Brings in the refreshed base (master merged through the stack after #203
and #205 landed), including the acp-snapshot extraction (#204), and
re-ports this PR's snapshot-suite extensions onto the extracted package:
- dsh-acp-snapshot's Scenario gains headerClass and configPath; the suite
factory pins the request header PER CLASS (construction rejects a
missing or duplicated class pin), forwards a scenario's configPath to
the harness (RunOptions.configPath overrides AgentUnderTest.configPath),
and a new fixtures meta-test asserts every pinning fixture carries
exactly one request/header and no deltas.
- The acp-agent example's thin scenario table re-registers code-mode-turn
and both-mode-turn with their overlay configs and per-class pins; the
committed fixtures replay unchanged.
- The package's synthetic suites cover the new surface (explicit
headerClass on one suite, the default on the other, a configPath
override through the fake bin, and the two construction throws).
The master merge added a code-runtime/ package row while this branch adds
the timeout/ row; together they push packages/README.md over its 605-word
ceiling. Condense the timeout/ row to the terse sibling style and raise the
ceiling 605->610 for the genuinely-new package group, mirroring how the
code-runtime work raised architecture.md's ceiling in the same spirit.
Review discussion converged on the industry shape (Claude Code caches
user context per conversation; Codex separates initial context from
diffs; Kimi appends at continuation boundaries to protect prompt
caching): stable openers belong in a compose-once prefix, mid-session
changes belong in append-only history — not in a per-request slot.
agent/session-prefix fires ONCE per loop instance, lazily on its first
request-building step: the composed Message[] is deep-frozen, cached on
the transmission bookkeeping, recorded as EpochHeader.messagePrefix on
the anchoring 'initial'/'resume' snapshot, and reused verbatim for
every request the instance sends — prefix stability is structural, not
a producer discipline, and a resume recomposes with attributable drift.
The request is messagePrefix + boundary snapshot.
The per-step RequestAdvice/RequestAdviceContext surface and the
messageSuffix header field are dropped: the tail slot had no consumer,
and every current update pattern (new AGENTS.md discovered, memory
update, skills change) routes through the existing append-only history
channels — inject(), tools/post-execute additionalContext,
prompt-submit additionalContext — each paid once and prefix-cached
thereafter. The messagePrefix delta arm stays for codec totality; the
loop never produces one in practice.
The zero-config enforcer no longer imports schemastery (its Config was
removed); knip flags the stale dependency. Remove it from the manifest
and sync the lockfile.
fetchTimeoutMs/searchTimeoutMs (default 30000) resolve to each tool's
ToolDefinition.timeoutMs, moving the budget's declaration home onto the
owning tool plugin and preserving per-tool deployment override without a
mistypable central tool-name map.
The enforcer now reads ctx.tools.get(exec.name).timeoutMs instead of a
free-text tool-name config map, so a mistyped name is impossible and the
tools/change warn-once apparatus is gone. exec.name always resolves in the
registry during dispatch, so there is no unknown-name path to warn about.
Both bot criticals verified against the code and rejected as exploit
paths — pinned instead of patched:
The bindings loop already excludes run_code (the skip predates the
finding), and the runtime host resolves forged port calls as own
properties of the bindings record, so an absent binding is unreachable
from a program under any mode. A new both-mode test pins the invariant:
the record has no run_code key on any lookup path.
The drain await cannot mask a run failure: `queue` is the folded tail
(every link swallows its rejection), so `await queue` never rejects and
the runtime's own result.error always reaches the CodeRunFailedError
conversion — the existing abort test exercises exactly the
queued-abandonment-plus-run-failure scenario. Stated at the drain site so
the fold's purpose is explicit.
A tool declares its cooperative timeout budget on its own definition
rather than a deployment naming it in a central config map. The field
never reaches the model (schemas() whitelists name/description/parameters)
and defineTool rejects a non-positive-finite value at authorship.
