Repository navigation
🌱 chore: top up v6 from v5 (2026-11-17) - #18620
Conversation
The sign-in page (where /activity and other signed-in views send anonymous visitors), the two access-needed pages and the link-preview page now carry the shared LF Projects legal footer too. Signed-off-by: Andy Anderson <andy@clubanderson.com>
The scanner reached owner-session PRs as unattributed FIX-BEFORE-NEW rows (they default to scanner) and pushed its own commits onto branches another worker was still pushing to, turning 'Sign-off survives the squash' red (#18419), racing the worker's next push and duplicating its fix. A red or conflicted PR the hive App did not open is now listed as 'suggest, do not push' in FIX-BEFORE-NEW, the CI_FAILING push queue and PR_LIST while its branch is busy: a live claim by another worker on the PR or a linked issue (worker-claim ledger), or a head pushed within HIVE_SCANNER_BUSY_BRANCH_MINUTES (default 60, from the head's earliest CI activity, no new GitHub calls). App-authored lane branches, in-place adoptions and idle branches are listed exactly as before. Fixes #18488 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 v5.288.0 entry. See src/docs/releases.md. Signed-off-by: hive-release-bot <actions@github.com>
release: v5.288.0
…tatus (#18498) * ✨ agentsmd: raise AGENTS.md cap to 32 KiB, add per-repo override and status Raise the default AGENTS.md injection cap from 8 KiB to 32 KiB so a 13 KB file reaches the kick in full. Add repos.<repo>.agents_md_max_bytes as a per-repo override over project.agents_md_max_bytes, an env default (HIVE_AGENTS_MD_MAX_BYTES) and a 256 KiB hard ceiling. Log the truncation warning once per file content and cap instead of every kick, and expose each repo's file size, effective cap and dropped sections in the /api/status repos payload. Fixes #18495 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 📖 docs: refresh api-reference citations after server.go shift Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🌱 testutil: allowlist crux/agentsmd in the dashboard import ratchet Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
… drafter (#18520) Describe the reference-repo credential state for a hive whose App is only on cncf/techdocs (read-only until the owner grants access), note that the Accepted state wears stage/020-claimed without a claim (#18434), and document how a hive owner starts the drafter by hand since nothing dispatches the techdocs-drafter profile yet. Signed-off-by: Andy Anderson <andy@clubanderson.com>
Vexa refuses a Zoom bot without meeting_url (it cannot build a Zoom URL from the id). Quorum v0.2.1 (hivecommons/quorum#27) sends the https join link for every Zoom meeting. Bump QUORUM_VERSION and QUORUM_SHA256 from the release's quorum.mjs.sha256; @snazzah/davey stays at 0.1.12. Signed-off-by: Andy Anderson <andy@clubanderson.com>
…18518) * 🐛 Stop the in-progress sweep stripping workflow-model stage labels A repository bound to a workflow model derives its stage through the model; TechDocs 'Accepted' (number 20) wears stage/020-claimed. The in-progress sweep read that as a claim, found no start mark and removed it; the stage sync re-added it two seconds later, every cycle (cncf/techdocs#395, ~298 flips). The sweep now leaves a trigger label the bound model derives for the issue alone and never withholds the issue for it (a live start mark still withholds). The default FactSink derives through the same per-repository model lookup as the sweeps, installed from the reconciler's options. Fixes #18434 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 Install the stage workflow lookup in configureGitHubClient The client-hook guard requires every Set* hook to be applied where the client is (re)built, so a rebuilt client keeps it (#9614). Signed-off-by: Andy Anderson <andy@clubanderson.com> Co-authored-by: nate-double-u <4453979+nate-double-u@users.noreply.github.com>
Automated release commit. Compiles changelog.d/ fragments and moves the CHANGELOG.md Unreleased section into a dated v5.289.0 entry. See src/docs/releases.md. Signed-off-by: hive-release-bot <actions@github.com>
release: v5.289.0
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
* ✨ prompt sources: owners can approve their own managed repos (#18490) Adds an owner-only path to allow one of the hive's project.repos for agents' prompt_source and definition_source without editing the config volume: Settings -> Security -> Prompt sources plus GET /api/config/prompt-sources and PUT /api/config/prompt-sources/owner-repos. Approvals require acknowledge_trust, cover one repo only, are audited, persist to variables.security.owner_github_prompt_repos via Save, and lapse if the repo leaves project.repos. GitHubPromptAllowed keeps its signature and admits seed or owner-approved repos. Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 prompt sources: update runtime golden, avoid dialog-ratchet false match Signed-off-by: Andy Anderson <andy@clubanderson.com> * 📖 openapi: document prompt-sources routes Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
…, hive audio retention note (#18500) * 🐛 Inbox recording: saved panel follows the transcript, plain previews, hive audio retention note - The Saved to the Inbox panel updates to "Transcribed. See the item below." (or the failure/off reason) and auto-dismisses with the toast countdown (toastCountdownBar, TOAST_DURATION_MS, pause on hover/focus); Close stays. - Inbox excerpts drop the machine-transcript provenance line and the untrusted-inbox-data envelope (display only; the fact keeps them). - hive-recording and upload transcripts get a provenance line without the "replaced when the source produces its own" clause. - Items expose audio_expires_at; the player note for hive-stored audio says it is kept until that day, then deleted. Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 Inbox audio stream test: load inboxAudioNoteText into the sandbox Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
Adds bd topic claim/get/list/release backed by the agent's bead store.
