P1: both merge parents shipped SCHEMA_VERSION=3 for different layouts (surface
columns vs seed_length), so an on-disk 3 was ambiguous and wrongly accepted.
Bump to 4 (merged layout) so the version check rejects both sibling v3s.
P2: a surface-eligible event with no surfaceOp lands in the log but vanishes
from deriveMessages() (surface is the sole derivation path). The typed append
overload enforces the marker only when the type arg is a literal; it collapses
to optional when widened to the union (a caller iterating raw events). Guard at
runtime in both append() and the seed constructor — no backward-compat for
surface-less logs. Shared seed fixtures carry surfaceOp explicitly and the
appendLog helper forwards it verbatim (no synthesized default). Exports
isSurfaceEligibleType. Regression tests for all three, each verified to fail
on the unfixed code.
Gates: typecheck, test (1115), snapshot (14), doc-sync, lint, build, hygiene green.
The seed-boundary change made fork-child replay route correctly but shipped
with no recorded fork scenario — the seedLength slice was exercised only by
llm-replay unit tests and a persistence round-trip, never by the full-transcript
snapshot tier. Add two recorded scenarios that drive a real fork child through
it:
- subagent-fork: parent completes a turn, then forks one child (child fixture
carries a non-zero seedLength, the boundary the replay slice consumes).
- subagent-mixed: parent completes a turn, then delegates once via spawn
(seedLength 0) and once via fork (non-zero seedLength) in one transcript —
the first scenario to drive two subagent backends at once, exercising both
branches of the slice.
Both need a completed turn-1 so the fork seed is a non-empty completed-turn
prefix (a turn-1 fork seeds empty = spawn, which would not exercise the slice).
Removing the slice turns both scenarios red (the fork child receives the
parent's recorded chunks), proving the guard bites.
ACP (out-of-process) subagent replay remains a different shape, still tracked
as TODO(acp-subagent-replay).
A fork subagent seeds its child session with a prefix of the parent's log, and
that seed becomes the child's persisted log — so a fork child's .jsonl begins
with the PARENT's events, including the parent's assistant/chunk events. The
snapshot replay harness derived a child's script from its whole log, which would
replay the parent's recorded responses as the child's model calls. Spawn-only
scenarios never hit it, but a fork snapshot would mis-route silently.
Record the seed boundary and skip the inherited prefix at replay:
- SessionHeader gains an optional `seedLength` (how many leading events were
inherited via a seed), threaded through CreateSessionOptions/CreateAgentOptions
meta and stamped by the fork backend (= seeded-prefix length; absent for spawn).
It is EXPLICIT, never inferred from seed.length: a resume seeds the whole stored
log, so the resume path passes the persisted boundary back.
- Both persistence backends round-trip it: JSONL header line, SQLite seed_length
column. The SQLite table change bumps SCHEMA_VERSION 2->3; per the pre-release
stance the backend rejects an older user_version on open with NO migration.
- llm-replay's parseSessionHeader reads seedLength and loadSessionScripts derives
a child script from events AFTER the boundary. seedLength is 0 for spawn, so
spawn replay is byte-for-byte unchanged.
Closes the routing-correctness gap the per-session snapshot replay RFC under-
stated; a recorded fork scenario remains a future addition but now derives
correctly. RFC: docs/rfc/implemented/testing/2026-06-22-fork-child-replay-seed-boundary.md.
Regression coverage: a fork child fixture whose seeded prefix carries a parent
chunk (derived script must exclude it, proven red without the slice); a seedLength
persistence round-trip through the shared coordinator contract (both backends);
the fork backend stamping it; resume preserving it from the persisted header.
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.
