Побочные находки инцидента #400 (правка большой таблицы через MCP подвешивает всё).
1. Гонка connect-vs-unload в @hocuspocus/server 3.4.4 (вероятный источник 25s
connect-таймаутов): createDocument проверяет loadingDocuments/documents, но НЕ
ждёт unloadingDocuments -> новое соединение может захендшейкаться на умирающий
Document -> redis-sync идёт по пути 'doc не загружен', провайдер висит до
таймаута. Апстрим (main) не починен. pnpm-патч (инфра как у yjs-патча): в начале
createDocument await in-flight unload (обёрнут в try/catch — отклонённый unload
не отравляет открытие, поведение как до патча), в ОБОИХ рантаймах (cjs+esm).
Тест hocuspocus-unload-race: реальный createDocument с засеянным in-flight
unload -> не грузит пока unload не осел; при откате патча тест краснеет.
2. Двойная перекодировка в onLoadDocument (persistence.extension.ts): хук строил
НОВЫЙ Y.Doc и возвращал его -> hocuspocus делал applyUpdate(encodeStateAsUpdate)
ВТОРОЙ раз (315КБ на каждую холодную загрузку); в JSON-ветке результат encode
выбрасывался (мёртвый вызов). Теперь стейт применяется прямо в data.document,
возврат undefined (hocuspocus мержит только при возврате Doc); мёртвый encode
убран. Содержимое документа не меняется — только меньше encode/alloc.
3. isDeepStrictEqual по 84КБ JSON на каждом store: замерил — 1.32мс на 90КБ
(immaterial, <50мс порога; доминируют fromYdoc+encodeStateAsUpdate). Изменений
кода НЕТ по правилу задачи (dirty-флаг только при material).
closes#401
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
mutate трекает in-flight одним полем inflightReject; наложенный второй вызов
перезаписал бы рехджектор первого -> при disconnect отклонился бы только второй,
первый бы висел до PERSIST_TIMEOUT_MS (20с). В проде безопасно (оба call-site
сериализуют через per-page withPageLock), но это футган на разделяемом примитиве.
Гард в начале mutate (после ready-проверки, до касания inflightReject): наличие
in-flight -> reject нового вызова без порчи состояния первого. Docstring CONCURRENCY.
Последовательные mutate не задеты (localFinish синхронно чистит inflightReject до
резолва). +2 теста: конкурентный второй реджектится, первый цел; последовательные
оба успешны.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Инцидент: агент заполнял большую таблицу десятками table_update_cell, каждый —
полный цикл connect/auth/load/initial-sync/store/unload; часть падала с 25s connect-
timeout, event loop lag до 1.7с. Вариант B: кэшировать живой HocuspocusProvider
per page на серию правок — пока провайдер жив (connections>0), сервер не входит в
store->unload->reload, дебаунс реально коалесцирует записи (N ячеек -> 1-2 store).
Новый модуль collab-session.ts: класс CollabSession (connecting->ready->dead) +
реестр (ключ wsUrl+pageId+token) с idle-TTL/max-age/LRU-evict. mutatePageContent
(collaboration.ts) и mutateLiveContentUnlocked (client.ts, replaceImage) переведены
на acquireCollabSession; one-shot Promise-машина (~360 строк дублирования) удалена.
5 инвариантов (подтверждены внутренним ревью): (1) read->write атомарна — между
fromYdoc и applyDocToFragment нет await; (2) per-edit ack сохранён дословно (гард
connectionLost от false-success при реконнекте); (3) disconnect=смерть сессии (без
авто-реконнекта, in-flight реджектится теми же текстами ошибок); (4) изоляция
identity (токен в ключе); (5) валидация при reuse. replaceImage работает под
внешним page-локом без дедлока (acquire лок не берёт). Тексты ошибок == develop.
Env: MCP_COLLAB_SESSION_IDLE_MS (0=выкл кэш, точное легаси), _MAX_AGE_MS, _MAX_ENTRIES.
