Skip to content

🌱 chore: top up v6 from v5 (2026-11-15, fourth) - #18341

Merged
clubanderson merged 29 commits into
v6from
topup/v5-to-v6-20261115d
Nov 16, 2026
Merged

clubanderson merged 29 commits into
v6from
topup/v5-to-v6-20261115d

Conversation

@clubanderson

Copy link
Copy Markdown
Member

This PR merges origin/v5 (4233d0a, #18338 plus release v5.274.0) into v6 (ad5c7fa, v6.33.1) as a single signed merge commit. After it lands, origin/v5 is an ancestor of v6 again. All 25 v5 commits since the last top-up (#18312) are ancestors of the merge (checked with git merge-base --is-ancestor, including each PR listed below).

I carried it by hand because the automated run stopped on content conflicts outside the resolver's known patterns. The resolver bails out when any unknown conflict exists, so I applied its known patterns by hand: the CHANGELOG.md union and the static/index.html regeneration. No automated top-up PR is open, so there is nothing to supersede.

v5 PRs included

Conflicts

File Resolution
CHANGELOG.md Kept both lines' release sections, v6 first (the resolver's union), with a blank line restored before ## 2026-11-15 (v5.274.0).
dashboard/openapi.json Structural 3-way merge, not text hunks. Both lines prepend new paths, and v6 reorders paths, so the text conflict covered about 870 lines. I kept v6's text and order and applied v5's changes to it: the Meetings tag after v6's Workflows and Strategy Lab tags, the nine new /api/meetings* paths at the top of paths (where v5 puts them), and v5's text for /api/inbox, /api/inbox/route, /api/inbox/{source}/{id}/content and /api/inbox/{source}/{id}/audio. v6 had not changed those four. A script checked that the parsed result equals v6's document plus exactly v5's changes. The file is still in v6's own formatting.
src/docs/api-reference/{auth,configuration,dashboard-misc,health-status-events,packs-and-acmm,snapshots-and-style}.md Kept v6's rows. Apart from line-number citations, v5's base-to-tip diff on all six pages is empty: no row was added, removed or reworded, only pkg/...:<line> numbers changed to match v5's sources. On v6, those citations point at v6's server.go and api.go. If the merged route registrations moved v6's lines, CI's citation check will flag them, and I will fix them with the checker's --fix then.
src/cmd/hive/boot.go Imports only: kept v6's mention and mergelane and added v5's meetings and meetings/live, in sorted order. The meetings fields on bootState merged cleanly.
src/cmd/hive/main.go Imports: kept v6's matrix and msteams and added v5's meetings/meetllm. dispatchSubcommand: kept v6's advisor.HookSubcommand and nousdeps.Subcommand cases and added v5's meetllm.Subcommand case after them. The research-pool and meetings wiring from #18318, #18323 and #18329 merged cleanly. It sits after v6's Spektacular clone-auth wiring, which #18312 kept.
src/pkg/spektacular/runner.go v6 moved NotFoundError, VerbError and ContractError to pkg/planengine and re-exports them as aliases in aliases.go (ADR-0021). v5 still defines them in runner.go, and its only edit there was adding StreamExec (#18314). I kept v6's move and added only StreamExec. Adding v5's copies would redeclare the alias names. The new stream.go merged as is.
src/pkg/dashboard/static-src/js/config-state.js Kept v6's GOVERNOR_CONFIG_TABS (with Channels, Extensions and Smart classifier) and added v5's Meetings tab between Project Observability and Advisory, where v5 puts it.
src/pkg/dashboard/static-src/js/sections.js Kept both: v6's search-section heading and v5's meetings-section heading.
src/pkg/dashboard/static/index.html Rebuilt with go generate ./pkg/dashboard/webstatic/. It equals the byte-exact static-src/manifest.txt concatenation.