A second open claim for the same topic is refused; the governor follows
a claimed PR's renamed head branch, releases the claim when the PR
merges or closes, lists claims in the kick (and as ${TOPIC_CLAIMS}),
and tags claimed PRs in the FIX-BEFORE-NEW and held-PR blocks.
Fixes #18493
Signed-off-by: Andy Anderson <andy@clubanderson.com>
* ✨ Inbox: delete hive recordings
DELETE /api/inbox/hive-recording/{id} removes a finished recording's
stored audio and chunks, its Knowledge fact (machine transcript), its
transcript record and the Inbox item state. Owners delete any recording,
read-write members only their own, read never; 409 while in progress or
being transcribed. Shared copies are not touched. Audited as
inbox_recording_delete. The Inbox shows 🗑 Delete with an in-page confirm.
Signed-off-by: Andy Anderson <andy@clubanderson.com>
* ✨ Inbox recording delete: literal route string for the citation guard, inbox-item test loads inboxCanDeleteRecording
Signed-off-by: Andy Anderson <andy@clubanderson.com>
* ✨ Inbox recording delete: inbox package tests for DeleteRecording, Store.Forget and TranscriptStore.Delete
Signed-off-by: Andy Anderson <andy@clubanderson.com>
---------
Signed-off-by: Andy Anderson <andy@clubanderson.com>
… image (#18519) * ✨ Supply hive's sandbox relays (hive-review) on every sandbox launch Job runtime: an init container from the running hive image (by digest, read from hive's own pod) copies hive-review into an emptyDir that the kick container mounts read-only at /opt/hive/relays. The init container is restricted-SCC safe: runAsNonRoot, read-only root filesystem, no privilege escalation, all capabilities dropped; the uid is left to the SCC on OpenShift and is 65534 elsewhere. job.relay_image and job.relay_run_as_user override, relay_image: none opts out. Podman: hive stages its own hive-review in a private temp dir and bind-mounts it read-only. Both wrap the kick command in a /bin/sh preflight that fails the kick (exit 78) with a clear message when bash or python3 is missing, and prepends the relay dir to PATH. Fixes #18512 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🌱 Split kubejob's API client helpers into client.go (file-size ratchet) Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🌱 Simplify Podman relay staging and cover its temp-dir failure Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
…ackend) (#18516) (#18521) * ✨ reference repos: per-hive write identity (bot, bot only, owner) (#18516) project.remote_write_identity picks whose name a granted write-tier reference repo's issues, PRs and comments carry: - bot (default; a missing key reads as bot): the hive's GitHub App bot on a repo whose owner installed the App themselves (GET /repos/{o}/{r}/installation with the App JWT, cached like org discovery), else the owner's connected account, else read-only. - bot_only: the App bot where installed, read-only elsewhere; the owner token is never read. - owner: the owner's connected account everywhere (the #18197 behaviour). The setting never widens a grant: only entries already at the write tier (hand-written or derived from an accepted request issue, #18139) get a token source, so an App installation shared by many hives is not a grant by itself. The bot token is minted on the installation covering that one repo, scoped to that repo and to the tier's permissions (metadata read; issues write; contents and pull_requests write for prs), cached until near expiry, and injected only by the existing reference-repo proxy path. A transient lookup failure serves the entry read-only for the pass (state unresolved) instead of silently switching identity. write_token_file still wins on hand-written entries. Every proxy refusal is unchanged; a bot login never names a fork, so bot PR branches go to the repo itself under the default-branch guard. The kick names the bot flow (no fork) beside the owner fork flow, the request-issue read-only notice names no cure under bot_only, and the own-item follow-up poller keeps to owner-identity repos. Refs #18516 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 reference repos: keep a hand-written entry's write opt-in on a transient App lookup (#18516) The resolved list replaces the config in memory and the editor saves from it, so clearing Write on a blip would persist as an edit. The entry keeps Write, is marked unresolved, and gets a read-only proxy policy for the pass; the kick says it is read-only this pass. Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
…ssions) (#18523) Signed-off-by: Andy Anderson <andy@clubanderson.com>