Teardown: destroyAllSessions обвязан в stdio (exit/SIGINT/SIGTERM) + реэкспорт из
index для встраивающего хоста.
closes#400
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- text-normalize.ts: closestBlockHint был вставлен между docstring'ом
stripInlineMarkdown и его определением -> docstring осиротел. closestBlockHint
перенесён ПОСЛЕ stripInlineMarkdown, каждый docstring снова примыкает к своей
функции. Поведение не менялось (только порядок объявлений).
- comment-anchor.ts: header-коммент завышал маршрутизацию — countAnchorMatches НЕ
зовёт resolveAnchorSelection, у него своя параллельная реализация exact-wins.
Коммент уточнён: can/get/apply идут через resolveAnchorSelection, count держит
свой счётчик-примитив, синхронный с ним; обе реализации exact-wins должны
держаться в синхроне при правках.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Ревью #426: секция «How tool calls are stored — READ THIS» всё ещё утверждала,
что единственные ключи элемента — toolName/input/output и «нет error», хотя этот
же PR добавляет error и подробно описывает его ниже. Заглавный абзац приведён в
соответствие: error — возможный ключ для брошенных ошибок на строках после #407;
подсчёт инвокаций и пайринг учитывают error как парный результат.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
createComment — топ-хотспот ошибок агента (промахи по якорю, слепые ретраи).
Портированы аффордансы самокоррекции из editPageText:
- Closest-block hint: общий хелпер closestBlockHint вынесен в text-normalize.ts
(json-edit.ts теперь тоже его зовёт), подключён во все 3 throw-а createComment.
- Markdown-strip fallback в comment-anchor.ts согласованно по всем 4 функциям
(can/count/apply/get) через единый resolveAnchorSelection: exact-verbatim wins
глобально, stripped — только если raw не якорится нигде; soft warning как в
editPageText. Инвариант уникальности suggestion (0/1/>=2) сохранён: raw-unique
никогда не запускает fallback -> не может стать ambiguous. Хранимый selection
остаётся СЫРОЙ подстрокой документа (strip только для поиска).
- Multi-block detection: явное сообщение 'selection spans multiple blocks' когда
per-block поиск провалился, но выделение есть в объединённом тексте блоков.
- tool-spec createComment: копировать selection дословно из getPage/searchInPage.
Известное мелкое ограничение (нит внутреннего ревью): детектор multi-block
использует raw selection, поэтому markdown-стилизованное выделение через границу
блоков получит generic-подсказку вместо spans-multiple-blocks (редко, guidance
всё равно корректный).
closes#408
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
В ai@6 упавший тул — это tool-error часть в step.content ({type,toolCallId,
toolName,input,error}), а не элемент toolResults. Раньше serializeSteps писал
только toolCalls+toolResults (ошибка терялась, orphan tool-call без результата),
а assistantParts эмитил заглушку 'Tool call did not complete.' (реальный текст
ошибки терялся для мультиходового реплея — модель не знала, почему упало, и
повторяла ошибку).
- StepLike расширен полем content; новый хелпер normalizeToolError (Error/string/
object -> строка, обрезка через существующий compactValue/лимиты).
- serializeSteps: на каждый tool-error пушит парный {toolName, error} тем же
паттерном, что успешный {toolName, output} -> колонка tool_calls фиксирует сбой.
- assistantParts: при наличии tool-error эмитит output-error с РЕАЛЬНЫМ текстом;
заглушка остаётся только для по-настоящему непарных вызовов (прерванных).
- docs/reading-ai-logs.md обновлён под новую форму + cutover-оговорка.
Обратно совместимо: старые строки читаются как раньше, error-элемент аддитивен.
closes#407
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Убран отдельный столбец-маркер .definitionMarker (order:-1, min-width:1.5em),
дававший висячий отступ. Номер сноски теперь рисуется инлайн в начале первого
параграфа через .definitionContent > :first-child::before из CSS-переменной
--footnote-number (в модель документа не попадает, экспорт не затрагивает).
