Introduce web access as a first-class capability seam so the model-facing
web tools stay stable while backends change. dsh-web owns ctx.web as a
provider registry with registration-order-independent selection and the
WebError taxonomy; dsh-web-search-exa, dsh-web-search-perplexity, and
dsh-web-fetch-local register capabilities into it; dsh-tool-web is the sole
owner of the model-facing web_search/web_fetch schemas, prompt sections, and
HTML-to-markdown presentation. Search and fetch are deliberately one seam.
Providers ship as namespace plugins that register into ctx.web (like an
LlmAdapter into ctx.llm), not key-owning services, since multiple search
providers cannot each own the key. Tool registration follows product
enablement, not backend availability, so load order/credentials never enter
the model contract; the seam resolves the provider at execution time and
surfaces a structured WebError otherwise.
Moves the RFC to implemented/ amended to match what shipped. Example/app
configs are intentionally not wired yet (RFC migration step 6).
The first OUT-OF-PROCESS subagent backend, proving the seam generalizes past the
in-process backends. @deepseek-ai/dsh-subagent-acp runs each child agent in a
spawned subprocess, driven over the Agent Client Protocol as the CLIENT — the
direction-inverted twin of the dsh-acp server bridge. Point the configured
command at the acp-agent example and the harness talks to its own process.
- Fresh process per run: start spawns, runs one ACP session (initialize →
newSession → prompt), dispose kills the subprocess and awaits its exit.
- Minimal client stub: advertises no fs/terminal; accumulates agent_message_chunk
text as the result output; auto-answers session/request_permission by a
configured policy (reject default / allow). No start-time capabilities (an
out-of-process child can't enforce the parent's depth/tool-filter); ignores
request.parent; injects only `subagents`.
- StopReason mapping (end_turn→completed, cancelled→aborted, …); result resolves
error/aborted on a child failure, never rejects (seam contract).
- Security: credential-shaped ambient env vars are scrubbed; the child's own key
is forwarded only via explicit config.env. A spawn-level error (ENOENT) is
captured and raced against the ACP drive so a bad command settles error rather
than crashing the parent.
Testing designed at every tier: keyless integration drives a scripted mock ACP
server subprocess (cancellation incl. the pre-newSession race and a
torn-pipe-after-cancel, permission auto-answer, non-message updates, spawn
failure, HMR, export shape) at 100% coverage; a with-key e2e drives the REAL
acp-agent example process (PONG + real file write, verified on disk) — the
harness driving itself. Snapshot coverage of an ACP child is deferred as
TODO(acp-subagent-replay) (each child is its own process with its own replay).
Stayed on @agentclientprotocol/sdk 0.25.1: the proposed 0.28.x bump only
deprecates the stable ClientSideConnection/AgentSideConnection API this layer
uses (33 sites incl. the server bridge), turning no-deprecated red across code
this PR shouldn't rewrite — that fluent-API migration is its own follow-up. The
backend needs nothing 0.28.x adds.
This completes the subagent seam stack (PR1 interface → PR2 in-process → PR2.5
snapshot infra → PR3 ACP); the seam RFC moves to implemented/, amended.
Two merge-blocking bugs in the shared in-process run driver, both rooted in
`readResult` scanning the whole child session and deriving the stop reason only
from `turn/end`:
- A pre-turn `cancel()` cleared the queued prompt before any `turn/end` was
logged, so the run settled `error` instead of `aborted`, violating the
`SubagentRun.cancel()` contract. The driver now tracks that a cancel was
requested and maps the no-turn case to `aborted`.
- A fork child whose own turn produced no `assistant/message` returned the
SEEDED parent's last message as a `completed` success. `readResult` now scopes
to the child's OWN events (after the seed prefix), so a message-less child
yields empty output.
Both fixes carry a regression test proven to go red on the pre-fix driver.
Also: correct the `SubagentRun.id` / event-payload docs (it is the child AGENT
id, not a session id — the backend mints distinct tokens); refresh the stale
`coding-agent` welcome string (subagent is now a tool); and replace the stale
`TODO(sub-agents)` "deferred" prose in the Agent interface, core.md, and
architecture.md with an accurate pointer to the realized seam.
BLOCKER — the published lib/bin.js (stdio + acp) was exercised only via tsx
(demo:* / the src/bin.ts smokes); the built artifact under plain `node` was
unguarded. Root-cause on the BUILT bin:
1. Settle race: boot() returned once loader.create() registered the include
ENTRY, but the include loads its child plugins asynchronously — so boot()
(and main()) resolved while the app plugins (stdin reader, agent loop, ACP
bridge) were still mounting. A CLI with no attached handles yet exits 0
silently, and a load error surfaces as an unhandled rejection AFTER boot.