The createdAt+recordedId child sort comment over-claimed "tie-safe". Codex
flagged that a same-millisecond sibling tie would be broken by random session
id, which does not recover first-call order. In the current synchronous cut that
tie is unreachable — the subagent tool awaits one child's result and disposes it
before the parent starts the next, so siblings' createdAt values are strictly
ordered and match first-call order. Restate the comment to that real invariant
(at both the replay sort and the harvest sort), note that the id tiebreak only
makes a degenerate collision deterministic, and flag the concurrent-subagent cut
that would need a real first-call ordinal with XXX(concurrent-subagents). The RFC
records the same limitation. Comment/doc only — no behavior change.
The snapshot tier was built single-session: dsh-llm-replay served calls from
one global positional cursor, and the harness harvested one session log. A
subagent runs as a second agent with its own session, so a parent→child
scenario could neither replay deterministically nor harvest the child's log.
This resolves the TODO(subagent-snapshots) deferral from the subagent RFC.
- Stamp the calling session id onto the model request: GenerateOptions.sessionId
(typed Branded<'SessionId'> to avoid the dsh-llm↔dsh-session cycle), set by the
agent loop from agent.session.id. Adapters ignore it; an llm/stream listener
routes by it.
- Key replay per session: dsh-llm-replay loads the parent log plus one per child
(childFiles / $DSH_SNAPSHOT_CHILD_FILES), derives a script per recorded session,
and binds each live (freshly-random) session to a recorded script by first-call
order — parent first (earliest createdAt, first to stream). Keys by WHO calls,
so it survives a future concurrent/backgrounded subagent; a global cursor would
not. An unrecorded extra session fails loud.
- Harvest every log: the harness collects all .jsonl across cwd buckets, ordered
primary-first (top-level, then children by createdAt), and RunResult exposes the
plural sessionLogs. The spec writes each back on record (session.jsonl +
session.<n>.jsonl) and diffs each against its fixture on replay.
- Wire the subagent seam + spawn + fork + tool into the acp-agent example (both
cordis configs) and add two nested scenarios recorded against the real API:
subagent-spawn (parent + 1 child) and subagent-multi (parent + 2 children, 3
sessions). Both replay keyless in the default gate.
A new RFC documents the design (docs/rfc/implemented/testing/). Single-session
replay is unchanged (a call with no sessionId is one anonymous primary session).
TODO follow-up: a dedicated branded-ids package could own the SessionId brand and
dissolve the cross-package cycle note; out of scope for this testing PR.
The required-fixture-guard description still said session.jsonl was needed only
for model scenarios, but the harness passes <dir>/session.jsonl to llm-replay
unconditionally, so loadReplayScript() fails for a no-model scenario without it.
The code already requires it for all scenarios; align the RFC prose.
Implements docs/rfc/.../2026-06-20-extract-example-app-packages.md. Each
example was thick — a hand-rolled start.ts, an infra preamble, nested
base.yml/base-core.yml/acp-tail.yml includes, and a coupled front-door
cluster enforced only by prose. This moves the composition into packages so
each example is a thin leaf cordis.yml: pick the swappable backends, load one
app package.
New packages:
- @deepseek-ai/dsh-agent-core (packages/core/agent-core): one bundle plugin
that loads the providerless/executor-less/UI-less spine (timer + llm +
sessions + system-prompt + tools + agents + invariants + tool-bash +
agent-loop) via ctx.plugin(...) inside apply(), and forwards agent-loop's
`agents` list as its own Config (export const Config = AgentLoop.Config,
default []).
- @deepseek-ai/dsh-stdio-agent (packages/ui/stdio-agent): terminal chat APP —
agent-core + console logger + readline UI + a pre-created `main` agent, with
a bin. The demo:echo/coding front door.
- @deepseek-ai/dsh-acp-agent (packages/ui/acp-agent): ACP server APP —
agent-core + JSONL persistence + the acp bridge, NO stdout logger, with a
bin. The stdout-purity footgun is structurally unreachable from the leaf.