- syncTools: on paginated listTools failure, unregister any tools already
registered in the current sync before rethrowing (prevents orphans)
- Effect disposer: call client.close() directly without awaiting startup
completion — aborts a hanging connect promptly on HMR/dispose
Three findings from the PR-4 convergence round:
(A) A root-undefined binding argument passed normalization untouched, so
the sub-call DISPATCHED and only then failed the tool/code-dispatch append
(Session.append rejects undefined event data) — a sub-call executed with
no log record, violating the nothing-executes-unlogged contract. And the
tool received the SAME object later handed to the append, so a tool
mutating its args desynced the logged record from what was dispatched (or
re-poisoned the append). jsonNormalizeArgs now rejects undefined up front
with a model-correctable message and returns TWO independent parses of the
canonical JSON text: the tool gets one, the event logs the sibling —
identical by construction, mutation-proof.
(B) The bridge built its bindings record with plain-object assignment, so
a registered tool named __proto__ hit the prototype setter and silently
vanished (the runtime host resolves binding names as own properties). The
record is now null-prototype with defineProperty, mirroring the
worker-side namespace build.
(B) The header-pin sanity assertions ran only inside NON-pinning
scenarios, so a class consisting solely of its pinning scenario (the two
Code Mode classes) would accept a re-recorded pin carrying several headers
or a header-delta. A fixtures meta-test now asserts every pinning fixture
directly.
- syncTools: on paginated listTools failure, unregister any tools already
registered in the current sync before rethrowing (prevents orphans)
- Effect disposer: call client.close() directly without awaiting startup
completion — aborts a hanging connect promptly on HMR/dispose
- syncTools: on paginated listTools failure, unregister any tools already
registered in the current sync before rethrowing (prevents orphans)
- Effect disposer: call client.close() directly without awaiting startup
completion — aborts a hanging connect promptly on HMR/dispose
Prove the full MCP protocol flow works end-to-end against real servers:
- Self-written fixture server: tool discovery, execution, error handling,
image placeholder, toolPrefix, and clean disposal
- @modelcontextprotocol/server-everything: echo, get-sum, get-tiny-image
- @modelcontextprotocol/server-filesystem: write_file + read_file round-trip,
list_directory with world-verification
All 15 tests keyless and deterministic (no API key needed).
Addresses CI coverage gap: exercises the latest-wins resync coalescing
(pendingResync branch), the client.onclose callback, and ensures index.ts
is loaded without module mocks for stable v8 coverage across environments.
The dsh-tools half of the Code Mode RFC (its fourth, final change): the
registry gains its first config — mode: native | code | both — and OWNS how
its tools reach the model. 'code' contributes exactly one wire tool,
run_code, plus a lazy tools:sdk prompt section declaring every other tool
as a generated TypeScript API (jsonSchemaToTs: total over the defineTool
subset, unknown degradation, lexicographic byte-identical rendering);
'both' ships both representations; 'native' is byte-for-byte the old
behavior. Non-native modes fail every assembly loudly without a
typescript-language ctx.codeRuntime.
run_code's dispatch bridge: JSON-normalizes each binding argument before
dispatch (what dispatches is what the tool/code-dispatch event logs — the
append can never fail on payload shape; BigInt/circulars reject that one
call), serializes all program tool calls through a per-run queue (even
Promise.all — no concurrency-safety metadata yet), routes every sub-call
through tools/pre-execute → tools/post-execute (a deny rejects the
program-side promise), drops sub-call additionalContext (no safe outlet
mid-run; pinned), owns a run-scoped abort that follows the outer signal in
and fires on settlement (in-flight sub-dispatch aborted, queued abandoned,
queue drained before returning), and converts a failed run into
CodeRunFailedError → a structured isError carrying kind + captured logs.
tool/code-dispatch joins SessionEventMap by declaration merging (log-only;
deriveMessages ignores it).