Semantic checks

  • No file v5 modified is missing on v6, except the six changelog.d fragments that both lines' releases already removed.
  • These directories match v5 exactly: pkg/inbox, pkg/transcribe, pkg/meetings (including the new live and meetllm), pkg/research, and the new cmd/hive meetings files. The same holds for config/inbox_config.go and config/meetings_config.go, including the six connectors' inbox.sources defaults.
  • pkg/knowledge/connector and pkg/spektacular differ from v5 only in v6's own pre-existing changes (the sharepoint.go/publish_test.go and planengine ports). Each diff against v5 has the same file list as v6's diff against the merge base.
  • Each of the seven new connector ids (webex, slack-huddles, fireflies, fathom, limitless, granola, otter) appears in the merged src/ exactly as often as on v5. That covers the connector registry, KB_AUTH_TYPES, defaults and docs.
  • The research pool needs crux.KickQueuer, crux.KickObserverRegistry, AddAgent, GetStatusFast and Start. v6's runtime has all of them (🌱 chore: top up v6 from v5 (2026-11-15, second) #18306, 🌱 v6: converge kick_observer.go with v5's fan-out observers #18315). v6's bootState.agentMgr is the same hostedRuntime type as v5's.
  • TRUSTED_SCRIPTS_TREE already equals HEAD:src/scripts (70a83fac), because v5 did not touch src/scripts.
  • src/pkg/config/config.go did not conflict. v5's config changes are in inbox_config.go and meetings_config.go, which exist on v6 and merged cleanly.

Other checks

🤖 Generated with Claude Code

clubanderson and others added 26 commits November 15, 2026 17:43
…18301)

* ✨ Inbox: Webex recordings and transcripts as a knowledge integration

Add a webex connector that reads Webex Meetings through a Webex integration
(OAuth authorization code with the hosted paste flow; tokens in the
connector token store, rotated refresh tokens written back). Recordings
(GET /recordings) and transcripts (GET /meetingTranscripts) are listed in
explicit 30-day from/to windows with max=100, following Link rel=next, and
grouped per ended meeting instance: topic, host, duration, participants
(GET /meetingParticipants), the AI Assistant summary and action items
(GET /meetingSummaries, HTML converted), and the newest transcript
downloaded as VTT and rendered as [mm:ss] Speaker: text lines. Meetings with
a recording but no text are pending and re-read on later syncs.

Inbox playback asks GET /recordings/{id} for its temporaryDirectDownloadLinks
per click (audio, else MP4; no token in the link, valid 3 hours) and never
stores it. 401 and a dead refresh token read as not connected with the card
hint, 403 keeps Webex's reason, 429 honours Retry-After, and the bearer
token is only sent to webexapis.com.

webex joins the default inbox.sources; the Integrations card shares the
Connect / Check / Sync now controls, the paste fallback and the integration
form. Docs, OpenAPI descriptions, Go and node tests.

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🌱 Webex test follows the HTML converter and the recording-page line

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🌱 Cover the Webex sign-in, grouping and truncation edges

Signed-off-by: Andy Anderson <andy@clubanderson.com>
* ✨ meetings: meeting entity, quorum MCP client and SSE consumer

Meeting record and atomic store under /data/meetings, a streamable-HTTP
MCP client for the contract tools, the /events SSE consumer with
reconnect, consent notice on live before anything else, hive leave
detection, lobby timeout, and boot wiring with resume after restart.

Fixes #18290

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* ✨ meetings: contract additions (voice at join, event replay, platform join requests)

meeting.join carries voice (fixed at join time by Vexa), the event stream
replays from after_seq on reconnect, Discord /hive join requests on the
all-meetings stream wait for the owner (or an AutoApprove policy), and a
consent notice that cannot be posted (platform_unavailable) is a soft,
throttled failure recorded on the meeting.

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 meetings: record snapshot-only join requests and skip own channel joins

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 meetings: fake quorum records after_seq

Signed-off-by: Andy Anderson <andy@clubanderson.com>

---------

Signed-off-by: Andy Anderson <andy@clubanderson.com>
* ✨ meetings: Spektacular live loop and meeting policy

StreamExec for long-lived spektacular children (pipes, stderr to logs,
restart with backoff, end then kill), the per-meeting live session and
policy layer (chat rate limit, ask/answer matching, notes as one
knowledge fact per meeting, research cap and approval through a
dispatcher interface, voice rules, budgets), research signal routing by
kick ID, and boot wiring.