…ut, and read cleanly (#18527) * 🐛 dashboard: Meetings and Calendar wait for the role, sit in the layout, and read cleanly - Meetings and Calendar rendered their not-allowed notice before /api/role resolved and never re-rendered; they now show a loading state and subscribe to the role listeners gh-auth.js notifies (same contract as the Inbox fix, #18442). A network failure says the hive is unreachable instead of leaving "Loading meetings…". - Both sections were missing from DASHBOARD_LAYOUT_TEMPLATE. The generated page parses every section outside #hive-dashboard-root and layout.js moves only template sections back, so these two stayed outside the root where --dash-card-padding is undefined: their cards had zero padding. Add them after the Inbox, matching the sidebar order. - Meetings form: link and title fields, then the Voice toggle beside its checkbox and the primary button; Discord disclosure aligned with the form. Meeting rows: title, kc-pill state, platform and time, a plain-language error with the raw error under details. Empty states use empty-dashed. Calendar sub-sections get spacing. Existing tokens and classes only. Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 dashboard: collapse Calendar and Meetings by default like the Inbox; 24 layout sections Now that both are layout sections, the collapse-defaults classification requires every non-critical section in DASHBOARD_DEFAULT_COLLAPSED_SECTIONS (only for new browsers; existing ones keep their state, and the sidebar expands the section it opens). The stale-layout normalize test counts 24 sections. Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
…one review (#18537) * ✨ review: post reviewer findings as inline line-anchored comments in one review hive-review gains --comments-file: a JSON array of {path, line, side, start_line, body}. The review-request watcher submits the summary, the verdict event and every anchorable comment in ONE create-review call, so each finding is its own resolvable thread and the PR gets one notification. Comments whose line is not in the PR diff are folded into the summary; a 422 from GitHub resubmits once with everything folded. Edit-in-place paths and the per-head status-card fold carry comments in the body. Comment bodies get the same redaction, canary and mention guards as the summary. reviewer-queue.md and the dispatch publish prompt tell the reviewer to use it. Fixes #18531 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 review: parseHunkHeader returns zeros on a malformed header Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
…he reviewer promptly (#18539) The eval cycle now keeps each seen PR's label set and, when an already-seen PR gains labels, emits one pr.labeled event with the new labels in event.added_labels. A rule such as event.kind == "pr.labeled" && "hive-review" in event.added_labels then kicks the reviewer on the next eval after a maintainer opts a PR into review, instead of waiting up to a full cadence. First sight stays pr.opened, so a restart does not replay old labels. Fixes #18538 Signed-off-by: Andy Anderson <andy@clubanderson.com>
…file paths (#18540) * 🐛 inference: refresh rotated gateway keys and refuse ignored api_key_file paths The inference proxy resolved the LiteLLM/gateway key once, when an agent's route was set, so a rotated Secret kept 401ing until restart (#18532). Routes now carry an InferenceKeyResolver that re-stats the key file every few seconds (symlink target, mtime, size: a Secret update swaps ..data), re-resolves on change or every 30s, and on an upstream 401 re-resolves once and retries once with a changed key before counting an auth failure. The auth-failure WARN gains key_source and key_changed_since_start; no key material is logged. An api_key_file outside the allowed secret dirs was silently skipped (#18533). Dashboard/API saves now reject it with a message listing the allowed dirs and the mount remedy; boot warns (existing configs keep booting on their fallback key); runtime logs one WARN naming the skipped path and the effective source; the governor config API and gateway rows expose keySource, keySourceDescription and apiKeyFileAllowed. The audit N8 gate is unchanged. Fixes #18532 Fixes #18533 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🌱 crux/config: move api key resolution out of agent_runtime.go Keeps agent_runtime.go under the crux file-size ratchet baseline. Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