Fix: `await ctx.loader.await()` after create() — settle the whole tree.
2. Config-path robustness: hand the include the config's ABSOLUTE file:// URL
so resolution never depends on ctx.baseUrl / can never fall back to cwd.
Both bins fixed identically. NOTE: the cordis Loader resolves a config's bare
plugin specifiers via its internal module loader, active only under
`node --expose-internals`; the bin cannot add a node flag itself, so this is
documented in the bin JSDoc + both package READMEs (the demos already comply).
The repo `examples/*/cordis.yml` are tsx-only artifacts (workspace plugins
resolve through the tsconfig paths map, not node_modules), so they are not a
valid plain-node bin target — the smokes use a real-install-shaped temp dir.
Fail loud on a load failure: boot() previously exited 0 SILENTLY when a config
path's directory does not exist — the include plugin fails to IMPORT, the cordis
Loader catches+LOGS it and leaves the entry with no fiber (no rejection), and
`loader.await()` does not rethrow (EntryTree.await uses Promise.allSettled). Fix:
boot() now calls assertEntriesLoaded(ctx) after the tree settles and throws on
any entry with no fiber, so a typo'd config dir exits non-zero with a clear
message. main() also installs an unhandledRejection guard (installFailLoud) that
replaces Node's stack dump with a single labelled stderr line for the
companion case (a missing config FILE in a real dir, whose include-init throw
surfaces as a rejection Node already exits non-zero on). Regression tests added
to both built-bin smokes (missing dir + missing file → non-zero exit + stderr);
verified the missing-dir test fails on the pre-fix bin.
Built-bin smokes (the reviewer's ask): packages/ui/{stdio,acp}-agent/tests/
built-bin.e2e.ts run the REAL lib/bin.js under `node` (NOT tsx) in a temp
consumer dir, asserting the stdio echo round-trip / the acp initialize response
+ stdout purity, plus the fail-loud cases above. They build-gate (skip if lib/
absent) and run in a new ci.yml step after the build.
Issue 2 — packages/README.md + docs/architecture.md said "plugins depend on
interfaces, never on the concrete loop", but dsh-agent-core imports the concrete
dsh-agent-loop. Scope the rule to EXTENSION plugins and carve out the sanctioned
COMPOSITION/bundle exception (dsh-agent-core composes the concrete spine); note
it in the implemented RFC too.
Issue 3 — examples/acp-agent/tests/acp.snapshot.ts fixture-guard claimed
no-model scenarios need no session.jsonl, but runScenario() always boots
llm-replay with the session.jsonl path and loadReplayScript() throws when it is
absent. Require session.jsonl for ALL scenarios (no-model ones ship a
header-only fixture) and rewrite the comment to match reality.
The session event vocabulary carried two standalone trace-only events that
were not load-bearing as separate records. Fold their facts into nearby
load-bearing events and delete the standalone variants.
- Token usage now rides on `assistant/message` as an optional `usage` field —
the assembled model output and its accounting travel together. The loop folds
`assembler.usage` onto the append instead of emitting a separate `usage`
event.
- The max-tokens path is the no-data-loss host: a step cut off with usage but
EMPTY content (e.g. only a dropped tool call) previously emitted a standalone
`usage`; it now records an empty-content `assistant/message { content: [],
usage }`. `deriveMessages()` skips empty-content assistant messages, so the
usage host never injects a spurious content-less assistant turn into the
provider transcript. A step with neither content nor usage appends nothing.
- An operational error's step number now rides on `turn/end.reason` for
`kind: 'error'` (`{ kind: 'error', step, message, code? }`) — the durable
turn outcome ACP and resume already consume. `failTurn` sets the reason
directly (no separate session `error` event). `agent/error` + logging are
unchanged for live diagnostics.
- No format-version bump: pre-release, no persisted data, so per the format
policy there is nothing to migrate or reject (the RFC's "refresh the format
version" criterion over-reached). `version` stays 1.
- ACP fixtures + goldens re-recorded (keyless replay): dropped standalone
usage/error lines, usage folded onto assistant/message, error step on
turn/end.reason.
RFC moved proposed -> implemented with an implementation note recording the two
scope refinements.
Codex's second pass caught that the prior doc fix swapped one wrong primitive
for another: framing teardown as cancel()+whenIdle() (or awaiting
agent.whenIdle() on disposal) is still wrong. whenIdle() only OBSERVES
quiescence; cancel() only stops queued/in-flight work. Neither unregisters the
agent or detaches the session. Real teardown is AgentHandle.dispose(), whose
disposer does `stop(); await agent.done` — stop the loop, await its exit, and
unregister (packages/core/agent-loop/src/index.ts:271). Copying the old framing
would reintroduce the orphaned-agent/session leak the AgentHandle seam exists
to prevent.