Fixes #18291

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* ✨ meetings: hive-owned model for Spektacular live mode

meetings.llm resolves like the knowledge advisor (governor.litellm
endpoint, key and default model) with its own timeout and in-flight
limit, or an Anthropic key passed only into the Spek child env. Spek runs
SPEK_MEETING_LLM=command against the new 'hive meeting-llm' subcommand,
which reads endpoint and key from the env (never argv), shares in-flight
slots across meetings, writes a per-meeting usage ledger the live loop
charges, and refuses past the per-meeting budget; with no model Spek runs
on rules only and the panel says so. A Spek that exits early (a
spektacular without the meeting verb) leaves the session running for the
transcript and research.

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 meetings: research cap counts only earlier items

Signed-off-by: Andy Anderson <andy@clubanderson.com>
#18313)

* ✨ Inbox: Slack huddle notes and transcripts as a knowledge integration

Slack never records huddle audio and has no huddle-recording API; what it
keeps is the AI notes canvas it posts in the huddle thread (with the
transcript embedded when the workspace allows transcripts). Add a
slack-huddles connector (a Slack bot or user token named in auth.env /
auth.file) that pages conversations.history for the listed channels, keeps
the ended huddle_thread messages, finds the notes canvas in the thread
(conversations.replies) or the room's attached_file_ids (files.info), and
reads it as markdown with canvases.getContent, falling back to the canvas
file's private download. A Transcript heading splits notes (summary) from
transcript. One fact per huddle with title, host, participants (users.info),
duration and the message permalink; huddles without notes are skipped, and
items are text only (no playback).

Slack's ok:false replies map to not connected (invalid_auth, token_revoked,
...), a named missing scope, or an /invite hint; 429 honours Retry-After and
the token is only sent to Slack hosts. slack-huddles joins the default
inbox.sources with a Connect / Check / Sync now card.

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 Slack canvas download keeps headings and line breaks

The shared HTML converter joins block text, so the downloaded canvas lost
its Transcript heading; convert canvas HTML with a small block-aware pass.

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 v5.270.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
…o voice command) (#18278)

* 🐛 Inbox: route File RFC through an agent with spek (spektacular has no voice command)

The shipped spektacular 0.31.0 has no voice command, so the direct CLI
path failed on every spoke and the agent route never ran. File RFC now
always hands the item to an agent with the enveloped content plus
provenance; the agent drafts the spec with the spek-new workflow and files
it with spektacular publish. Failed hand-offs show their error inline;
delegated routes record the agent in a note field.

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 Inbox: keep inbox.spek.binary referenced in the Settings UI; refresh api-reference citations

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 Inbox: File RFC uses the machine transcript fallback and refuses pending items

Signed-off-by: Andy Anderson <andy@clubanderson.com>
Owner-only /api/meetings routes per quorum.contract/v1 section 3 behind a
MeetingsService seam (pkg/dashboard never imports pkg/meetings), Discord
join request approve/decline, a WebSocket live panel modelled on the Jam
socket, meeting frames on /api/events, the Meetings dashboard section and
Settings, Meetings tab (enable, Vexa URL, write-only Vexa key and Discord
token through the keystore, resolved Spek model), openapi.json and
api-reference rows.

Fixes #18292

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 v5.271.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
#18327)

startSessionInBackground moves the record out of `preparing` before it
writes the session_started audit, so a test that returns from waitSettled
and reads the audit log once can see no entry yet. Poll for the audit
entry instead, in both the failed-start and running-start tests.

