The watch lane settled the investigation with `settleOpenInvestigations`, then
appended the terminal card as a separate write — and swallowed that write's
error, logging it and reporting success. So the row went terminal, the card
never arrived, the stale sweep stopped selecting the row because it was no
longer `in_progress`, and the user was left on "Working…" with nothing able to
repair it. Nothing in production calls the action again on its own, which is
exactly why the swallowed error mattered.
The lane now writes through `settleInvestigationStateAndCloseCard`: the terminal
revision and the closing card commit in one transaction, under the lane's own
message id, and the error propagates so the task's retry is a real retry. If the
card cannot be rendered the settle rolls back, leaving the `in_progress` row the
sweep still selects. The duplicated revision bump is gone with it — one outcome
is now one revision.
The regression test drives a failing close in one action; it no longer proves
recovery by calling the action a second time by hand.
`persistTurn` and `persistMessages` stored the whole `messages` array they were
handed. The array is the snapshot the turn started from, so anything another
process appended in between was deleted: a wake delivery, a watch consent
record, or the terminal card of an investigation the stale sweep had just
settled. That last one is unrecoverable — the row is already terminal, so the
sweep never selects it again, and the panel is back to "Working…" for ever.
Both writes now read the row under `select ... for update` inside the
transaction and merge by stable message id: incoming order is kept, a stored
message the snapshot does not have goes at the end, and no id appears twice. A
message with no id falls back to its content so it cannot be carried over twice.
An append-only `chat_messages` table is the better long-term shape; merging
under the lock is enough for this architecture and needs no migration.
Covers the two halves the earlier pass left open. Against a real database, a
stale card that cannot be rendered now proves the sweep rolls the settle back
with it: the row stays `in_progress` at revision 0, the chat stays empty, and
the row is still in the next run's selection.
The panel half is covered over `liveProgress`, the code that decides whether
"Working…" is shown: a mounted panel holding the unconcluded card re-reads the
stored transcript, merges by stable id, and the progress line goes away without
a reload. Re-reading repeatedly cannot add a second copy of the card.
The row settle, the transcript and the session state were three separate
operations. Once the row was terminal, a failed transcript write left a card
that read `in_progress` forever: the stale sweep only selects `in_progress`
rows, so nothing was left to repair it.
Both lanes now commit the pair atomically. The live turn hands its pending
settlements to `persistTurn`, which upserts the revisions, appends their cards
and writes the session in one transaction; the process-local entry survives
until that commits, so a retried `onTurnComplete` still settles. The sweep goes
through `settleInvestigationAndCloseCard`, whose rollback restores the
`in_progress` row the sweep already selects.
That also removes the last reader of the per-run `chatOwners` map, which had no
`delete` and grew for the life of the worker: the failure record now travels in
the transcript write, which needs no userId.
Separately, the consented watch investigation gets the rolling step cache. Its
ten-step `streamText` re-sent every accumulated tool output uncached; the
breakpoint helper and the per-step cache telemetry now live in `step-cache.ts`
and both lanes use them, wrapping any `prepareStep` the resolved options carry.
Binding a delegated token to its environment claim on the project-wide
routes also refused any token that carries no claim at all, which the
public PAT exchange used by MCP and the CLI is allowed to issue. Line the
project-wide helper up with its neighbours: a claimless dashboard-agent
token is still refused, everything else stays project-wide.
The sweep settled the row and appended nothing, so it visibly fixed nothing: the
chat kept rendering the last card it had, which was still "Working…". The settle
now returns the state and revision it wrote, and the sweep appends that as the
closing card revision on the chat — id-deduped on
`investigation-settlement:{id}:{revision}`, so a retried run can neither stack a
second card nor open a second investigation.
The append is scoped by chat id: a sweep runs off any session and has no user in
context, unlike the turn lane.
Settling the investigations row was invisible to the user. The panel builds the
winning revision from the transcript's own `tool-render_view` parts and never
reads that table, so a turn that ran out of steps left the card at
`in_progress` forever: the database believed the investigation had finished
while a refresh still showed "Working...".
`settleOpenInvestigations` now returns the revisions it committed, and
`onTurnComplete` appends each as one more card revision — after the transcript
write and id-deduped on `investigation-settlement:{id}:{revision}`, so a failed
append leaves the card visibly unclosed rather than silently lost, and a retry
can't stack a second card.
The card-building and the latest-revision reader move out of the watch lane and
into the runtime both lanes share, so there is one shape, not two. The watch
lane keeps its own message id: it dedupes on the action, not the revision.
The create-watch response matched any watch-alert channel in the project, so a
second member was told they were subscribed while the mail went to the first.
The channel's deduplication key is the only record of whose it is, so state
resolution, subscribe and unsubscribe now share one owner lookup.
replay() re-ran subscribe() for an already-recorded `created` submission. A retry that
succeeded could flip the ledger to `enabled`, but the confirmation in the transcript is
append-once, so the user kept being told email was unavailable while the system believed
it was on. The replay now reads recordedExternalNotification() and takes no external
decision. An attempt that dies before subscribing or before its outcome is recorded
leaves the row `pending`, and the normal creation path subscribes on the retry.
A user-actor token declaring no scope cap could mint an environment JWT with any
scopes it asked for. The exchange now clamps the minted scopes to the actor's own
ability, so a capless token is read-only, and only mints for its claimed environment.
The direct PAT authentication path returned identity only, so a user-actor token
reaching it (admin routes and other direct callers) lost its environment scope. The
claims now ride on the authentication result itself.
The environments and runs listings are project-wide, so an environment-scoped
user-actor token could read every environment its user can reach. Both routes now
resolve the claim into a mandatory filter, and a conflicting request filter is
refused rather than overridden.
A tick that could read nothing now moves the group's fairness key only, so a watch
with a permanently broken reader stops crowding out the rest of an over-cap group.
Dueness and the streak facts still follow the last real check.
A per-condition fallback identified the condition rather than the submit, so a
re-watch could replay a stale terminal outcome from the retention window.
Classify a failed tool result locally into one of seven categories and send only
the label; the message is dropped with every other free-text field. Unrecognised
failures are unknown rather than guessed, and a bare string error field is now
withheld too.
Replaces https://*.googleusercontent.com (a host with public write access)
with CSP_IMG_SRC_ALLOWLIST: exact origins only, https outside development,
deduplicated, bad entries warned about instead of failing the boot.
Route conflicts were the tab-title work meeting the agent page-context handle: both
sides kept, duplicate meta exports resolved to pageMeta, duplicate imports merged
with unused bindings dropped. Lockfile regenerated from the merged manifests.
The hand-built row stopped satisfying every column the authenticated-environment
mapper reads, so both proxy cases failed. Stubbing the lookup keeps the fixture
independent of the row's columns.
A step-level 5-minute cache breakpoint rolls onto the growing history once it is large enough for the write to pay for itself, and each step logs the provider's cache write and read counts.
The judged turn now passes an allow-list of structural fields instead of a deny-list of known-sensitive names, unfolds the tool-output envelope so error text is redacted too, truncates inputs as well as outputs, and caps the whole turn.
The winners map was rebuilt on every streamed token and handed to the memoized turns as a fresh reference, re-rendering the transcript. The reference is now reused while the winners hold.
A remote image in model output fetches on render, so its URL is an exfiltration channel. No image is rendered at all, and the document CSP now names the hosts images may come from.
The chat id a fresh submit derives, the transcript appends, the create's chat read
and the ledger replay are all owner-scoped now, the fire callback alerts once per
terminal outcome, and a failed email subscription is stated instead of dropped.
Katia Bulatova