Merge remote-tracking branch 'origin/master' into codex/enforce-tool-cancellation

# Conflicts:
#	docs/event-producer-consumer.md
#	examples/acp-agent/tests/snapshots/bash-spill/session.jsonl
#	examples/acp-agent/tests/snapshots/escalation-approved/session.jsonl
#	examples/acp-agent/tests/snapshots/escalation-rejected/session.jsonl
#	examples/acp-agent/tests/snapshots/fs-escalation-approved/session.jsonl
#	examples/acp-agent/tests/snapshots/hook-cc-pretool-ask/session.jsonl
#	packages/context/workspace-context/tests/workspace-context.spec.ts
#	packages/core/agent/src/index.ts
#	packages/support/invariants/tests/invariants.spec.ts
#	packages/ui/acp/src/index.ts
#	packages/ui/tui/tests/harness.ts
#	packages/ui/tui/tests/tui.spec.ts
This commit is contained in:
Tianyi Cui
2026-07-21 20:03:00 +08:00
623 changed files with 21217 additions and 3024 deletions

View File

@@ -4,9 +4,9 @@ The **DeepSeek Harness SDK** builds on Cordis: **everything is a plugin**, inclu
## Overview
Harnesses are [Cordis](cordis-primer.md) contexts. Packages contribute services (`ctx.llm`, `ctx.tools`, `ctx.sessions`), typed events (`agent/request`, `tools/pre-execute`, `session/event`), and disposable prompts, tools, providers, adapters, and listeners.
Harnesses are [Cordis](cordis-primer.md) contexts whose packages contribute disposable services, events, and registrations.
`packages/core/` groups the default agent flow; surrounding capabilities are equally first-class Cordis plugins.
`packages/core/` groups the default agent flow.
### Default Services
@@ -30,6 +30,7 @@ Harnesses are [Cordis](cordis-primer.md) contexts. Packages contribute services
| `ctx.sandboxPolicy` | [`sandbox/`](../packages/sandbox/README.md) | shared sandbox policy home |
| `ctx.codeRuntime` | [`code-runtime/`](../packages/code-runtime/README.md) | model-written program execution |
| `ctx.fs` | [`fs/`](../packages/fs/README.md) | filesystem provider primitives and policy events |
| `ctx.lsp` | [`lsp/`](../packages/lsp/README.md) | semantic navigation registry |
| `ctx.skills` | [`skill/`](../packages/skill/README.md) | skill provider registry and progressive disclosure |
| `ctx.web` | [`web/`](../packages/web/README.md) | search/fetch provider registries |
| `ctx.compact`, `ctx.toolResultPrune` | [`compact/`](../packages/compact/README.md)/[`compact-tool-result-prune`](../packages/compact/compact-tool-result-prune/README.md) | summary compaction; optional model-free result pruning |
@@ -39,6 +40,7 @@ Harnesses are [Cordis](cordis-primer.md) contexts. Packages contribute services
| `ctx.goals` | [`goal/`](../packages/goal/README.md) | persisted same-session goals |
| `ctx.sessionPersistence` | [`session-persistence/`](../packages/session-persistence/README.md) | durable storage for session logs |
| `ctx.sessionQuery` | [`session-query/`](../packages/session-query/README.md) | live-preferred logical-corpus exact reads and relationship traces |
| `ctx.invariants` | [`support/invariants`](../packages/support/invariants/README.md) | registry and package-name selection for package-owned runtime checks |
## Event
@@ -118,7 +120,7 @@ Pruning precedes summaries; overflow retries require durable progress. Bounded t
The turn contains failures. Adapter failures close the step before `agent/request-error`, which receives exact `Error`, `LlmFailure`, and history. Retry opens another step; success clears history; exhaustion stores failure on `turn/end`. Failed chunks commit no message/tool.
Other failures use `agent/error`. Cancellation and disposal beat recovery; undispatched model tool calls receive synthetic `tool/call` and `ABORTED_BEFORE_DISPATCH` result pairs. One signal spans the turn until retirement before `turn/end`. Effective `cancel()` emits its typed `user | parent` cause before clearing queues and aborting; observers cannot veto it, idle calls emit nothing, and durability records only `aborted`. Disposal awaits quiescence before unregistering ([decision](../.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.md)).
Other failures use `agent/error`. Cancellation and disposal beat recovery; undispatched model tool calls get synthetic `tool/call`/`ABORTED_BEFORE_DISPATCH` pairs. One turn signal retires before `turn/end`. Effective `cancel()` emits its typed `user | parent` cause before clearing queues and aborting; observers cannot veto, idle calls emit nothing, and durability records only `aborted`. Disposal awaits quiescence before unregistering ([decision](../.agents/notes/implemented/architecture/2026-07-16-explicit-turn-cancellation.md)).
Every session event is turn-enclosed. Reloading preserves an interrupted tail and closes it with a synthetic `interrupted` turn end. Failures after durable turn close report only through `agent/error` because no safe in-turn position remains. Each turn has one `TurnEndReason`; [TurnEndReasonMap](core-data-structures/session.md#why-a-turn-ended-turnendreasonmap) owns the variants.
@@ -136,7 +138,7 @@ Every live agent owns a scoped `agent.ctx`. Its registrations shadow globals, re
The session log is the source of truth. `deriveMessages()` projects session events into the `Message[]` sent to the model; raw `assistant/chunk` events stay in the log for replay and UI fidelity. Replay, fork, resume, transcript rendering, telemetry, and persistence all derive from the same event stream.
**Model-visible ⟺ logged**: the log reconstructs every request — messages at `step/start` fronted by the header's session prefix, headers by folding `request/header` — and dev invariants assert this ([reconstructability](../.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md)).
**Model-visible ⟺ logged**: the log reconstructs every request — messages at `step/start` fronted by the header's session prefix, and headers by folding `request/header` — and the package-owned `dsh-agent-loop/invariant` can assert it through `ctx.invariants` ([reconstructability](../.agents/notes/implemented/architecture/2026-07-05-reconstructable-requests.md)).
Durability is a plugin concern. Backends buffer synchronous `session/event` notifications; the loop awaits a turn-end checkpoint. `SessionPersistence` stores `SessionEvent` directly and metadata in `SessionHeader`; JSONL defaults to checksummed Zstandard, with SQLite under one contract.