- docs/architecture.md: whenIdle() is a non-owner quiescence-observation hook,
explicitly NOT teardown; teardown is `await AgentHandle.dispose()`.
- docs/cookbook/extension-cookbook.md (prose + the ts comment): tear agents
down via AgentHandle.dispose(), not agent.whenIdle().
- docs/rfc/proposed/feature/2026-06-14-acp-agent-client-protocol.md: the
lifecycle/disposal paragraph routes teardown through the handle's dispose().
Codex's no-ship was a completeness/docs-sync gap, not loop behavior:
- docs/architecture.md: drop the public abort() handle row; the teardown
signal is now cancel() then await whenIdle().
- cancel.spec.ts: the module doc and the turn-start comment contrasted
cancel() against a public abort() verb that no longer exists — reword to
name the loop's private step AbortController.
- packages/ui/acp/src/index.ts: the post-resume-leak comment cited abort();
cancel() is the surviving stop verb that likewise does not unregister.
- Move the RFC proposed -> implemented/simplification with amended text:
Status flips, the both-removal proposal is narrowed to abort-only, and an
implementation note records why whenIdle() is retained (load-bearing
quiescence primitive with live ACP consumers). Update docs/rfc/README.md.
- AGENTS.md "RFCs are proposals, not golden truth": add the concrete
abort/whenIdle worked example now that the implemented RFC exists to link.
- Regenerate the cordis catalog (line-number drift from the rebase).
The public Agent handle exposed abort() (step-only) and cancel() (queue-aware).
No production caller used abort() — ACP maps session/cancel to cancel(), and
lifecycle owners tear down via AgentHandle.dispose(); the loop's own stop paths
abort their per-step AbortController directly. So abort() is latent generality
that keeps a private loop mechanic public.
RFC-premise correction: the public-agent-stop-surface RFC proposed removing
whenIdle() too. Implementation found whenIdle() load-bearing — a real
quiescence primitive with a deliberate loop contract (settle-without-transition,
the replacement-turn race) and ACP test consumers; its proposed replacement
("observe the running->idle transition") is exactly the async-state race
AGENTS.md warns against. So only abort() is removed; whenIdle() stays. The RFC
is amended on the way to implemented/ to record the narrowed scope, and the new
AGENTS.md "RFCs are proposals, not golden truth" principle (PR1) gets its
worked example.
- Remove Agent.abort() from the interface + the ReactLoopAgent impl; the no-arg
'aborted' default goes with it (cancel() keeps its 'cancelled' default).
- Migrate tests: empty-queue abort() -> cancel(reason); the two review-fixes
tests whose subject is the in-flight step's AbortController drive that
controller directly via the private currentAbort field (cancel() would clear
the inbox and destroy the queued steering one of them proves survives a step
abort). The no-arg-default test is dropped (cancel()'s default is already
covered in cancel.spec.ts).
- Resulting public stop surface: cancel() + whenIdle(). Update agent/agent-loop
READMEs, architecture.md, core.md type-equiv, the extension cookbook, the
lifecycle RFC (short note), and the proposed ACP RFC.
Implements docs/rfc/implemented/simplification/2026-06-20-public-agent-stop-surface.md
The LLM service exposed three call surfaces (stream/streamBlocks/generate) but
the only production consumer — the agent loop — uses stream() exclusively,
feeding raw chunks through its own BlockAssembler for replay fidelity. Drop the
speculative convenience surfaces and the registry-change event that no listener
consumed, leaving stream() as the single model-call contract for both
production and tests.
- Remove LlmService.streamBlocks() and generate(), the llm/generate waterfall,
and GenerateResult.
- Remove the llm/adapter-change event (declaration + emits) and the
listener-throw rollback ordering that existed only to protect it; keep the
HMR rollback disposer.
- Remove BlockAssembler.flushReady()/flushRemaining()/result() and the flushed
cursor — the streaming-flush slice existed only for streamBlocks().
- Adapter tests drive a stream()+BlockAssembler helper (tests/assemble.ts)
instead of generate(), exercising the same path production uses.
- Land the AGENTS.md "RFCs are proposals, not golden truth" principle and move
both RFCs proposed -> implemented.
Implements:
- docs/rfc/implemented/simplification/2026-06-20-drop-unconsumed-llm-adapter-change-event.md
- docs/rfc/implemented/simplification/2026-06-20-drop-unconsumed-llm-assembled-surfaces.md
Add a second axis to every RFC — its class (feature, bug-fix,
simplification, architecture, process, testing) — encoded in the path
as docs/rfc/{lifecycle}/{class}/file.md. The folder is the label, so
the closed set is enforced by structure rather than a parsed field.
