Implements the approved simplification Agent Note: sse.ts now pipes the
response body through TextDecoderStream and EventSourceParserStream
(eventsource-parser/stream) and keeps only the DeepSeek protocol shim —
yield each event's data, terminate on [DONE], throw
LlmError('STREAM_CLOSED') on EOF without the sentinel. The SSE
spec-conformance tests are deleted; sse.spec.ts pins only the
[DONE]/STREAM_CLOSED/EOF contract, including the new spec-strict verdict
that an unterminated trailing event is truncation (the old parser
flushed it — a robustness nicety no real provider shape needs).
eventsource-parser@^3.1.0 becomes llm-deepseek's second runtime
dependency (already in the lockfile transitively via the MCP SDK).
Docs: the Agent Note moves proposed/ → implemented/ and is rewritten per
the lifecycle contract; the rejected NIH roll-up note's inbound links
follow. The twin-adapters note, dsh-llm LlmAdapter JSDoc (and its
type-equiv fences), cookbook, group/package READMEs, root AGENTS.md
layout line, sdk-helper comments, and the regenerated config catalog
drop the "hand-rolled fetch + SSE" claim in both languages; all eight
touched pairs re-recorded.
Validate replay sidecars and cross-copy failure facts, make browser console tripwires and macOS temp paths deterministic, and wait for asynchronous TUI resume details. Keep the owning docs, translations, and generated catalog aligned.
Both ds-review-bot findings were real:
- markLlmAdapterFailure's carried-facts cross-check read error.code
directly; a foreign Error with a valid own failure payload but a
throwing code accessor would replace the original adapter error with
the accessor exception, breaking the error-identity guarantee. The
read now goes through foreignErrorCode(), which contains the trap and
falls back to the normalized snapshot (test: hostile code accessor
beside a valid failure payload -> original identity kept, UNKNOWN
facts).
- live-interactions' afterEach caught scaffold.close() into undefined,
silently disabling ReplayHandle.assertConsumed() — the fixture-drift
tripwire — and hiding cleanup defects. Teardown now runs every step,
collects failures, and rethrows (AggregateError when several).
Downstream packages lost the tests that exercised their agent-loop-facing
edges when the loop was rewritten. Restore 100% per-file coverage with
behavior tests through public seams: llm-retry config validation and
cancellation races, goal replay drift/staleness/teardown edges, plan-mode
disposed-flush, workspace-context empty-change commits, api-proxy
synchronous send failures, acp-snapshot spill-path extraction and refresh
write-back, ACP injection-triggered turns, cli-demo and tui inbox
lifecycle edges, and agent-loop retry/settlement/lifecycle branches.
The only source changes are narrowly-justified v8 ignore annotations on
invariant guards and one redundant-guard removal (workspace-context).
markLlmAdapterFailure gated the own-`failure` data property on instanceof
HarnessError, which drops the validated facts exactly when class identity is
lost — two copies of this package in one process (e.g. a source-plane replay
harness throwing into a lib-plane boot) make the replay-thrown LlmError's
SERVER/AUTH code arrive as UNKNOWN and defeat llm-retry's retryable-code
match. The snapshot is already validated field-by-field and cross-checked
against the error's own code, so honor it on any Error.
- dsh web gains --provider/--model: a non-deepseek provider mounts the
matching pi-ai catalog route (ambient credentials), making image input
reachable from the shipped Web assembly; requires an explicit --model
- attachment-local syncs the publication directories after the hard-link
publish so a reported durable reference survives a crash (POSIX; Windows
relies on filesystem metadata journaling)
- the attachment seam gains storage-free validateImage; the host validates a
complete multi-image prompt before persisting any member, so one malformed
image cannot strand valid members as unreferenced objects
- startSession sends before navigating: a rejected first send keeps the empty
state, its error strip, and the complete draft mounted
- the webserver rejects an undeclared-length body the moment it crosses the
configured limit instead of draining a potentially endless stream to EOF
- regenerate docs/module-graph.md for the attachment packages
- pack @deepseek-ai/dsh-attachment in the packed-install e2e closure so npm
resolves the new dsh-llm peer from the tarball set instead of the registry
- gate the pi-ai anthropic e2e profile strictly on ANTHROPIC_API_KEY: the
DeepSeek endpoint does not serve anthropic-messages, so the DEEPSEEK_API_KEY
fallback turned the keyless skip into a 404 across the whole suite
A well-formed provider stream that ends with finish_reason stop and zero
content blocks previously became a successful empty assistant message: the
turn completed silently, and drivers like goal-session counted the no-op
round. Both adapters now map that degenerate completion to a finish
{kind:'error'} with the new canonical EMPTY_RESPONSE code from dsh-llm, and
dsh-llm-retry adds the code to its default retryable set, so the existing
closed-step recovery path retries it and fails loud once the budget is
exhausted.
Covered by adapter unit tests, an llm-retry default-policy test, and a new
authored keyless ACP snapshot (empty-response-retry) with a deterministic
1 ms zero-jitter retry overlay.
Tianyi Cui