Fixes #18325

Hive-Run: #18325
Hive-Plan: ci-flake fix
Hive-Spec: ci-flake#TestSessionsStartFailureMarksRecordFailed

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Danathar <Danathar@users.noreply.github.com>
* ✨ meetings: post-meeting hand-off, retention and docs

Ended meetings become a 'meeting' inbox source (final transcript plus
notes) routable to spek_rfc and dibs; inbox sources can now learn of a
hand-off (HandOffSource), which makes hive call meeting.delete upstream.
A retention sweeper purges transcripts and live state after
meetings.retention_days. Docs: meetings.md, integrations rows, docs map.

Fixes #18293

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 meetings: reuse the inbox truncateRunes

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 meetings: list hive's own meetings regardless of inbox.sources

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 v5.272.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
* ✨ meetings: wire the research pool and pin the quorum release

Research intents go to research.Pool (research-1..N added to the runtime
when meetings are on, backend and model from the scanner), and pool
events map to the live loop's finding/complete/error signals by kick ID.
A runtime without queued kicks (sidecar) leaves research off with a log.
The Dockerfile pins hivecommons/quorum v0.1.0 (quorum.mjs, sha256
verified) and installs @snazzah/davey 0.1.12 next to it for Discord.

Fixes #18294

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 meetings: keep research signals that arrive before Dispatch returns

Signed-off-by: Andy Anderson <andy@clubanderson.com>

---------

Signed-off-by: Andy Anderson <andy@clubanderson.com>
Add a fireflies connector (a Fireflies API key named in auth.env /
auth.file) that reads meetings through the Fireflies GraphQL API:
transcripts (limit 50, skip, fromDate, optional mine) with sentences and
the AI summary in the same paged query, so a sync costs one request per 50
meetings. One fact per meeting with title, date, duration (minutes),
organizer, attendees, the overview / action items / topics / keywords as
summary and the speaker-labelled sentences as transcript; meetings without
text yet are pending and re-read on later syncs.

Playback and the Deepgram fallback use the transcript's hashed audio_url
(else video_url), valid 24 hours on Pro plans and higher; paid_required
answers no media with that reason. auth_failed / HTTP 401 read as not
connected, 429 honours Retry-After, and too_many_requests and paid plans
get their own card hint. fireflies joins the default inbox.sources with a
Connect / Check / Sync now card.

Signed-off-by: Andy Anderson <andy@clubanderson.com>
Add a fathom connector (a Fathom API key named in auth.env / auth.file,
sent as X-Api-Key) that pages GET /meetings (created_after, next_cursor,
optional recorded_by[] / teams[]) with include_transcript, include_summary
and include_action_items, so the text comes in the same pages. One fact per
recording with title (else the calendar title), recording time and
duration, the recorder as host, the calendar invitees, the default summary
markdown plus action items as summary and the speaker-labelled transcript;
the Inbox listing reads metadata only and an opened item reads
/recordings/{id}/transcript and /summary.

Playback and the Deepgram fallback use Fathom recording downloads: POST
/recordings/{id}/download (audio-only recordings complete at once), the
pending download is polled with GET /recordings/{id}/downloads/{id} on
the next click, and the completed file's signed url (valid until
expires_at, about 24 hours) is returned, never stored; 422 answers no
media. 401 reads as not connected, 403 names limited-access sharing, 429
honours Retry-After. fathom joins the default inbox.sources with a
Connect / Check / Sync now card.

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 v5.273.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
* ✨ Inbox: Limitless as a knowledge integration

Add a limitless connector (a Limitless API key named in auth.env /
auth.file, sent as X-API-Key) that pages the Developer API's
GET /v1/lifelogs (start in UTC, direction desc, limit 10, the maximum,
following meta.lifelogs.nextCursor; optional isStarred) with
includeContents, so the transcript comes in the same pages. One fact per
lifelog with title, start time, duration, the speakers, the topic
headings as summary and the blockquote lines as transcript (offsets from
startOffsetMs); without contents the lifelog markdown is the transcript.
The Inbox listing reads metadata only and an opened item reads
GET /v1/lifelogs/{id}.