The composed surface: the tools config forwards through agent-core and
both app packages; examples/code-agent + demo:code run the worker runtime
under mode code (keyless boot smoke + a with-key e2e proving the collapsed
[run_code] header, the dispatch events, and the file the program wrote);
two new snapshot scenarios (code-mode-turn, both-mode-turn) record the SDK
section, collapsed header, dispatch events, and result card — each its own
header-pinning class (the harness gains per-scenario config overlays and
per-class pins). Catalogs, graphs, cookbook, hooks-bridge notes, and the
RFC (moved to implemented/, restructured to decision-era headings) updated
in the same change.
Two findings from the GitHub review bot on the ready PR:
The tsdown two-entry build emitted the shared bootstrap module as a
lib/bootstrap-*.js chunk imported by both bundles, which the package.json
files whitelist (deliberately exact) omitted — a packed install had
dangling imports. The package now runs two single-entry builds, so each
bundle inlines its own bootstrap copy and every shipped file is
self-contained.
prepareValue admitted any cloneable value whose BOUNDED inspect rendering
fit maxValueBytes, so a huge container with a compact rendering (a
50k-element array renders as '... N more items') crossed the port raw,
bypassing the cap on both sides. The cap now measures the value's real
cross-boundary size — exact bytes for strings, the structured-clone wire
size (v8.serialize) for everything else — and oversized containers cross
as their bounded rendering instead.
- Add client.onclose handler to unregister tools when the MCP server
disconnects (crash or intentional close)
- Replace bare `let disposers` with a shared mutable state object so the
effect disposer and notification handler always reference the current
generation
- Serialize tools/list_changed resyncs with latest-wins coalescing
(syncing + pendingResync flags) to prevent concurrent races
ds-review-bot flagged that a typo'd or stale config key (e.g. web_fech for
web_fetch) silently applies the timeout to nothing — the tools/execute lookup
just never matches. Mirror dsh-tool-subagent's lifecycle-driven handling of a
configured-but-unregistered provider: on every tools/change (and once at load),
logger.warn each configured name still absent from ctx.tools, warning each name
at most once so a late registration silences it. Not a load-time throw — the
tool set is dynamic (cordis.yml load order, HMR), so a real tool may register
later.
Declare inject = ['tools'] since the plugin now reads ctx.tools synchronously
in apply (previously only inside event callbacks). Regenerate config-catalog
(Requires: tools) and event-producer-consumer graph.
The host's message listener trusted the compile-time WorkerToHost shape on
traffic from a peer that runs model code: postMessage(null) threw in the
listener and crashed the host process; forged log/done messages bypassed
maxLogBytes/maxValueBytes (the worker-side LogBuffer and prepareValue cap
only honest flows); and the error-reply renegotiation re-echoed a forged
non-cloneable call id, throwing outside any catch.
Every inbound message now passes a runtime shape gate that validates and
REBUILDS it field by field (junk drops without a throw; call ids must be
numbers, so replies are always clone-plain; forged extra fields never ride
along). One host-side ledger bounds everything landing in logs — honest
port entries, forged ones, and stray pipe bytes — at the single documented
maxLogBytes, with the shared in-band truncation marker emitted host-side
when the ledger trips first; the completion value is re-capped host-side
through the same prepareValue (with exactly the truncation suffix as slack
so honest worker-capped values pass unchanged), and done error text is
bounded. Also folds the stray-capture budget into that shared ledger
(round-1 finding B: it was a second maxLogBytes on top of the documented
shared cap).
The shipped backend of the code-execution seam, per the Code Mode RFC's
worker-thread section: one fresh Node worker per run, executing the
model's TypeScript after a host-side type-strip (wrapped in an
async-function shell so top-level return/await parse, sliced back out
position-preserved), bindings bridged over the message port under
hostile-peer rules (own-property name lookup, at-most-once replies,
post-settlement drops, null-prototype namespaces), logs streamed eagerly
with an in-band truncation marker, and two independent budgets — measured
event-loop busy time (computeMs) plus a never-pausing wall ceiling
(maxWallMs) — funneling into worker.terminate(). env: {} and execArgv: []
keep the isolate hermetic; disposal aborts in-flight runs and awaits
worker exits.
