Every PR that changes product-user-visible GUI behavior now includes a
demonstration GIF with real provenance (that branch's built tree, real
key, real model rounds). Recording stays side-effect-free; the skill
gains a bounded final publication step: GIFs go on an append-only
orphan assets branch (one per PR series) and embed via the blob URL
with ?raw=true, never on the PR branch itself.
Folds in the operational lessons from the Code Mode UI series:
.playwright-mcp/ screenshot roots (now gitignored), per-PR staging and
precise server teardown, one-call DOM polling for transient states,
exact-text completion predicates, prompt engineering for UI states,
and the export-before-invoke GIF_SKILL_DIR encoder pitfall.
Agent Note: implemented/process/2026-07-26-gui-pr-gif-evidence-and-assets-branch (+ zh pair); the 2026-07-23 recording note now defers publication policy to it.
Implements the gate-consolidation Agent Note from the NIH dependency audit:
- Shared markdownFences helper in scripts/markdown.ts (mdast code-node visit);
doc-typecheck and verify-type-equiv extract fences through it; md-fences.ts
and the duplicated extractEquivBlocks regex scanner are deleted;
markdownProseLines derives fenced lines from parsed code-node positions
instead of a second fence regex.
- publint-all.ts and verify-built-package-invariants.mjs parse argv with
node:util parseArgs instead of hand-stepped parseOptions copies.
- Five straggler readdirSync walks become globSync: verify-runtime-closure,
dev-web discoverPluginDirs, verify-package-paths realPackageNames,
verify-client-domain-graph listSources, publint-all addPath. The
dirent-diagnostic walks in check-workspace-constraints.ts and clean.ts stay.
Behavior parity verified: pnpm run doc-sync and every rewritten gate produce
byte-identical output before and after on this tree.
Moves the owning Agent Note proposed -> implemented and re-records its pair.
- proposed/testing -> implemented/testing with the lifecycle rewrite
(Proposal->Decision in present tense, Acceptance criteria + Risks
folded into Consequences); zh counterpart mirrored and both pairs
re-recorded.
- the rejected NIH-audit roll-up pair now links the implemented/ path.
- loader-smoke README: captured output is bounded by execa's default
100 MB maxBuffer, no longer unbounded.
- acp-snapshot README: harness.ts now also imports vitest (vi.waitFor),
so the vitest-run-only constraint names both modules.
- jsonrpc keyless smoke: raise the invalid-env case's subprocess
deadline to 25s (the 9s pick starved a cold tsx boot on slow NFS).
Convert the hand-rolled corepack + store-path + actions/cache pnpm setups
to pnpm/action-setup@v4 + setup-node cache: pnpm (the landlock-run.yml
pattern) in e2e.yml, docs-pages.yml, pi-ai-provider-e2e.yml,
build-exe-for-python-sdk.yml, and ci.yml's node-compat and benchmark
jobs. The enterprise restore-only PR jobs and the Windows job keep their
documented asymmetries; serial-linux keeps its hand-rolled store cache
because it produces the exact key those restore-only jobs consume, and
only swaps corepack for the action.
Implements and moves the CI-caching Agent Note from PR #667 to
implemented/, recording the serial-linux producer exception and the
exact-key (no restore-keys prefix) behavior of setup-node's cache.
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.
Conflict resolution: scripts/type-equiv.manifest.json takes master's new
paired-derivative format (one primary entry per pair) and re-adds this
stack's CodeDispatchLog entry in that format. zh README pairs brought
along for the dispatch-log arm (spill-policy behavior/limitations bullets,
tools limitation bullet now pointing at the shipped bounding).
The public classifier existed to defend an exported reader against
arbitrary signals, but its only production caller is the loop reading
its own machine-private turn signal, where cancel() is the sole aborter
and always writes one frozen canonical cause. Delete the export and its
15-line structural validation: settle() states the slot invariant with
one cast, the boolean call sites ask signal.aborted directly, and the
retry veto drops entirely because a requested window already implies a
live signal (cancel() retires the window before aborting).
The abort(reason) channel and first-wins semantics are unchanged; only
the reader's publicness is gone, and with it the paranoia it required.
The sandbox note's deferred-phases plan for the Windows chain now points
at the proposed landstrip evaluation gate, so whoever picks up that rung
finds the pending evaluation. The landstrip note itself stays proposed;
this only tracks where the pending decision lives (a permitted
keep-implemented-notes-current edit). Mirrored in the .zh.md and the
pair re-recorded.