Amendment to the RFC: hmr stays a LEAF cordis.yml entry, not baked into
dsh-stdio-agent. hmr is a Loader-only dev plugin (throws without
--expose-internals; the in-process test tier can't even import its decorator
form), so a package statically importing it could never carry the per-file
coverage gate. Unlike the console logger, a stray hmr is not a stdout-purity
footgun, so leaving it at the leaf costs no safety. With hmr out, all three new
packages carry in-process unit specs at 100%.
Boot glue (Loader tail, .env load, snapshot-mode selection, stdin-dispose
lifecycle) moves into each app's bin; start.ts and base.yml/base-core.yml/
acp-tail.yml are deleted. Each app package gets a keyless real-load-path test
that boots through its bin + the cordis Loader (guarding the unwrapExports
export-shape bug class, postmortem 0001). ACP snapshot replay stays green
against the existing committed goldens (pure boot restructuring). RFC moved
proposed->implemented with the amendment recorded; package/example/architecture
docs and the module graph updated.
The dsh-llm property-suite bullet claimed the suite checks an ordered-prefix
contract (the blocks push() returns incrementally are a prefix of final
blocks(), in order) and streaming-vs-one-shot agreement on usage/finish. Both
died with flushReady()/flushRemaining()/generate()/streamBlocks(): the
ordered-prefix guarantee was provided by that flush pair, and push() never
guaranteed it (index 0 opened by a delta then index 1 closed by block-end has
push() return block 1 while final blocks() orders [0, 1] — the returned block
is not a prefix).
Rewrite the bullet to enumerate only what properties.spec.ts actually asserts:
blocks() count <= distinct indices, idempotent re-assembly with message().content
mirroring blocks(), blocks() never throwing and yielding valid tags, and finish
reflecting the last finish chunk (defaulting to stop).
Model-driving ACP snapshot scenarios shipped both session.jsonl (the
replay fixture) and session.golden.jsonl (the expected re-persisted log).
For recorded scenarios the normalized fixture and golden were byte-identical
— pure duplication. Remove session.golden.jsonl entirely: every model
scenario now has at most one committed session-log artifact, session.jsonl,
which doubles as the replay source AND the expected produced log.
The snapshot test compares the replay run's persisted log against the
session.jsonl fixture, normalizing BOTH sides — but each against its OWN
volatile values, not a shared context. A raw harvested fixture bakes in the
recording run's session id / cwd / timestamps, distinct from the live replay
run's; since normalizeSessionLog scrubs cwd by exact string match, the
fixture must be normalized against its own header (new fixtureContext helper)
or its stale recorded cwd would leak unscrubbed and the compare would fail.
The session side uses a normalized-string toEqual, NOT toMatchFileSnapshot,
so a run never overwrites the fixture.
Authored override scenarios (error-finish, cancel) now hold their expected
produced log in session.jsonl. Verified llm-replay ignores the fixture for
model chunks when an override exists: loadReplayScript() returns the override
array and never reads config.file, so committing the full expected log there
does not affect replay behavior.
The required-fixture guard is now per-kind: every scenario needs input.json +
stdout.golden.jsonl; model scenarios need session.jsonl; authored ones
additionally need replay.override.json. Updates the ACP-snapshot-tests RFC to
the reduced fixture set and moves the proposing RFC proposed -> implemented.
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.
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
Merge brought in the RFC-classification reorg and two new doc gates;
rewrite every drifted packages/<name> cross-link (Markdown link targets,
moved-README relative depths, and .ts comment paths) to the grouped paths.
Add two doc-sync/hygiene gates so the manual checks this restructure
needed become automated:
- verify-package-paths.ts: flags a packages/<path> reference (in Markdown
or a .ts comment/string) that does not resolve AND names a real package
in a segment — i.e. a stale path to a MOVED package. A path naming a
non-existent package (a forward-looking proposal) is left alone, so it
applies uniformly across proposed/implemented/rejected.
- check-workspace-constraints: assert the packages/<group>/<pkg> depth-2
shape (group dirs carry no package.json; no flat or over-nested
packages). Group names stay open; only the shape is fixed.
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.
Hypatia May
Tianyi Cui