Кегль сносок уменьшен до var(--mantine-font-size-sm). Инвариант #146 (contentDOM
первый в DOM) и логика мульти-бэклинков #168 сохранены.
closes#420
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Read-tool outputs were compacted at a 4000-byte gate before being stored
in metadata.parts, which is re-sent to the model every later turn. Whole-page
reads (tens of KB) got shrunk to a 500-char preview plus a "[truncated N chars]"
marker; on the next turn the model read that marker as a source truncation and
re-read the page, wasting tokens and producing wrong behavior.
- Raise MAX_TOOL_OUTPUT_BYTES 4000 -> 200_000 so normal reads are stored and
replayed verbatim; only a single >200 KB output is compacted as a backstop.
- Reword the inline string marker so it reads as a replay-history elision, not
a source truncation, and tells the model it can re-call the tool.
- Update doc comments and enlarge the compactToolOutput unit-test inputs above
the new 200 KB gate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Promoting insertFootnote/insertImage/replaceImage into the in-app
DocmostClientLike interface requires their mirror in the drift-guard's
HOST_CONTRACT_METHODS list (the bidirectional deepEqual — 40 vs 43 — was
red, exactly what the guard exists to catch). Added the three names and
corrected the header comment: they were MCP-only but are now in-app-consumed
and tracked; only deleteComment/updateComment remain untracked MCP-only.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The Researcher role wrote 40 literal `^[...]` and zero real footnotes: its
incremental write path (insertNode/editPageText) doesn't parse markdown, and
the footnote-capable tool was MCP-only. Promote three tools from inline
MCP-only to the shared registry so the in-app agent gets them too.
- tool-specs.ts: insertFootnote/insertImage/replaceImage added to
SHARED_TOOL_SPECS (mcpName/schema/description moved VERBATIM from the inline
registrations — MCP names + behaviour unchanged for external clients).
- index.ts: the 3 inline registerTool calls become registerShared; drop the
"MCP-only by design" comments.
- ai-chat-tools.service.ts: register the 3 in-app via sharedTool ->
client.insertFootnote/insertImage/replaceImage (imageUrl->url,
attachmentId->oldAttachmentId mapping).
- tool-tiers.ts: insertFootnote -> core (else the original asymmetry recurs —
footnote tool hidden while editPageText is core); images -> deferred.
- research/{en,ru}.yaml FOOTNOTES: `^[...]` parses ONLY on a whole-markdown
write (create/update/import); for a pinpoint citation to existing text use
insertFootnote; via editPageText/insertNode it stays literal.
- json-edit.ts guardrail: an edit_page_text `replace` containing a `^[...]`
token is refused into failed[] with an insert_footnote hint, mirroring the
existing formatting-marker refusal. (Slightly broader net than that mirror —
a literal `^[a-z]` regex class in a replace is also refused; accepted
defense-in-depth, has a no-false-positive test.)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add detailed "PROSE, NOT NOTES" instructions to the English and Russian researcher role bundles, clarifying report writing standards. Update the researcher role version to 8 in the index and content-hashes files.
Follow-up to the merged MCP-hang fix (16b476a2); post-hoc review DO. The
fix itself is unchanged and prod-working — this adds the coverage + one
coherence fix the review asked for.
Tests (ai-chat.service.setup-abort.spec.ts):
- onLateResolve late-close: a toolset that resolves AFTER the setup race
was lost has its leased clients released (close spy asserted).
- pure 60s deadline (signal NOT aborted): the turn proceeds Docmost-only
(reaches streamText, run NOT finalized 'aborted') — the defense-in-depth
backstop, previously untested.
- legacy no-runId: a setup abort does NOT re-throw (the `runId &&` guard);
together with the deadline test this locks both halves of the catch guard.
Coherence (external-mcp/mcp-clients.service.ts buildEntry):
- The per-server connects now run via Promise.all instead of a sequential
for-await, so total build time is bounded by the slowest single server
(~2×CONNECT_TIMEOUT_MS) instead of the sum. At >6 all-timing-out servers
the sequential build exceeded the outer 60s deadline, inverting the
"per-server bound primary, outer deadline is backstop" invariant. Merge
is done in a sequential post-Promise.all loop in server order, so tool-key
precedence/disambiguation, outcomes and client ordering are byte-identical
to the sequential build; each server keeps its own timeout + failure-close.