The pinned advisory (governor) issue was only ever ensured for the primary repo: the boot ensure, the per-cycle re-ensure and the digest post all keyed on advisory.PrimaryRepo, so every other repo in project.repos never got one. Ensure every non-primary managed repo each eval cycle until it resolves (repos added in Settings included), store the number per repo, post the same digest to each, and surface repos the hive cannot write to as per-repo dashboard warnings with an hourly retry. Fixes #18448 Signed-off-by: Andy Anderson <andy@clubanderson.com> Co-authored-by: nate-double-u <4453979+nate-double-u@users.noreply.github.com>
A review kick that runs in a sandbox now gets the PR it reviews. Hive reads the targets from the kick's hive-written header (review swarm and reviewer lane), fetches refs/pull/N/head and the base with the clone token, checks the head out on pr-N, and writes .hive/pr/N.json from its own GitHub client, marked untrusted, with metadata, changed files, a capped diff and review comments. The kick prompt names both. Review kicks never push: the manager wires no broker and the executor refuses one. Host writes into the checked-out tree go through os.Root. Fixes #18513 Signed-off-by: Andy Anderson <andy@clubanderson.com>
…e CLI and Settings picker (part 1 of #18517) (#18528) * ✨ meetings: agent-backend model mode with a warm tool-less Claude Code CLI (#18517) Betty and Spektacular live mode can answer through one of the hive's agent backends: meetings.llm.mode agent-backend (meetings.llm.backend, meetings.llm.model) or claude-cli. Each backend runs in a narrow completion profile (single turn, no tools, no MCP, no hooks, no settings files, no session, empty temp cwd, fixed system prompt, usage from the CLI result). Claude runs one warm stream-json process per meeting for Betty, recycled after N turns or on any error, and one claude -p per prompt for Spek's hive meeting-llm. LiteLLM maps onto the OpenAI path; other backends are refused at validation. Usage is charged to the meeting ledger, budget and ceilings. Settings, Meeting attendee gains a Model section fed by the agents' backend and model catalog, and the panel shows the model name. Refs #18517 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 meetings: fix CI for the agent-backend model mode (#18517) Send the CLI turn without HTML escaping, document GET/PUT /api/meetings/model in dashboard/openapi.json, keep backend discovery off the boot path, resync api-reference citations, and give the usage failure test a non-zero usage. Refs #18517 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 v5.289.1 entry. See src/docs/releases.md. Signed-off-by: hive-release-bot <actions@github.com>
release: v5.289.1
…on-needed (#18542) * 🐛 hive never applies hold itself; gates use ready-for-review / decision-needed (#18468) A hold placed by a person still means hands-off: the hive never implements, decomposes, relays or merges a held item and never removes the label. The hive's own gates no longer overload hold: - ACMM level gate (and cross-repo outreach) -> ready-for-review - #5117 self-authorization, #9665 reporter trust, holdguard re-hold, owner-only delegated decision -> decision-needed - SHA-hold sweep -> needs-sha for every reporter Both gate labels block merges exactly like hold (exact-match members of the hold set), are created with descriptions saying who acts next, are reserved from agents, and are renameable via governor.labels.ready_for_review / decision_needed. A boot migration swaps hold for the gate label only when the newest hold event is the App's and an App-authored gate notice is on the thread; every other hold is left alone. Fixes #18468 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 fix ambiguous ListOptions.Page and drifted docs citation (#18468) Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 gate the hold migration writes, update golden and stale test (#18468) Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🌱 drop unused test field (#18468) Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 update watcher and level-notice tests for ready-for-review (#18468) Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🌱 gofmt (#18468) Signed-off-by: Andy Anderson <andy@clubanderson.com> Co-authored-by: nate-double-u <4453979+nate-double-u@users.noreply.github.com>
…he status card (#18456) * ✨ Sentinel: fold informational notices into the PR status card In quiet notification mode, the sentinel's informational notice on a trusted author's PR (no label, merge not blocked) becomes the sentinel section of the hive:status card, edited in place. A hold still posts its own alert comment in every mode and drops any stale informational section. Normal mode is unchanged. Fixes #18453 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 Sentinel card test: use shortSHA for the head prefix Signed-off-by: Andy Anderson <andy@clubanderson.com>