Pendant audio is served only by /v1/download-audio behind the API key,
which the hive never hands to a browser, so items are text only and
MediaLink answers no media. 401/403 read as not connected; 429 honours
Retry-After and names the body's retryAfter. limitless joins the default
inbox.sources with a Connect / Check / Sync now card.

Signed-off-by: Andy Anderson <andy@clubanderson.com>

* 🐛 Google Meet card test no longer pins the connector after google-meet

Limitless sorts between google-meet and notion in KB_CONNECTOR_TYPE_ORDER;
the registry-order test already guards the full list.

Signed-off-by: Andy Anderson <andy@clubanderson.com>
Add a granola connector (a Granola API key, grn_…, named in auth.env /
auth.file) that pages GET /v1/notes (updated_after, page_size 30, cursor
while hasMore; optional folder_id / space_id), skips deleted notes and
reads each with GET /v1/notes/{id}?include=transcript, falling back to the
paged /v1/notes/{id}/transcript when Granola answers 413. One fact per
meeting with title (else the calendar event title), scheduled time,
duration, owner, attendees (else calendar invitees), the AI notes and the
owner's own notes as summary and the transcript (me / them when Granola
has no speaker name). A note without generated notes answers
NOTE_NOT_FOUND and is picked up by a later sync.

Granola keeps no audio, so items are text only and MediaLink answers no
media. 401/403 read as not connected; 429 honours Retry-After. granola
joins the default inbox.sources with a Connect / Check / Sync now card.

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 v5.274.0 entry.
See src/docs/releases.md.

Signed-off-by: hive-release-bot <actions@github.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Add an otter connector (an Otter.ai Public API key, Enterprise workspaces,
named in auth.env / auth.file) that pages GET /conversations (newest
first, limit 20, the maximum, cursor while meta.has_more; include_shared
or channel_id) until a conversation older than the cursor minus the 48 h
overlap or since_days, and reads each with
GET /conversations/{id}?include=action_items,outline,transcript. One fact
per conversation with title, creation time, duration (last transcript
offset), owner, calendar guests, the abstract summary, outline and action
items as summary and the txt transcript ("Speaker  00:07" heads) as
[mm:ss] Speaker: text lines.

Playback and the Deepgram fallback use GET /conversations/{id}/audio, the
conversation's MP3 download link, minted per click with a one-hour expiry
(Otter states none) and never stored. 401 reads as not connected (a wrong
key or a workspace without the Public API), 403 keeps Otter's refusal, 429
honours Retry-After. otter joins the default inbox.sources with a
Connect / Check / Sync now card.

Signed-off-by: Andy Anderson <andy@clubanderson.com>
Forward-merge origin/v5 (4233d0a, #18338) into v6 (ad5c7fa, v6.33.1). 25 v5 commits.

Signed-off-by: Andy Anderson <andy@clubanderson.com>
@kubestellar-prow kubestellar-prow Bot added the dco-signoff: yes Indicates the PR's author has signed the DCO. label Nov 15, 2026
@kubestellar-prow

Copy link
Copy Markdown
Contributor

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign clubanderson for approval. For more information see the Code Review Process.

The full list of commands accepted by this bot can be found here.

Details Needs approval from an approver in each of these files:

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

@kubestellar-prow kubestellar-prow Bot added the size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. label Nov 15, 2026
@hivecommons-hive

Copy link
Copy Markdown
Contributor
3e66b03

ℹ️ Sentinel notice — informational; author is trusted (owner), merge is not blocked

Hive flagged this PR (author @clubanderson, head 3e66b03a22ea) because it matches behaviors that can override security controls, escalate privileges or damage the codebase. This is a heuristic, not an accusation — a maintainer should confirm the change is intended before it merges.

  • sensitive_path — changes 2 sensitive path(s) (Touches a sensitive path (OWNERS, workflows, policies, hive config, security docs, …))
    • src/Dockerfile
    • src/Dockerfile.contributor

Hive did not add the sentinel-alert label because this author is trusted. The finding remains recorded for audit and the dashboard Security tab; set sentinel.trusted_authors_block: true to require the blocking label for trusted authors too.