Comments: the onLateResolve note now says it releases the lease (refcount),
not force-closes transports (the cache owns them); the invariant comment
reflects the parallel build.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace the researcher role instructions in both bundles with the new prompt:
- Budget measured in PAGES READ, not searches; snippets never enter the report.
- The document is working memory: live plan, "Log"/"Open Questions" sections,
hard flush cadence (~8-10 pages), context discipline with re-reads.
- Mandatory CRITICAL REVIEW PASS + BUDGET REMAINDER PROTOCOL (adversarial
verification, primary sources, lateral expansion).
- Source hierarchy, dates/staleness, dead-end handling, inline ^[...] footnotes,
report language/terminology rules, finalization checklist.
ru.yaml carries the text verbatim (Russian report); en.yaml is the English-
adapted mirror (report language + working-section names/examples translated).
Bump researcher role version and refresh the content-hash lock.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rework the researcher role prompt (ru + en bundles):
- Add a "CITING SOURCES INLINE (FOOTNOTES)" section: every non-trivial claim
must carry an inline `^[...]` footnote (the only form this system parses);
explicitly forbid the unsupported `[^1]` reference style.
- Add a HARD CADENCE rule: flush findings to the document at least every 10
searches, reinforced in the WORK LOOP.
- Raise the default search volume: STEP 0 estimate and the VOLUME floor now
default to ~50 searches (15 -> 50), with a carve-out for a single trivial fact.
- Replace the weak "FULL PAGES, NOT SNIPPETS" bullet with a strong rule to open
and read (extract) pages, not just run searches (~2-3 pages per search).
Bump researcher role version and refresh the content-hash lock.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A single NUL (U+0000) in model/tool output (e.g. a truncated multibyte
read of a web page) is rejected by Postgres in BOTH the content (text) and
toolCalls/metadata (jsonb) columns, so it failed EVERY write of the
streaming assistant row ("invalid input syntax for type json") and silently
dropped the turn's content from the DB while the live stream still showed it.
- add stripNulChars: deep-strips NUL from all strings, returns the same
reference when there is nothing to strip (no needless clone)
- apply it at the flushAssistant choke point (covers content + toolCalls +
metadata for the seed, per-step and terminal writes)
- tests: deep-strip, same-reference, end-to-end via flushAssistant
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reduce the default external-MCP silence timeout from 5 min to 1 min and the
overall call timeout from 15 min to 2 min. Update related tests and comments.
Increase Fastify's JSON body limit to 25 MiB (configurable via
HTTP_JSON_BODY_LIMIT) to accommodate large AI‑chat payloads.
BREAKING CHANGE: shorter MCP timeouts may abort long‑running tool calls that
previously succeeded with the older defaults.
DO-1: add an integration test (test/mock/tool-timing-server.test.mjs) that
constructs a real createDocmostMcpServer with a spy onMetric, links a Client
over InMemoryTransport, invokes get_workspace (no input schema, so the wrapped
handler always runs) and asserts onMetric fired with
("mcp_tool_duration_seconds", <number>, { tool: "get_workspace" }). This locks
that the monkeypatch wraps every tool AND labels with the registration name —
which the isolated timeToolHandler unit test does not. Mutation-verified
(mislabel -> test fails). The tool's expected backend failure (ECONNREFUSED)
is tolerated: the wrapper times in a finally on throw too, so the metric fires.
DO-2: reword the mcp.service.ts "zero overhead when disabled" comment to
"negligible overhead" — the registerTool wrapper still runs performance.now()
+ an async try/finally per tool call when onMetric is undefined (the
`onMetric?.()` short-circuits the label build; cost is immaterial at
tool-call rate), so "zero" was literally inaccurate.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
packages/mcp stays free of prom-client/server: it only calls an optional
DocmostMcpConfig.onMetric(name, value, labels?) sink the host provides.
- client.ts: onMetric on the config; fires collab_connect_timeouts_total
once, only inside the 25s connect-timeout callback (the connect-vs-unload
signal). cleanup() clears the timer on every other finish path, so no
double-count.
- index.ts: createDocmostMcpServer monkeypatches server.registerTool (before
registerShared + inline tools are registered) to wrap every handler with
timeToolHandler — times in a finally on success AND throw, re-throws
unchanged, emits mcp_tool_duration_seconds{tool=<registered name>} (bounded
cardinality). Single choke point catches all tools.
- mcp.service.ts: the per-request config resolver injects onMetric ONLY when
isMetricsEnabled(), routing mcp_tool_duration_seconds -> observeMcpTool and
collab_connect_timeouts_total -> incConnectTimeout. Disabled / standalone
(stdio, no onMetric) -> undefined -> zero-overhead no-op.
New node:test unit (tool-timing.test.mjs) covers the wrapper's value/throw
preservation and the standalone no-op. packages/mcp/build/ is gitignored,
not committed (CI rebuilds it).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Extends the #355 perf-metrics registry (all behind the METRICS_PORT hard
gate — nullable instruments, no-op helpers when unset). New families:
- collab_doc_load_duration_seconds{size_bucket} — onLoadDocument timed on
the real DB-load paths only (the already-loaded early-return is skipped).
- size_bucket added to collab_store_duration_seconds; storeDocument returns
the ydoc byte length (reusing its single Y.encodeStateAsUpdate, no second
encode) so onStoreDocument observes size without extra cost.
- collab_docs_open (gauge, read-on-scrape via collect() from
hocuspocus.getDocumentsCount() — never inc/dec'd, so it can't drift) +
collab_doc_loads_total / collab_doc_unloads_total (afterLoad/afterUnload).
- collab_connect_duration_seconds — onConnect->connected, correlated by a
WeakMap keyed on the shared request (leak-free; the current client uses
one socket per document).
- collab_auth_duration_seconds — wraps onAuthenticate (success and failure).
- sizeBucket() shared helper (lt64k|lt256k|lt1m|ge1m, 4 bounded values).
The mcp_tool_duration_seconds histogram + collab_connect_timeouts_total
counter helpers are registered here but wired in pass 2 (MCP callback).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A transient network blip during an external-MCP handshake left createMCPClient
pending forever (@ai-sdk/mcp does not settle on abort). getOrBuildEntry caches the
per-workspace build PROMISE, so the never-settling connect poisoned the cache and
every later turn hung at step_count=0 before streamText — the run never finalized,
the row stayed 'running', and the chat was permanently blocked with
A_RUN_ALREADY_ACTIVE (an explicit Stop could not interrupt the un-signalled setup).
- mcp-clients: wrap connect() in a settling timeout (connectWithTimeout) that
closes a late-arriving client, so a hung handshake rejects instead of poisoning
the build cache; the bad server is skipped and the build completes.
- mcp-clients: close a connected-but-unregistered client when tools() fails,
fixing a pre-existing transport leak in buildEntry.
- ai-chat.service: bound the toolsFor build with the run's abort signal AND a
deadline (raceAgainstAbortAndTimeout); re-throw an explicit Stop only when a run
exists (runId) so the run finalizes as 'aborted', and keep the legacy
socket-bound path unchanged; settle the run 'aborted' vs 'error' accordingly.
- tests: cache-not-poisoned + orphan-client-close-once + Stop-during-setup
finalizes the run once.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Review round: add the two missing test-locks the reviewer asked for —
(1) priority order of PRIMARY_INPUT_FIELDS is now pinned (`{query,title}`
-> "Q", so a reordering breaks the test); (2) the clamp boundary is pinned
exactly (140 chars -> unchanged, no ellipsis; 141 -> 140 + "…"), catching
an off-by-one / `>` vs `>=` regression. Test-only, no production change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The researcher now reuses the current/already-open document only when the
user explicitly asked for it, OR the page is empty/near-empty AND its title
matches the research topic; otherwise it creates a new document.
- Rewrite the reuse rule in the WHERE TO WRITE THE RESULT section
- Reword "Create this document" -> "Set up this document" so it fits both
the create-new and reuse-current-empty cases
- Apply identically to bundles/research/ru.yaml and en.yaml
- Bump researcher version 3 -> 4 in index.yaml; refresh content-hashes.json
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
For tools without a friendly label (esp. external MCP tools like
Search_web_search) the card showed a generic "Ran tool <name>" with no
sense of what was searched. Add a compact one-line, dimmed summary of the
call's arguments under the label, pulled from part.input.
New pure toolInputSummary(part): picks the first present "primary" field
(query/q/searchQuery/url/urls/title/name/text/prompt), collapses
whitespace, clamps to ~140 chars; arrays render "first (+N)". It returns
undefined during input-streaming (the input grows while state is fixed and
messageSignature doesn't track input, so a live summary would freeze) —
the state flip to input-available re-renders the row with the final value,
so message-signature.ts is left untouched. The value renders ONLY through
Mantine <Text> (React-escaped) — no markdown/HTML, no XSS.
A showInput prop (default true) is threaded MessageList -> MessageItem
(+memo) -> ToolCallCard; the public share widget passes showInput={false}
so an anonymous reader never sees the agent's raw query text (mirrors
showCitations). No JSON fallback when no primary field is present.
closes#392
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DO-1 (regression): the orderedList start hardening `Number(x)||1` let a
negative/fractional start through into the marker ("-3.", "2.5.") — marked
can't tokenize it, so the whole list re-imported as a paragraph (structure
corruption); start=0 churned. Both the markdown and raw-HTML export paths
now use `Number.isInteger(raw) && raw > 1 ? raw : 1`, so any degenerate
start collapses to the default "1." markers / bare <ol> (list always valid).
New ordered-list-start-normalization test pins {0,-3,2.5} → valid start=1
list, byte-stable; the round-trip fuzz keeps to integers >=2 (num 2,3,5,42).
DO-2 (nightly was non-functional): NUM_RUNS=5000 OOM'd the worker and the
crash was misreported as a "counterexample" issue whose prefix-dedup then
locked out all future issues. Reworked to shard 8 fresh vitest processes
(600 runs each, distinct seeds, --max-old-space-size) so deep fuzzing
never OOMs; a failing shard's output is preserved, and issue creation
discriminates a real fast-check counterexample from an infra/OOM failure
(distinct titles + scoped dedup). The two issue steps use `always() &&`
so they actually run on the failure path.
DO-3: envInt extracted to test/generative/env-int.ts + unit-tested.
DO-4: nightly dispatch inputs go through env: (no ${{ }} in run:).
DO-5: attr-arbitraries.ts docblock synced (column.width/orderedList.start
are fixed+fuzzed, not pinned it.fails).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Items 1 & 2 of the #351 remainder.
Item 1 — the flat/nested generative property tests now read SEED and
NUM_RUNS from PROPERTY_SEED / PROPERTY_NUM_RUNS (via an envInt helper that
honors an explicit 0 and falls back to the current defaults —
20250705/300 flat, /100 nested — on unset/empty/non-numeric). New
.github/workflows/nightly-property.yml runs the generative suite daily
(and on workflow_dispatch) with a random seed and NUM_RUNS≈5000; on
failure it files a Gitea issue containing fast-check's shrunk
counterexample (jq-escaped, dedup'd by title). No build step — the suite
imports the converter from src/, so a tsc error can't masquerade as a
property failure.
Item 2 — new packages/prosemirror-markdown/README.md documents the
counterexample process (surface -> shrink -> permanent fixture in
test/fixtures/counterexamples/ + counterexamples.test.ts -> fix the
converter, never weaken a property; maintainer-approved ACCEPTED/allowlist
entries carry a reason) plus the two golden layers, the coverage
allowlist, and how to run with the env knobs. Linked from AGENTS.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Item 4 of the #351 remainder — implement value-fuzz for the media
dimension/family attributes previously parked in the flat property
suite's ATTR_VALUE_FUZZ_ALLOWLIST as "round-trip candidates deferred to
a later PR": width/height/align/aspectRatio/size/caption/title/alt across
image, video, youtube, pdf, drawio, excalidraw, embed.