Two new doc-sync gates back it:
- verify-rfc-classification: every RFC sits in a valid class folder and
the README index lists it under the matching lifecycle→class heading.
- verify-doc-refs: every docs/*.md path cited in a packages|examples TS
comment resolves — closes a drift class verify-md-links can't see, and
catches the four comment refs this reorg moved.
The README gains a Classification section explaining the taxonomy and
per-class index sub-sections. A self-referential process RFC records why
the scheme is path-encoded and gated.
Add scripts/gen-cordis-catalog.ts: a fully-generated docs/cordis-catalog/
events-and-services.md cataloging every cordis event (exact signature + @mode)
and ctx.<key> service (exact interface), modeled on gen-module-graph's
--write/--check freshness gate. The harness tier renders in full from the
interface Events / interface Context declarations and their JSDoc; the inherited
cordis-core/loader/hmr/timer surface renders tersely from a curated table.
The generator hard-errors on a missing @mode tag and on a tag that contradicts
a conclusive signature shape (a trailing next param is structurally a
waterfall). Signature blocks use a ts cordis-catalog fence that doc-typecheck
skips. Type tokens cross-link to the core-data-structures catalog.
This supersedes the hand-maintained event-taxonomy table: verify-event-taxonomy
is deleted and verify-cordis-catalog joins doc-sync. architecture.md keeps the
Event taxonomy heading (TOC anchor) but points at the catalog; the Service-map
role table stays. RFC, AGENTS.md @mode authoring rule, and dependent doc/skill
references updated. Negative gate tests cover the missing-tag and
tag/shape-contradiction paths.
A new docs/core-data-structures/ folder: a self-contained core.md defining what
counts as a "core" data structure (the agent-loop spine) and covering the spine
vocabulary, plus per-seam sub-pages (llm-streaming, session, persistence, tools,
bash). Type definitions are pasted verbatim via `ts type-equiv` blocks and
drift-checked by verify-type-equiv. Cross-linked from architecture.md; the
`ts type-equiv` mechanics are documented in development.md.
The bridge now holds each session's `AgentHandle` disposer in its
`SessionRecord` and runs it on teardown (client disconnect or fiber dispose)
instead of the old `abort()` + `whenIdle()` drain that left agents
registered. A bare client disconnect now leaves NO registered agent and NO
session-store entry — not an idled-but-still-registered one. The queue-aware
`cancel()` inside the disposer also closes the former pre-step best-effort
window (a turn about to start is dropped), so teardown reaches true
quiescence.
The `session/load`-races-teardown leak is fixed: if the bridge closed while
`resume()` was pending, the just-resumed handle is disposed before throwing,
so it leaves no orphan (it has no SessionRecord, so quiesce() never sees it).
Tests: the disconnect test now asserts (through the SAME memoized teardown)
that the agent is unregistered AND its session removed; a durability test
re-loads the persisted log after dispose and asserts the closing turn/end is
on disk (guards the teardown-order contract); a sibling-isolation test proves
one handle's dispose() leaves other agents untouched. Docs: agent /
agent-loop / acp READMEs, architecture.md, and the stale in-code quiesce()
ownership comment updated to the per-agent disposal model; the now-resolved
TODO(rfc010-agent-disposal) / TODO(rfc010-cancel-prestep) teardown notes
removed.
abort() only kills the in-flight step, so a queued-but-not-yet-started prompt
ran to completion after a cancel and a prompt accepted right after could be
batched into the cancelled turn (the loop merges queued messages into one turn).
This closes TODO(rfc010-cancel-prestep) with a distinct cancel() verb.
cancel() clears the queued + steering FIFOs, aborts the in-flight step, and
drives a turn-scoped marker on the LoopHandle that the driver checks at EVERY
point a turn could start or continue:
- right after the idle wait (window 1): drop the about-to-run turn and settle
whenIdle() waiters directly (no running→idle transition fires, and no
agent/status is emitted, so an ACP listener can't see a spurious idle that
resolves a freshly-queued prompt as cancelled);
- after the synchronous setStatus('running') emit (window 2): a running listener
can cancel in the gap before runTurn;
- in the step-start window (before runStep, after setAbort): a synchronous
turn-start/step-start listener can cancel before any AbortController exists;
- at the continuation gate: a cancel during the continuation waterfall (the
finished step's controller already cleared) ends the turn aborted.
The marker is ARMED only when there is something to cancel (running, an
in-flight step, or queued/steering work) — an idle no-op cancel cannot leave it
set to drop a later prompt — and RESET unconditionally once per loop iteration,
so it governs exactly one turn and never leaks onto the next prompt (even when a
send() lands in the cancelled turn's flush window).