Two new keyless scenarios for the functionality master gained since this
lane's base (#644 websettings, #643 workspace browser rework), both zero
model calls:
- settings-chrome: the modal shell (sidebar-foot trigger aria states,
role=dialog, aria-current section switch to the deliberately empty
Models, Escape + close-button paths, dialog aria golden); the Appearance
row as the REAL theme gesture — retiring lifecycle-chrome's
TODO(web-theme-gesture): clicking 深色 runs aria-pressed -> persisted
dsh.theme -> body[data-ds-dark-theme] -> alias-token flip, survives
reload, and 'system' follows the emulated OS scheme both ways; the
Language row switches the settings-scoped copy to English (dsh.locale
persisted, survives reload) and restores zh. Intentional reloads tear
the SSE stream, so the spec drains exactly its own reconnect warnings —
the tripwire still fails on unexpected connection loss.
- workspace-management: create-by-name twice through the region-header
dialog (host-durable via ctx.workspace.list()); rename end to end —
hover-revealed row menu (the button is display:none until the row
hovers), duplicate-name pre-check (inline role=alert + disabled primary
before any wire call), then workspace.rename through the real RPC,
row update, host durability, reload survival; the flat 'In one list'
view (section label flips, group headers drop, dsh.workspace.view
persists across reload, grouped restored); the session hover card
(dwell to open, closes on pointer leave). The one session row reuses
seeded-history's committed seed — no new recording. Deliberately not
driven: the inert menu rows and drag reorder (deferred in the note
with re-entry triggers).
Agent Note gains scenarios 8-9 and the drag-reorder deferred item in both
languages; llm-replay README's zh side catches up with the { patches }
paragraph; pairings re-recorded.
Faithfulness pass from the translation review: 接入 over 搭乘, first-mention
disposition annotation, 原始的 Readable/Writable, 留在原地, 评审 over 审阅;
pairing records re-recorded.
Responding to ds-review-bot round 2 on #662:
- LANG_ALIASES is a Map: an assistant-authored fence label like
constructor or __proto__ now misses (plain render) instead of resolving
an inherited object property and crashing shiki mid-conversation. Test
sweeps the inherited-key labels.
- The singleton is pre-warmed in a deferred task at plugin boot (the
~120-175ms engine+grammar construction long task moves off the first
finalized fence's render); the lazy path remains the correctness
fallback, and unref keeps non-browser imports from pinning the loop.
Agent Note updated (both languages).
Responding to ds-review-bot round 2 on #661:
- logWork is bounded: past maxParallelSubCalls pending shaped-append tasks
the ordered commit lane holds (Promise.race drains one), so a slow spill
backend backpressures the run instead of accumulating unbounded pending
I/O and retained results. Tasks self-remove on settlement; run
settlement still drains every task inside the open turn. New spill test
drives three oversized reads against a hung backend at cap 1 and proves
the third dispatch cannot start until a save drains.
- shapeDispatchLog's catch uses errorMessage() (total), so a thrown value
with a throwing toString cannot escape the containment and lose the
settle event.
- CodeDispatchLog.content documented as the RENDERED result projection
(native tool/result vocabulary), not what the program received — the
program gets the structured value; doc pair + type-equiv re-synced.
- New RECORDED tui-agent snapshot scenario code-mode-dispatch-spill: the
real Loader-visible composition (worker runtime + spill-local + policy)
drives an oversized bash sub-call end-to-end; replay proves the durable
dispatch copy is bounded to preview + locator while the program value
stays whole (the outer result carries just the line count).
Agent Note updated (both languages).
Three review rounds each found a new silent-loss path in the same
wrapper state (dispose racing an in-flight flush, overlapping hints
displacing the retained promise, the provider's fixed 30s flush timeout
rejecting while the processor still drains). Every path exists only
because forwarding the seam's turn-boundary hint to forceFlush() made
this backend the process's second flusher against undocumented SDK
internals from the upstream experimental tree.
The backend now implements no flush(): the batch processor is the only
flusher, its scheduledDelayMillis (already deployment-tunable through
the processor passthrough) governs export cadence, and shutdown()'s
drain is complete by construction. The two race-pin tests collapse into
one dispose-during-in-flight-batch drain pin; the seam's optional
flush() contract now tells implementers they own the concurrent-flush/
shutdown interaction. Removal rationale and the reinstatement trigger
(a stated turn-boundary latency requirement scheduledDelayMillis cannot
meet — and then via the processor's own forceFlush(), never the
provider's timeout-wrapped one) are recorded in the revival Agent Note,
both languages.
Resolutions: regenerate docs/capability-seams.md (the only conflict);
master widened the bilingual pairing scope to package READMEs, so this
merge adds the Chinese counterparts and i18n records for the three
telemetry READMEs and patches packages/README.zh.md with the telemetry
group row.