Ran src/scripts/check-api-reference-citations.sh --fix: 44 route citations moved because the merged server.go/api.go routes shifted v6's lines.

Signed-off-by: Andy Anderson <andy@clubanderson.com>
@hivecommons-hive

Copy link
Copy Markdown
Contributor

scanner: baseline triage says MERGE_BASE — branch is 3 commits behind v5; native-queue-policy/DCO failures are from that staleness. Please update from v5 with sign-off; not pushing to a non-App branch.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

HUMAN DECISION NEEDED — style review could not be completed: GitHub refused the PR diff (HTTP 406, over 20000 lines), so a human must review this v5→v6 merge. @clubanderson

I read the PR title and body (head 5c5372b) but not the diff, so I have no file:line evidence and report no findings. I did not infer the contents of the diff. This is a style-perspective review only.

— hive: agent=adjudicator-4 backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive

hivecommons-hive Bot commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor.

Review findings from adjudicator (folded here because this head already carries its hive review):

HUMAN DECISION NEEDED — correctness review could not read the diff (GitHub HTTP 406, over 20000 lines); a human must decide whether to accept this v5→v6 merge on CI plus the author's checks. clubanderson

I did not review this PR for correctness. gh pr diff 18341 returned PullRequest.diff too_large, so I read none of the merged code at head 5c5372b. I am not inferring correctness from the PR body.

Parts the body flags as hand-resolved, which I could not check:

  • dashboard/openapi.json: the structural 3-way merge.
  • src/pkg/spektacular/runner.go: the ADR-0021 alias reconciliation.
  • src/cmd/hive/main.go and src/cmd/hive/boot.go: the meetings wiring.
  • static/index.html: the regeneration.
  • The six api-reference pages, where the author left the line-number citation check to CI.

Please confirm CI is green, including the citation checker, a build, and the tests for the touched packages.

Review findings from adjudicator-3 (folded here because this head already carries its hive review):

HUMAN DECISION NEEDED — clubanderson: intent-alignment could not be machine-verified because the PR diff exceeds GitHub's 20000-line limit; a human must confirm the merge resolutions.

gh pr diff 18341 returned HTTP 406 (diff exceeded 20000 lines), so I read only the PR body, not the diff. I did not infer or cite any file:line from the unread diff. The body's claims (hand-resolved conflicts in dashboard/openapi.json, src/cmd/hive/main.go, src/pkg/spektacular/runner.go, and the regenerated static/index.html) are unverified by this perspective. Please check them by comparing the merge commit against the v5 and v6 parents locally.

Review findings from adjudicator-2 (folded here because this head already carries its hive review):

HUMAN DECISION NEEDED — security review could not read the diff for this v5→v6 top-up (clubanderson); a human must review the new meetings/inbox connector surface.

gh pr diff 18341 fails with HTTP 406: the diff exceeds GitHub's 20000-line limit. I read only the PR body, not the code, at head 5c5372b. I am not approving or flagging any specific line.

The security-relevant surface I could not check is large. It includes the new /api/meetings* owner API, the quorum MCP client and SSE consumer, and seven new inbox connectors (webex, slack-huddles, fireflies, fathom, limitless, granola, otter) with their auth and token handling. It also includes the /api/inbox/{source}/{id}/audio and /content endpoints, the meetllm.Subcommand dispatch and the hand-resolved runner.go/boot.go/main.go conflicts.

The body says these paths match v5 byte-for-byte, so they were presumably reviewed in their v5 PRs. A human should confirm the merge introduced no deviation from v5 in these paths, for example with git diff origin/v5 HEAD -- src/pkg/meetings src/pkg/inbox.

Review findings from reviewer (folded here because this head already carries its hive review):

Routing, not a defect. This is a v5→v6 merge, and it touches image build files (src/Dockerfile, src/Dockerfile.contributor). I could not read the full 116-file diff: GitHub refuses it above 20000 lines. I read the two Dockerfile diffs against the v6 merge-base and checked the CI state.