The worker entry loads unbuilt via Node's native type stripping
(src/worker.ts, erasable-only) and ships built as a sibling tsdown bundle
(lib/worker.js); tests/built-lib.e2e.ts pins the built load path under
plain node and joins the built-artifact smoke gate. Unit suites cover the
bootstrap in-process (fake port) and the runtime over real workers,
per-file 100%.
Codex flagged that the load-path smoke disposed the fiber only at the end,
so a leaked stale tools/execute wrapper would still pass. Add an explicit HMR
test: after fiber.dispose(), a configured tool receives the caller's own
signal unwrapped (the derived deadline is gone), matching the repo's
"dispose must reach quiescence" rule.
Model-facing tool-call budgets were tangled into each capability's schema
(bash timeoutMs, web_fetch timeout_ms) with no shared home. Add a
tools/execute around-dispatch waterfall to dsh-tools whose base next() is
the dispatch-with-normalization thunk, and a new @deepseek-ai/dsh-timeout-policy
plugin (packages/timeout/) that arms a per-tool deadline on exec.signal and
returns a structured TOOL_TIMEOUT when it wins. Migrate web_fetch (drop the
model-facing timeout_ms) and web_search onto it; the fetch provider keeps its
timeout only as a resource backstop for direct callers. bash and hook command
execution keep BASH_TIMEOUT unchanged.
Named the plugin timeout-policy (not the RFC's tool-timeout) so it does not
trip the gen-tool-catalog packages/*/tool-* completeness guard, and replace
exec.signal by in-place mutation before next() since cordis waterfall next()
ignores passed arguments. RFC moved to implemented/ recording both deviations.
tianyicui's review: the seam name did not say what the event or its
types do. 'advice' reads both ways — advisory content for the model,
and AOP before/after advice woven around a join point (here the
derived history) without modifying it — so RequestAdvice.before/after
are self-describing. Types follow: RequestAdvice / RequestAdviceContext;
the logged EpochHeader fields keep their positional names
(messagePrefix/messageSuffix).
Also sharpens the core.md wording the review flagged as ambiguous:
before-advice sits in front of the ENTIRE derived history, directly
after the system slot (the conventional home for session-stable openers
— an AGENTS.md digest, a skills catalog), after-advice follows the
history's last message. Catalogs and doc graphs regenerated.
A client-callback throw only becomes a JSON-RPC error RESPONSE to the
agent's session/request_permission — runScenario itself kept going, so a
tolerant agent could treat the error as a denial and the scenario would
pass, or worse, record: the impossible click baked into fixture and
golden, green on every replay. The mismatch is now captured as a harness
error while the agent is answered plain cancelled (a well-defined path
it cannot reinterpret), and the step loop rejects the run on it as soon
as the in-flight step settles. The spec asserts the rejection instead of
the agent-side error echo.
InputScript gains an optional permissionAnswers queue, consumed FIFO by
the harness's requestPermission handler. Each entry selects by option
KIND (allow_once, reject_once, …): option ids are agent-issued randoms a
committed script cannot know, while kinds are the ACP-stable vocabulary,
so the client maps kind → the offered optionId at answer time. An absent
or exhausted queue answers cancelled — existing scenarios and goldens
are untouched — and a scripted kind the request never offered throws,
surfacing as a JSON-RPC error on the permission request: the scenario
scripted an impossible click.
This is what lets an approval-flow suite (the sandbox composition) drive
allow/reject round-trips deterministically from input.json, per the
shared-acp-snapshot RFC.
A scripted fake ACP agent bin (tests/fixtures/fake-acp-agent.ts) speaks
real newline JSON-RPC through the REAL runScenario spawn path (tsx
loader, temp cwd, env plumbing); every behavior — prompt outcome,
session/new rejection, persisted logs, filesystem noise — comes from a
behavior.json beside the fixture, so specs script whole subprocess runs
from data. harness.spec.ts drives every step op, both expect-error arms,
the permission-stub default, env forwarding, workspace seeding, and the
harvest ordering/noise/fallback branches. suite.spec.ts runs the factory
for real at collection time: a replay suite over committed synthetic
fixtures and a record suite over a temp copy (write-back never touches
the committed tree; ACP_SNAPSHOT_SPEC_BOOTSTRAP=1 re-bootstraps it),
plus direct cases for the exported pure helpers. The suite factory's
pure helpers (childFixturePaths, fixtureContext, normalizedHeaders,
headerDeltaCount) are exported for those direct specs.
