 agent 校验、比较并交换拒绝、所有生命周期转换、上限执行、清除与替换、种子回放和 `SessionStore.fork()` 继承、会话启动时解除激活与活跃目标重新激活、FIFO 延迟变更协调、重入追加观察、注入拒绝回滚、损坏事件的稳定回放、服务与监听器销毁、监听器隔离、挂钟后退钳制、严格记录解码、生命周期连续性、来源与内容一致性,以及连续目标回合归属。无密钥 Loader/stdio 进程测试通过测试专用 `cordis.yml` 挂载服务与生命周期消费者,再从外部读取持久 JSONL以验证模型可见快照以及不存在未经请求的目标回合。包源码受仓库逐文件 100% 覆盖率门禁约束。
单元测试固定创建默认值、精确实时 agent 校验、比较并交换拒绝、所有生命周期转换、阻塞原因校验与保留、恢复时的上限执行、清除与替换、种子回放和 `SessionStore.fork()` 继承、会话启动时解除激活与活跃目标重新激活、FIFO 延迟变更协调、重入追加观察、注入拒绝回滚、损坏事件的稳定回放、服务与监听器销毁、监听器隔离、挂钟后退钳制、严格记录解码、生命周期连续性、来源与内容一致性,以及连续目标回合归属。无密钥 Loader/stdio 进程测试通过测试专用 `cordis.yml` 挂载服务与生命周期消费者,再从外部读取持久 JSONL以验证模型可见快照以及不存在未经请求的目标回合。包源码受仓库逐文件 100% 覆盖率门禁约束。
## 考虑过的替代方案
@@ -52,7 +52,7 @@ Status: implemented
- 恢复与 fork 会暴露同一持久阶段,但在显式恢复变更激活目标前不会执行任何操作。
- 完整快照便于检查和严格回放,但在压缩隐藏它们之前,会在模型历史中重复目标描述与状态字段。
- 修订号与生命周期校验会尽早拒绝遭篡改、部分写入或生产者不一致的目标记录。
- 回合上限只约束继续执行次数token、费用、时间提供方限制仍属于独立策略
- 回合上限只约束继续执行次数;当回合、token、费用、时间提供方限制停止工作时,策略消费者会把它们映射为不同的阻塞原因
## 已知限制与延期工作