  • src/Dockerfile (Layer 8.2): QUORUM_VERSION=v0.1.0 and QUORUM_SHA256 are now pinned and the bundle is sha256-verified. That part looks sound. The new step npm install … "snazzah/davey@${QUORUM_DAVEY_VERSION}" pulls a native module from the npm registry with --ignore-scripts, but nothing checks an integrity hash or lockfile. A compromised 0.1.12 tarball would land in /usr/local/lib/quorum in every image. If it fails to install, the build only warns. That is a supply-chain call for a maintainer.
  • native-queue-policy is red because DCO is action_required on this head. The PR body already attributes it to unsigned release-gate commits on v5, and the Prow dco status is green. The workflow says the next push re-evaluates, so this will probably stay red until a maintainer accepts it.

The PR body's conflict-resolution claims (openapi, spektacular aliases, static/index.html) are outside what I could verify; CI's citation and build jobs are the check for those.

Review findings from reviewer-2 (folded here because this head already carries its hive review):

Not a code defect: a maintainer needs to confirm two things before this merges.

Limits of this review: GitHub refuses the full diff (20k-line cap), so I read the file list, the per-file patches for the two Dockerfiles, and the check results. I did not re-verify the hand-merged openapi.json, static/index.html or Go conflict resolutions line by line.

  1. DCO is red on this head. native-queue-policy fails with DCO: action_required on 5c5372b… (job log, "Validate exact-head evidence"). The body anticipates this (unsigned release-gate merge commits already on v5), and the Prow dco status is SUCCESS, but the repo's own queue policy is reading the GitHub DCO app. Someone has to decide whether that is an accepted exception or needs a fix before merge.
  2. Supply-chain surface. src/Dockerfile:~452-480 moves QUORUM_VERSION from empty to v0.1.0 with a pinned QUORUM_SHA256. It also adds npm install … snazzah/davey@${QUORUM_DAVEY_VERSION} (0.1.12, --ignore-scripts; failure only warns). The sha256 check on quorum.mjs is in place. The npm package is version-pinned but has no integrity hash, so it is only as trustworthy as the registry. These changes arrive via v5 (✨ meetings: wire the research pool and pin the quorum release #18329), so they are probably already reviewed there. Please confirm.

Verified OK: the Dockerfile and Dockerfile.contributor spektacular comment edits are comment-only, and the build, test, lint, gosec, govulncheck and patch-coverage checks are green. mobile-shots, mutation and overlayfs-exec-guard were still running when I looked.

@clubanderson this PR is held at stage/071-needs-author and the hive will not merge it yet.

Failing checks:

  • native-queue-policy

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.

hive/awaiting-ci: merged v6 into the branch (it was 12 behind); CI pending — sweep will merge when green.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: 3eb2dd1

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@github-actions

github-actions Bot commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

Mutation testing

Head 3eb2dd102.

⏱️ The 900s mutation budget ran out, so 4 package(s) were not fully mutated and the results below do not cover them:

  • pkg/inbox: skipped (not started)
  • pkg/knowledge/connector: skipped (not started)
  • pkg/meetings: skipped (not started)
  • pkg/spektacular: skipped (not started)

❌ gremlins did not produce a report for 1 package(s), so no verdict can be derived from this run (see the job log):

  • pkg/dashboard: gremlins exited 137

9 mutant(s) in changed files are not reached by any test (NOT COVERED). A maintainer may apply the mutation/acknowledged label to accept this.

File Line Mutation
src/cmd/hive/main.go 3815 CONDITIONALS_NEGATION
src/cmd/hive/main.go 3820 CONDITIONALS_NEGATION
src/cmd/hive/main.go 3824 CONDITIONALS_NEGATION
src/cmd/hive/meetings_dashboard.go 25 CONDITIONALS_NEGATION
src/cmd/hive/meetingswire.go 125 CONDITIONALS_NEGATION
src/cmd/hive/meetingswire.go 182 CONDITIONALS_NEGATION
src/cmd/hive/meetingswire.go 200 CONDITIONALS_NEGATION
src/pkg/config/meetings_config.go 233 ARITHMETIC_BASE
src/pkg/config/meetings_config.go 234 ARITHMETIC_BASE

Verdict: error (#17349).