Two branches carry justified v8 ignores, both structurally unreachable:
the waiter in-bounds guard (noUncheckedIndexedAccess) and waitForExit's
already-exited race guard (both call sites sit one synchronous frame
after stdin.end()/kill()). The fake bin substitutes the session/new cwd,
not process.cwd(), into scripted logs — the realpath difference
(/private on darwin) is exactly what the real bin's header carries.
packages/support/acp-snapshot/src is at 100% statements, branches,
functions, and lines under the per-file gate.
The snapshot tier's machinery leaves examples/acp-agent/tests for
packages/support/acp-snapshot (@deepseek-ai/dsh-acp-snapshot), where the
coverage gate measures it and a second example can consume it instead of
forking it: harness.ts (runScenario, parameterized by an AgentUnderTest
{binScript, configPath, tsconfigPath} instead of module constants),
normalize.ts (moved verbatim), and suite.ts (defineAcpSnapshotSuite — the
per-scenario golden/log compares, record write-back, per-suite header pin
with its uniformity guard, and the fixture guard block, lifted from
acp.snapshot.ts). The example file collapses to its scenario table plus
one factory call; env reading (DSH_SNAPSHOT) stays at that edge.
The exactly-one-pin meta-test generalizes from the hardcoded text-turn
name to "exactly one per suite" — which scenario pins is the scenario
table's reviewable choice (per-suite pinning per the proposal RFC).
Extraction parity: pnpm run test:snapshot is 36 passed + fs-policy-reject
failing BEFORE AND AFTER (BSD-sed environment failure, reproduced at the
base commit in a clean worktree — the recorded golden's sed -i syntax is
GNU-only), with zero byte changes under examples/acp-agent/tests/snapshots/.
Coverage for the new src files lands in the next commit.
New group packages/code-runtime/ with the interface package
@deepseek-ai/dsh-code-runtime, per the Code Mode RFC: abstract CodeRuntime
service (run() resolves program failures as an error field, rejects only
for seam misuse), the CodeRunRequest/CodeBindingNamespace/CodeRunResult/
CodeLogEntry/CodeRunFailure vocabulary, and readonly language/isolation
backend descriptors. Registered in the tsconfig maps, packages/README,
architecture service map, and the doc-graph service-role classification;
catalogs regenerated.
The RFC's one forward path token to the worker package becomes an npm-name
mention until PR3 creates that directory (verify-package-paths is
drift-scoped: the now-existing group made the token checkable).
docs/architecture.md ceiling 1630 -> 1640: the doc gained a genuinely new
capability-service row; the row itself is already minimal.
The Date.now()/Math.random()/argless-new-Date() bans existed solely to
keep scripts resume-compatible for the deferred journaling/resume
feature. Pre-support for an unimplemented feature is speculative cost:
scripts may now read the clock freely; implementing resume reintroduces
the bans as a script-contract tightening. The RFC's shipped-state
description is updated in place, the tool DESCRIPTION drops the
constraint sentence (the pinned text-turn header follows), and the
engine README's trust-premise paragraph now leans on absent globals
alone.
Connects to an external MCP server and registers its tools on
ctx.tools. Supports stdio (child process) and Streamable HTTP
transports. Credential-shaped env vars are scrubbed before forwarding
to child processes.
- Plugin lifecycle: connect, sync tools, re-sync on ToolListChanged,
dispose unregisters and closes
- Full JSDoc on all exports (@param/@returns on functions)
- 100% per-file coverage (apply lifecycle, args coercion, env scrubbing)
- Config catalog regenerated