ACP session/cancel now maps to agent.cancel() (keeping the synchronous
settlePrompt). Teardown/disconnect still use abort('disposed') until PR D, so
the ACP README narrows the remaining best-effort window to teardown only.
Tests (agent-loop/cancel.spec.ts) cover every window unit-level (the F1 hang
guard: a whenIdle() waiter registered before a pre-step cancel resolves; the F2
leak guard: idle cancel then a prompt runs; mid-step, continuation, both
pre-step windows, turn-start-listener, steering-cleared, marker-reset). ACP
turns.spec.ts adds the through-bridge tests with NO intervening whenIdle (idle
cancel→prompt runs; mid-stream cancel→immediate next prompt runs) and updates
the stale pre-step test to the queue-aware guarantee. The existing cancel
snapshot golden is byte-identical (it drives the new cancel() path end-to-end
through the real subprocess), so no new golden is needed. 100% coverage.
SessionSummary (updatedAt/title/firstPrompt) and SessionPersistence.update()
were dead state: zero production callers of update(), no production reader of
updatedAt/firstPrompt, and ACP's title comes from a tool-call presenter, not
storage. The live Session.header was already typed SessionHeader, so the
summary only ever existed in the persistence layer, written and read by nothing
but its own contract test.
Delete it entirely (no SessionMeta alias — SessionMeta collapses to
SessionHeader everywhere). This removes the JSONL .summary.json sidecar
machinery, the SQLite title/first_prompt/updated_at columns and per-append
updated_at bump, and the update() method from the abstract service and both
backends. SQLite SCHEMA_VERSION goes 1->2 and openDatabase now rejects any
non-current user_version (older or newer) — no migration, unreleased software.
Net -400 lines, and it erases the JSONL-sidecar-vs-SQLite-column durability
divergence that the upcoming write coordinator would otherwise have to model.
Records the decision in docs/rfc/implemented/2026-06-19-drop-mutable-session-summary.md
and migrates the 2026-06-14 session-persistence RFC's facts to current truth.
Adds a standalone AGENTS.md section "Tests document behavior, not golden truth"
(a passing test pins current behavior, not necessarily correct behavior) with
the summary-drop as its worked example, and reinforces the no-migration
pre-release stance.
Rename the concrete Agent class to make its ReAct-style reasoning loop
explicit in the name. Package name, default-export plugin (`AgentLoop`),
and the `ctx.agentLoop` service key are unchanged.
Collapse docs/adr/ and docs/rfc/ into a single docs/rfc/ with proposed/,
implemented/, and rejected/ subfolders. Every file is renamed to
yyyy-mm-dd-topic-title.md, where the date is when the topic was first
proposed (from git history). ADRs and RFCs that covered exactly the same
topic are merged (property-based testing, session persistence); the
umbrella RFC 005 stays split across its three implemented decisions, and
RFC 006's deferred part-3 (API extractor reports) splits into its own
proposed RFC. All cross-references become machine-checkable relative
links instead of bare "ADR NNNN" / "RFC NNN" prose.
Add a verify-md-links doc-sync gate (scripts/verify-md-links.ts) that
checks every relative Markdown cross-link resolves, wired into doc-sync
alongside verify-md-wrap. This makes the reorganization self-verifying:
the same change that rewrote ~forty inter-doc links adds the check that
proves none dangle. Document the cross-link convention in a new
docs/AGENTS.md and record the gate as an implemented RFC.
doc-sync, typecheck, lint, and the full test suite (667) all pass.
Per review feedback (use a real markdown parser with an AST linked to
source positions), rewrite verify-md-wrap to parse each file with
mdast-util-from-markdown (the CommonMark parser behind remark) + the GFM
extension, then flag any `paragraph` node whose source span covers more
than one line.
Why a parser over the hand-rolled line scanner:
- It is a checker, not a formatter — it reports and never rewrites, so
zero cosmetic churn (no emphasis-marker or table-delimiter
normalization, which is why Prettier was rejected for this).
- The AST owns every structural exemption (fenced code of any fence
length, tables, lists, blockquotes, HTML, headings, reference defs),
fixing both bugs the regex version had: it now catches wrapped
list-item / blockquote prose (a `paragraph` inside those nodes) and no
longer false-positives on a longer ```` fence wrapping an inner ```.
Also unwrap two pre-existing hard-wrapped blockquotes (architecture.md,
adding-a-tool.md) that the stricter AST check correctly surfaced.
A crash can leave a durable log whose final turn never closed. The old
behavior truncated everything after the last turn/end as a "crash tail".
But a single turn can be HUGE in a long-horizon task (many steps, large
tool output), so truncating it silently destroys real, durably-written
work — truncating a turn is wrong.