@hivecommons-hive

Copy link
Copy Markdown
Contributor

scanner triage: native-queue-policy and DCO fail, and the branch is 3 commits behind v5 (baseline check: not shared, action MERGE_BASE). Merge v5 in (signed off) and re-check. This PR isn't hive-authored, so I'm not pushing.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive

Copy link
Copy Markdown
Contributor

Scanner: native-queue-policy red is a stale-branch signal (baseline helper action MERGE_BASE, PR is 5 behind v5; siblings #17512/#18273 also behind), and DCO is ACTION_REQUIRED. This branch is not hive-authored and is marked requires_human, so I am not pushing — a maintainer should merge v5 into it with sign-off.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive

Copy link
Copy Markdown
Contributor

scanner triage: native-queue-policy and DCO are failing, and the baseline check says MERGE_BASE (5 commits behind). Please merge the base in (signed off) and re-check; not pushing to a non-App branch.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive

Copy link
Copy Markdown
Contributor

Scanner triage: native-queue-policy and DCO fail on a stale branch (5 behind base; helper verdict MERGE_BASE, not shared). Authored by a human so I won't push; please sync with the base branch (signed-off merge) and re-check.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
@hivecommons-hive

Copy link
Copy Markdown
Contributor

hive/scanner: build-gate is red because this branch is 9 commits behind v5 (baseline triage: MERGE_BASE, not shared). Please merge origin/v5 (signed off) and re-check; this PR is not hive-App-authored so I am not pushing.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive

Copy link
Copy Markdown
Contributor

hive/scanner: DCO reports action_required on f5d6b8f (the bot's merge-of-v6 commit), which fails native-queue-policy. Branch is also 9 behind v6. Deferring: this is a multi-commit maintainer top-up, so a signed-off base merge or DCO approval by a maintainer is needed rather than a blind rewrite.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive

Copy link
Copy Markdown
Contributor

scanner: baseline triage says MERGE_BASE: the branch is 10 commits behind v6 and native-queue-policy fails on stale branches (sibling PRs are equally behind), plus DCO is red. Suggest merging v6 into this branch with sign-off (git merge -s ... --signoff) and re-checking. Not pushing to a human-authored branch.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

@hivecommons-hive

Copy link
Copy Markdown
Contributor

hive/scanner: failing DCO and native-queue-policy; branch is 10 commits behind v5 (baseline triage: MERGE_BASE, not a shared incident). This branch is not App-authored, so I am not pushing: please merge v5 in with sign-off (git merge -s ort --signoff origin/v5) and re-check.


🐝 Hive Agent: scanner | Instance: hosted-available-oke-11-placeholder-r05x | SHA: unknown

— hive: agent=scanner backend=claude model=claude-sonnet-5-5 claude=2.1.325

Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>

@hivecommons-hive hivecommons-hive Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Delta re-review at 3eb2dd1 (correctness)

  • The new PruneClaudeTmp code is correct. tmpprune.go:41-76 reuses pruneHomeTarget and newestModTime, which uses WalkDir and does not follow symlinks (homeprune.go:128). It only acts at claude-*/<project>/<session> depth. A missing root or maxAge<=0 is a no-op. The tests cover an idle session, a live session with a deep write, symlinks, and the no-op cases.
  • Still present: native-queue-policy fails. The log says DCO: action_required on 3eb2dd10 ... fix it on this PR. The head commit is a bot-signed merge, so the probot DCO check is red. Re-sign the head with your own sign-off, for example git rebase --signoff or git commit --amend -s on the merge, then push. The check should re-run on the next push.

— hive: agent=adjudicator backend=claude model=claude-sonnet-5-5 claude=2.1.325

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

dco-signoff: yes Indicates the PR's author has signed the DCO. size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants