Repository navigation
π± chore: top up v6 from v5 (2026-11-15, second) - #18306
Conversation
Automated release commit. Compiles changelog.d/ fragments and moves the CHANGELOG.md Unreleased section into a dated v5.266.0 entry. See src/docs/releases.md. Signed-off-by: hive-release-bot <actions@github.com>
release: v5.266.0
* β¨ Inception (spek): voice notes in the session input bar Write roles get a Voice button in the Sessions input bar that dictates into the Send textarea through the browser Web Speech API. Interim text is shown live, the final transcript is an editable preview, and nothing is sent until Send / Send and run (same queued /input path). Audio never reaches the Hive backend; a first-use notice says the browser vendor's speech service transcribes it. The gateway Permissions-Policy for the dashboard document becomes microphone=(self); /terminal/ (ttyd iframe) and /api/ keep microphone=(). Signed-off-by: Andy Anderson <andy@clubanderson.com> * π Voice notes tests: load the sandbox with a write role Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦ct (#18281) native-queue-policy went red on #17512, #18273, #18275 and #18278 and the scanner filed it as one shared-CI incident (#18280). The gate was right every time: #17512 is a draft carrying do-not-merge/work-in-progress, #18273 and #18275 have probot DCO action_required, and #18278 has changelog-fragment-guard failure. The defect was the surface. A red provider was thrown as a bare `Error('DCO: action_required')` and printed as "Unhandled error" with a stack trace, which reads like a broken workflow. evidenceForHead and assertProwDCO now throw a ProviderVerdict naming the provider, its conclusion, the head, and that this is a per-PR verdict and not a shared CI incident. The evidence-wait step reports it with a job summary and core.setFailed instead of rethrowing. The conclusion stays failure: this is a required check. Hive-Run: #18280 Hive-Plan: shared-ci incident triage Hive-Spec: native-queue-policy#provider-verdict Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Signed-off-by: Danathar <Danathar@users.noreply.github.com>
β¦18274) * β¨ Inbox: Deepgram transcription for recordings without text (manual) Adds spoke-local machine transcription for Inbox items that arrive without text. Each hive owns its transcriptions: the owner's own Deepgram key (env var, never in hive.yaml), transcripts stored only on the spoke, per-hive budget, nothing shared or billed centrally. Off by default. - pkg/transcribe: Transcriber interface, Deepgram pre-recorded adapter (nova-3, diarize, smart_format, utterances, mip_opt_out=true, streamed body, timeout, 401/403 key error naming the env var, 429/5xx retryable), Retry wrapper, LimitReader, Fake adapter, RenderSegments/WriteLine. - Plaud transcript renderer now uses the shared line formatter. - config: inbox.transcription with named defaults, env overrides for the limits, validation and the consent gate. - inbox: transcript store (0600, atomic) under connector-state/transcripts, usage ledger, state machine, budget and size checks before upload, content fallback where upstream text wins. - dashboard: POST /api/inbox/{source}/{id}/transcribe (owner or read-write, audited per run as inbox_transcribe), settings field, Transcribe button, state pill and the settings section with the consent notice. - Docs, OpenAPI, API reference, env vars, changelog fragment. Signed-off-by: Andy Anderson <andy@clubanderson.com> * β»οΈ Inbox transcription: renderVTT on the shared line formatter; fix role test - renderVTT (meeting.go, now on v5) writes its lines through transcribe.WriteLine/FormatOffset like Plaud and machine transcripts. - Zoom/Teams/Discord pending notes mention the Transcribe button too. - JS test passes an empty role instead of undefined (a default param turned undefined into owner). Signed-off-by: Andy Anderson <andy@clubanderson.com> * π config: omit an unset inbox.transcription from the JSON view (agent runtime golden) Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