New crash recovery (ADR 0018): load() PRESERVES the interrupted turn's
events and CLOSES the orphaned turn by durably appending synthetic
boundary events — a step/end if a step was open, then a turn/end carrying
the new merge-extensible TurnEndReason {kind:'interrupted'}. load()
returns the balanced log, so a resumed session is immediately usable. Only
a never-fully-written TORN tail fragment is discarded; corruption in the
committed region is still unloadable.
- dsh-session: TurnEndReason {kind:'interrupted'} + shared
interruptedTurnClosers() repair helper.
- JSONL backend: scanLog preserves the longest contiguous prefix
(including a partial final turn); loadCore truncates a torn fragment and
durably writes the closers, returning the balanced log.
- runPersistenceContract gains a crash-recovery test (both backends + mock).
- Docs: ADR 0018/0017, architecture.md, package READMEs.
Also (review #33): RFC 013 records the "move event vocabulary to Zod"
question (merge-extensible maps → runtime schema registry) + blast radius;
deferred, not done here.
Implements the RFC 010 MVP: a new `@deepseek-ai/dsh-acp` package bridges the
harness agent to the Agent Client Protocol (JSON-RPC 2.0 over newline-delimited
stdio), so Zed and other ACP editors can drive the coding agent — streaming
render, tool-call display, and resumable sessions via `session/load`.
- packages/acp: AgentSideConnection wiring; initialize/newSession/loadSession/
prompt/cancel; a total TurnEndReason→StopReason codec; settle-once with a
fallback chain (agent/turn-end → logged turn/end → idle); single-session
guard; cwd-must-equal-launch-dir validation; load replays from the persisted
event log (assistant/chunk→agent_message_chunk, tool/call/result→tool_call*).
- agent: add Agent.whenIdle() quiescence signal to the interface; LoopAgent
implements it (resolves on the first running→idle/disposed transition). The
bridge awaits it on disposal so teardown reaches quiescence, not just abort.
- examples: extract the shared provider/tool core into examples/base.yml;
coding-agent nest-includes it; new examples/acp-agent serves the agent over
ACP with JSONL persistence and no stdout logger (stdout is the protocol).
- Permission gate deferred (TODO(rfc010-permission-gate)): tools run with the
executor's full authority; only the Agent→sessionId ownership seam is laid
down. Cancel is best-effort for a not-yet-started queued turn
(TODO(rfc010-cancel-prestep)). RFC 010 stays `proposed`.
- Docs: package README + Zed snippet; client-driver cookbook section; root and
packages layout/commands; RFC 010 implementation-status note.
48 bridge tests + whenIdle coverage; 100% per-file coverage; e2e boots the
example as a subprocess and verifies a written file on disk (key-gated, with a
no-key stdout-purity check).
Add a `max-tokens` variant to `TurnEndReasonMap` and carry the model
finish reason up from `runStep` to `runTurn`, applying the rule "any
max-tokens step in the turn surfaces as max-tokens" (disposed/aborted/
error still take precedence). This lets consumers distinguish a clean
stop from a truncated one — the contract RFC 010's ACP bridge maps to
the `max_tokens` stop reason.
Also add an AGENTS.md rule: write an ADR when (and only when) a PR makes
a durable, contested, surprising decision.
Add a SQLite SessionPersistence backend (node:sqlite), a SECOND
implementation built to prove the abstract seam + the shared
runPersistenceContract suite are genuinely backend-agnostic. Each
SessionEvent maps 1:1 onto an events row (session_id, seq, type, time,
data); append is an INSERT inside a transaction asserting the
contiguous-seq contract; the mutable SessionSummary lives in the
sessions metadata row.
It satisfies the SAME contract semantics as the JSONL backend, expressed
over rows instead of file bytes:
- Lazy materialization: create() records intent in memory; no row until
the first append (a never-appended session is absent from has()/list()
via a materialized flag set inside the first append transaction).
- Crash-tail-on-load: load() returns events only through the last
complete turn/end and deletes the uncommitted tail; a seq gap in the
committed region makes the session unloadable.
- Transactional append: a mid-batch failure (a UNIQUE seq collision from
a concurrent writer) rolls back entirely, keeping the cursor truthful.
Like the JSONL backend it is also the write-path plugin (session/event →
buffer → session/flush drain, onCreated seed/adopt/collision handling,
HMR seeding, dispose-to-quiescence). The package runs the shared
runPersistenceContract suite plus SQLite-specific tests (transaction
rollback, crash-tail cut, schema version, HMR adoption).
Docs flip every "SQLite is future/deferred" reference (ADR 0016,
architecture.md, the persistence module doc + README) to "implemented;
the contract holds both backends to identical semantics".
Add the agent-creation factory seam on ctx.agents (AgentRegistry):
setFactory/create/resume plus the AgentFactory interface and
CreateAgentOptions/ResumeAgentOptions. AgentLoop implements AgentFactory
and registers itself via ctx.agents.setFactory(this), so plugins
create/resume agents through the interface without depending on the
concrete loop package.
- create({ agentId, sessionId, meta?, agentOptions? }) — programmatic
create on a caller-supplied session id (e.g. an ACP-generated id).
- resume({ agentId, resumeSessionId, agentOptions? }) — load a persisted
session via ctx.sessionPersistence (RFC 009) and resume an agent on it;
the live session id is the resumed id, turn numbering and derived
history continue from the loaded log. sessionPersistence is NOT
hard-injected (non-persistent demos still work); resume rejects with a
typed error when it is absent. assertAgentIdFree runs before any
session is created (and again after the load await) so a duplicate id
never leaves an orphaned live session.
Adds the runtime dsh-session-persistence dependency to agent-loop.
Add the durable session-persistence capability seam (ADR 0016): an
abstract SessionPersistence service (dsh-session-persistence,
ctx.sessionPersistence) defining create/append/load/list/has/delete/
update over the existing SessionEvent — no parallel persisted type — and
a first implementation (dsh-session-persistence-jsonl): an append-only
JSONL log per session with crash-safe atomic writes, truncation-repair
of a never-committed crash tail, and a read/replay path. SessionMeta
(format version, cwd, lineage) travels out-of-log via session.header.
A shared runPersistenceContract suite holds every backend to the same
append-only / contiguous-seq / lazy-materialization / serializability
semantics.
Config-driven create() now uses a per-run ${id}-session-<uuid> session
id so a fixed name no longer collides with an on-disk log once a durable
backend is loaded; each run is a new session (a demo simplification). The
examples drop their hand-rolled session-jsonl.ts and load the JSONL
backend via cordis.yml; CI smoke-loads it too.
The agent-facing create/resume factory that consumes load() is a
separate seam, deferred to a follow-up; this change stops at the load
primitive and does not reach into the loop.
Every session event now lives inside a turn (between turn/start and its
turn/end). The loop records queued user/message events AFTER turn/start;
an idle agent.inject() wraps its context/message in a one-shot injection
turn. This makes the turn the single durability/replay boundary so a
persistence backend can treat anything after the last turn/end as a
crash tail without dropping legitimate between-turn context.
A failure once the turn is already closed (rejecting session/flush, a
throwing agent/turn-end listener) has no in-turn position for a session
error event, so it is reported via agent/error + logger only; the turn
stays balanced. failTurn appends an error event only while the turn is
open.
The dsh-invariants plugin enforces turn-enclosure via a default case:
every non-boundary event type — including plugin-added merge-extensible
keys — must sit inside an open turn or it throws.
Documented in ADR 0017 + architecture.md.
Every loop boundary appends the session event before emitting the Cordis
event (ADR 0003's append-before-emit rule) — except step/start, which was
inverted. A listener on agent/step-start that inspected session.events could
not see the step it was just told had started.
- Swap the two lines so session.append('step/start') precedes the emit.
- Fix the two stale pseudo-code copies (the runLoop JSDoc STEP-loop block and
docs/architecture.md) so neither shows step-start emitted before the append.
- Regression test: a step-start listener observes the matching step/start
event already at the tail of session.events. Verified the test fails on the
pre-fix (emit-first) order.
Two tsx CI gates make doc/code drift fail fast:
- doc-typecheck extracts every fenced ts block from README/docs/package READMEs,
compiles them with tsc --noEmit against a temp project (vendor->lib, harness->src
paths from tsconfig.typecheck.json), and fails on errors. Deliberate sketches opt
out with ```ts ignore-check; the opt-out ratio is reported and capped.
- verify-event-taxonomy asserts the docs/architecture.md taxonomy table names
exactly the events declared in the interface Events blocks. This surfaced three
events the table had been missing (tools/change, llm/adapter-change,
system-prompt/change), now added.
Doc snippets made compilable with stub imports/declares (1 genuine sketch ignored).
Wired into CI after typecheck. API reports (RFC 006 pt 3) deferred. Graduates RFC
006 pts 1-2 -> ADR 0014.
- AGENTS.md Commands: fix typecheck/build descriptions; add lint, lint:fix,
test:coverage, knip, publint, hygiene (were undocumented).
- Drop the bare `yarn demo` for explicit `demo:echo` + `demo:coding`; update
README, examples READMEs (and document coding-agent in examples/README).
- New cookbook guide: adding-a-vendored-package.md (the missing "add" half of
vendor/README's update-only procedure).
- architecture.md: add a table-of-contents and extract the Extension cookbook
to docs/cookbook/extension-cookbook.md (link-preserving); drop the completed
"restructure this document" TODO.