Each gets a non-default-value arbitrary in attr-arbitraries.ts and is
removed from the allowlist; the P1 (semantic) + P2 (byte-stability)
property gate now exercises them. All round-trip green — the comment-JSON
serialization (stable key order, String()-stringified) carries every one
faithfully, so no new pins or accepted-limitations were needed.
embed.width/height are fuzzed as numeric STRINGS (not numbers): a
non-default embed dimension round-trips through the comment JSON as a
string (the numeric 800/600 default is still omitted/re-materialized),
so authoring a number diverged under P1.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Item 3 of the #351 remainder — resolve the two pinned converter
counterexamples, fixtures-first.
orderedList.start (genuine P1 loss): the converter always emitted "1."
and dropped attrs.start. Now the markdown path emits `${start + index}.`
(Number-coerced, guards a stray non-numeric start) and the raw-HTML path
emits `<ol start="N">` when start>1; a default (start=1) list is
byte-unchanged. tiptap StarterKit reads both forms back. Un-pinned as a
passing regression test; orderedList.start is now value-fuzzed.
column.width: the "50% churn" counterexample rested on a false premise —
the canonical editor (editor-ext column.ts) stores width as a unitless
flex-grow NUMBER (parseFloat, `flex:${width}`), never a "%" string, and
docmost-schema.ts is a vendored mirror that MUST match it. The original
parseFloat was correct parity and already byte-stable for numeric widths.
Reverted the mirror to parseFloat, deleted the fabricated counterexample
fixture, and now value-fuzz column.width as a number (real type). No src
behaviour change for column.width — parity with editor-ext preserved.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Snapshot the editor selection at send time (same live-ref pattern as
openPageRef in prepareSendMessagesRequest), carry it nested inside
openPage so it dies with the page on a fail-closed resolve, and surface
it to the model only through the existing core getCurrentPage tool.
The selection TEXT is returned exclusively in the tool result (untrusted
collaborative-page content, treated as data by SAFETY_FRAMEWORK); the
system prompt gets only a fixed one-line flag, never the text/before/
after. sanitizeSelection caps text (4000), before/after (200), blockIds
(<=64 chars, <=20). Selection is a hint, not ground truth — the tool
description tells the agent to localize the fragment before editing.
closes#388
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A user-provided "research budget" now defines the required search volume:
it is binding and must be spent in full, overriding the default "stop at
saturation" rule.
- STEP 0: split the budget into user-set (binding, must be used up) vs
self-estimated (when the user gave no number)
- VOLUME: limit "stop at saturation" to the no-budget case; add a
MANDATORY BUDGET block requiring the full budget be spent on genuine
broadening/lateral/primary-source/verification searches, not padding
- Apply identically to bundles/research/ru.yaml and en.yaml
- Bump researcher version 2 -> 3 in index.yaml; refresh content-hashes.json
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rework the "WHERE TO WRITE THE RESULT" section of the researcher role so
the report document is created at the very beginning of the run (right
after the plan, before any searches) and filled dynamically after each
finding, instead of being dumped in one pass at the end.
- Rewrite the section identically in bundles/research/ru.yaml and en.yaml
- Bump researcher version 1 -> 2 in index.yaml
- Refresh scripts/content-hashes.json via check.mjs --update-hashes
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The AI chat SSE endpoints (POST /api/ai-chat/stream, GET
/api/ai-chat/runs/<chatId>/stream, POST /api/shares/ai/stream) must
bypass response buffering AND compression at every proxy in front of
the app — a compressing proxy silently buffers SSE frames until the
response closes (pending request, tokens arriving in one burst,
reloaded tabs degrading to polling). Document the affected paths, the
DevTools tell (Content-Encoding on text/event-stream), and concrete
nginx/Traefik configuration in both READMEs.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
vvzvlad
agent_coder
agent_vscode