Automated release commit. Compiles changelog.d/ fragments and moves the CHANGELOG.md Unreleased section into a dated v5.267.0 entry. See src/docs/releases.md. Signed-off-by: hive-release-bot <actions@github.com>
release: v5.267.0
β¦ation (#18279) * β¨ Inbox: Google Meet recordings and transcripts as a knowledge integration Add a google-meet connector that reads Google Meet REST API v2 conference records (user-level OAuth with the hosted paste flow, or a service account with domain-wide delegation signing an RS256 assertion per impersonated user). Each ended meeting that produced a recording, a transcript or Gemini notes becomes one fact: event title from the Meet-created Drive file name (else the meeting code), duration, participants, the newest smart note exported from Docs as the summary, and the structured transcript entries rendered as [mm:ss] Speaker: text lines (Docs export as the fallback). Recording-only meetings are pending and re-read on later syncs. Meet stores the MP4 as a Drive file served only behind a bearer token, so Inbox playback answers no playable media and names the Drive page. 401 reads as not connected with the card hint, 403 keeps Google's reason, 429 honours Retry-After through the shared HTTP client, and the bearer token is only sent to the Meet and Drive API hosts. google-meet joins the default inbox.sources; the Integrations card shares the Connect / Check / Sync now controls, the paste fallback, the service-account note and the registration form. Docs, OpenAPI descriptions, Go and node tests. Signed-off-by: Andy Anderson <andy@clubanderson.com> * π Keep the Inbox pending-label map inside inboxSummaryHtml The inbox item tests extract inboxSummaryHtml into a sandbox; a module-level label table read from it threw ReferenceError there. Signed-off-by: Andy Anderson <andy@clubanderson.com>
β¦tml conflicts (#18277) * π ci: v6 top-up auto-resolves CHANGELOG, scripts-tree pin and index.html conflicts The v5 -> v6 forward-merge failed on every run since 2026-11-15 04:09Z. Two of its conflicts recur on nearly every top-up: CHANGELOG.md (both lines prepend release sections) and the TRUSTED_SCRIPTS_TREE literal in .github/scripts/integration-governance.cjs (pins HEAD:src/scripts, so it differs whenever both lines touch src/scripts). The generated static/index.html is mechanically regenerable too. Move the known-pattern resolution into .github/scripts/v6-topup-autoresolve.sh (run from a copy taken off the triggering ref) and extend it: union for CHANGELOG.md, re-pin to the merged src/scripts tree when every hunk in the pin file is the literal, regenerate index.html from the manifest. Any other conflict still fails for a human. Add a throwaway-repo test and a workflow that runs it. Signed-off-by: Andy Anderson <andy@clubanderson.com> * π ci: classify v6-topup-autoresolve-tests.yml in release-lines.yml Signed-off-by: Andy Anderson <andy@clubanderson.com>
* β¨ feat(crux): kick IDs and a bounded per-agent kick queue Add Manager.EnqueueKick / KickStatus / KickQueueDepth / InFlightKickID and the optional crux.KickQueuer interface. Each agent gets a bounded FIFO (HIVE_KICK_QUEUE_SIZE, default 8) drained by one pump goroutine that delivers when the agent is idle, through the same exactly-once dispatch slot SendKickAsync uses. Status by kick ID moves queued -> pending -> delivered -> completed (on the turn's outcome) or failed. SendKick and SendKickAsync are unchanged. Fixes #18285 Signed-off-by: Andy Anderson <andy@clubanderson.com> * π± test(crux): wait on time.After, not time.Sleep, in kick queue tests Keeps the sleep ratchet at its baseline. Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
Add Manager.AddKickObserver(fn func(crux.KickEvent)) (remove func()) and the optional crux.KickObserverRegistry interface. Each observer gets an ordered, buffered channel and its own goroutine; a full buffer drops instead of blocking. Events carry the in-flight queued kick ID, and the kick queue emits kick-queued, kick-completed and kick-failed. SetKickObserver keeps its single slot and its two events unchanged. Fixes #18286 Signed-off-by: Andy Anderson <andy@clubanderson.com>
Automated release commit. Compiles changelog.d/ fragments and moves the CHANGELOG.md Unreleased section into a dated v5.268.0 entry. See src/docs/releases.md. Signed-off-by: hive-release-bot <actions@github.com>
release: v5.268.0
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
|
Hive status β updated in place; edits do not notify. Full history is in the Activity Monitor. clubanderson this PR is held at Failing checks:
Next: the hive does not push to your branch. Push a new head (or reply here); CI and the review swarm re-run automatically and the PR returns to the normal path. |
875c954
βΉοΈ Sentinel notice β informational; author is trusted (owner), merge is not blockedHive flagged this PR (author @clubanderson, head
Hive did not add the |
|
Scanner baseline triage: not a shared incident. The branch is 1 commit behind π Hive Agent: β hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325 |
β¦dboxKickLocked Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
hive/scanner: DCO is failing and π Hive Agent: β hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325 |
Mutation testingHead β±οΈ The 900s mutation budget ran out, so 4 package(s) were not fully mutated and the results below do not cover them:
β gremlins did not produce a report for 1 package(s), so no verdict can be derived from this run (see the job log):
Verdict: |
Merges
origin/v5(cd7d951, release v5.268.0) intov6(48596cc) as a single signed merge commit (13 commits), soorigin/v5becomes an ancestor ofv6again. Carried by hand because the first automated run after #18277 (run 43737993411) stopped on a real source conflict insrc/crux/agent/kick_observer.go. That file is outside the known patterns, as intended. This supersedes the automated #18302, which carried an older v5 head.v5 PRs included
Conflicts
CHANGELOG.mdsrc/crux/agent/kick_observer.gosrc/crux/agent/kick_observer_test.goerrors(v5) andfmt(v6)kick_observer.go: what each side changed
v6, from π v6: @-mention triggers β let a human summon an agent from a GitHub threadΒ #7483 ("mention: reply when summoned runs complete", a v6-only feature). The single-slot
SetKickObserverchains: a second call runs after the installed observer instead of replacing it, because v6 has two consumers, the mention responder (cmd/hive/mention_runtime.go) and the Linear agent (dashboard/api_linear_agent.go). Archive events also carry the kick source in their detail (kickObserverArchiveDetail,KickObserverDetailSource/Reason). The responder uses that to recognise the kicks it summoned.v5, β¨ Crux: fan-out kick observers with kick IDsΒ #18303 ("fan-out kick observers with kick IDs", fixes β¨ Crux: fan-out kick observers with kick IDsΒ #18286) adds:
AddKickObserverwith per-observer buffered channels and a remover;crux.KickObserverRegistry;kick-queued/kick-completed/kick-failed;crux.It leaves
SetKickObserveras the single-slot compatibility path ("its two events unchanged"), with the Linear agent as its only v5 consumer.Neither supersedes the other: #18303 is new fan-out machinery, and #7483's chaining and source detail are v6 behaviour that v5 never had. Both survive:
crux-backed constants, buffer constant, registry assertion,kickObserverSet,AddKickObserver,emitKickEventand the newnotifyKickObserver(single slot plus fan-out with kick ID).SetKickObserver(it merged cleanly). Its doc comment says the method is the compatibility path, independent of fan-out, and that it chains on this line for its two consumers.Every v6
notifyKickObservercall site (deferred-mentionkick-dropped, archive with source, sandbox) is unchanged and now also reaches fan-out observers.Other checks
TRUSTED_SCRIPTS_TREEalready equalsHEAD:src/scripts(70a83fac), andstatic/index.htmlequals the manifest concatenation.gofmt -lis clean on the hand-edited files. Nothing was built or tested locally; CI is the gate.action_requiredfor the unsigned release-gate merge commits already on v5 (as on π± chore: top up v6 from v5 (2026-11-14)Β #18193 and π± chore: top up v6 from v5 (2026-11-15)Β #18275). The Prowdcocommit status is the gate.π€ Generated with Claude Code