- ADR 0009 (capability seams) + 0010 (twin LLM adapters), and a "when to write
an ADR" standard in adr/README.
- Add a committed dsh-code-review skill under .agents/skills, exposed to Claude
Code via a tracked .claude/skills symlink (gitignore carve-out).
Hard line breaks mid-paragraph make docs harder to edit and diff — a
one-word change reflows and re-diffs the whole paragraph. Reflow all
tracked non-vendor Markdown (plus vendor/AGENTS.md) so each prose
paragraph is a single line; soft-wrapping is the editor's job. Fenced
code, tables, and list structure are preserved (wrapped list items fold
to one line per bullet). Documents the convention in AGENTS.md.
The first real agent wiring: DeepSeek V4 + the bash tool suite + stdio
chat + JSONL persistence, runnable via yarn demo:coding (reads the
gitignored repo-root .env through process.loadEnvFile).
- examples/coding-agent: cordis.yml wiring both real plugin families
(llm-deepseek with !!js env secrets; bash-local + tool-bash), a
bash-only coding system prompt, a max-steps-guard plugin (bounds
runaway turns via the agent/turn-continuation waterfall — abort()
from step-end is a no-op by then), and a stdio UI with dimmed
reasoning and exit-on-idle for piped stdin.
- e2e (yarn test:e2e, key-gated): full-loop.e2e.ts runs a real model
against the real bash tool; coding-task.e2e.ts is the swebench-style
smoke — the model fixes a buggy add.js in a temp dir and the test
re-runs node add.test.js itself rather than trusting the agent.
- docs/cookbook: adding-a-package (the verified checklist),
adding-a-tool (execute() contract, background pattern, seams),
adding-an-llm-adapter (protocol obligations, mock-server testing,
e2e policy). AGENTS.md layout/commands/secrets sections updated;
architecture.md points at both examples and the cookbook.
- vitest.e2e.config.ts: serialize test files + retry twice — parallel
e2e files trip the shared internal key's concurrency quota.
- fix: the !js YAML tag spelling in docs/JSDoc is actually !!js
(js-yaml resolves custom tags under tag:yaml.org,2002:js).
The first real LlmAdapter implementations, shipped as a deliberate pair:
same models and wire protocol, completely different internals, so the
StreamChunk protocol is verified across independent implementations.
- dsh-llm-deepseek: hand-rolled fetch + SSE parser + chunk-translation
state machine against the official chat-completions format (thinking
mode via top-level thinking/reasoning_effort; the empty-string
reasoning_content first chunk; usage attached to the finish chunk or
trailing; reasoning_content passback on tool-call turns; disjoint
cache-token accounting).
- dsh-llm-pi-ai: the same endpoint through @earendil-works/pi-ai,
mapping its event vocabulary (parsed tool arguments, in-stream error
events, folded reasoning tokens) onto the same chunks.
The agent loop now honors the in-band error path: an adapter that ends
its stream with finish {kind:error|aborted} (the only option for
adapters that can't throw mid-stream, like pi-ai) is translated into a
step error, so the turn ends error/aborted with a logged error event
instead of a normal completed assistant message. This makes the
StreamChunk error contract real for both adapters; docs/architecture.md
and the StreamChunk doc are updated accordingly.
New yarn test:e2e (vitest.e2e.config.ts, *.e2e.ts) runs key-gated
real-API matrices for both adapters across V4 Flash/Pro and all
thinking/effort levels; it self-skips without DEEPSEEK_API_KEY. Unit
suites run against local node:http mock SSE servers at 100% per-file
coverage.
Three packages following the new capability-seam pattern (interface /
implementation / consumer, now documented in docs/architecture.md):
- dsh-bash: abstract BashExecutor service (ctx.bash) + vocabulary types.
- dsh-bash-local: local subprocesses — bash -c per call in a detached
process group, SIGTERM→SIGKILL group kills, tail-keep truncation with
full-stream spill files, model-friendly env, background task registry.
- dsh-tool-bash: the bash / bash_output / bash_kill tool schemas with
runtime arg validation and background completion notices via
agent.inject(). Non-zero exits are reported, not errored.
Design surveyed against the bash tools of Claude Code, OpenCode, Codex,
and pi (notes in the package READMEs). Permissions/sandbox stay TODO on
the tools/execute waterfall seam; stateful-shell alternatives recorded
in run.ts.
The architecture cookbook's tool-plugin example now uses defineTool
with typed args (the raw-JSON-Schema + `args: any` example contradicted
the type-safety policy it sits next to); a note explains raw schemas
remain the MCP interop path. The echo-agent README's mock-llm row now
matches the code (registerAdapter(['mock-echo'])) and the echo-tool
row mentions the typed registration.
Dudu-0223
Hypatia May
Tianyi Cui