… operator-token Claude, refusal reasons (#18549) (#18556) Agent-backend mode now covers every backend the agents' picker offers: - copilot (1.0.88) and omp (18.20.0) answer through one tool-less CLI per call with the agents' sign-in, prompt on stdin, usage charged to the meeting ledger; copilot runs in a throwaway home. - vllm, llm-d and named model gateways map onto the OpenAI-compatible path (gateway key, default model and CA bundle). - A hive with only HIVE_CLAUDE_OAUTH_TOKEN passes it to the meetings CLI as CLAUDE_CODE_OAUTH_TOKEN, as the agent launcher does. - codex, gemini, watsonx and the other CLIs are refused at validation, in PUT /api/meetings/model and in the Settings picker with the reason. Signed-off-by: Andy Anderson <andy@clubanderson.com>
…est spool (#18586) * ✨ Sandbox request spool carries issue and topic requests Generalize the sandbox review spool into a typed request spool. Envelopes now name a kind (review, issue, topic); every kind but review binds its name into the MAC, so a request signed as one kind is refused as another, and the seen-MAC replay guard covers all kinds. Kindless envelopes are reviews MACed as before, so hive-review is unchanged. - issue requests run through the issue relay's own handler with every gate, attributed to the manifest's agent and authorized by the issue authorizer installed for the spool. - topic requests are applied to the manifest agent's bead store. A claim without a PR is pending on the kick's branch (now recorded in the manifest) and the governor binds it to the PR opened from that branch. - hive-open-issue writes signed requests to the spool when HIVE_SANDBOX_REQUEST_DIR and HIVE_SANDBOX_SPOOL_KEY are set; the new hive-topic relay mirrors bd topic and falls back to bd in-pod. - Both ship on the sandbox PATH (Podman bind mount, Job init container). Fixes #18572 Fixes #18573 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 📖 Refresh docs citation shifted by the spool fields Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 Fix CI: generated workflows, duplicate test helper, relay-owned setter Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 Track the issue relay's write site under its new name; fix topic refusal test Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
…dited to their agent (#18587) * ✨ sandbox: push fix kicks to the agent's own PR and credit sandbox PRs A sandboxed fix kick that names one of the agent's open PRs (explicit [hive:target-pr] line, red-PR or red-CI kick header, first FIX-BEFORE-NEW row, or a single topic claim) now checks out that PR's head branch and the push broker adds the kick's commits to it. The push needs the PR open, same-repo, credited to this agent by its agent_pr_created record, still at the checked-out head (re-read before the push and --force-with-lease in the push), and the usual broker scans to pass; otherwise the kick fails with the reason and nothing is pushed. A PR a sandboxed kick opens is credited like a relay-opened PR: attribution trailer, agent_pr_created audit entry and PR-opened hooks through github.Client.CreateAgentPR, or a runtime-emitted agent_pr_created entry when the PR client cannot record it. The governor's head-ref fallback reads the agent out of hive/<agent>/<stamp> work branches. Fixes #18574 Fixes #18575 Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🌱 pushbroker: move workspace config guard to its own file pushbroker.go passed the crux 800-line limit with the lease support. Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com> Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
"Betty, how many open issues are there?" went to the research pool, which only reads the public web. The persona now has a hive-status retriever backed by dashboard.Server.HiveStatusBrief, built from the same in-memory status snapshot as the Overview and the grounded chat (chatBuildSnapshot): open/actionable/held/blocked counts, agents, working and blocked agents, recent completions and open PRs. A question about the hive gets that snippet, the prompt says to answer from it, and no deep dive is started. Part of #18590 Signed-off-by: Andy Anderson <andy@clubanderson.com>
The panel only showed a Spektacular status once the child's lines channel closed (after it gave up restarting), so a meeting whose Spek child was down showed nothing while every forward failed with "stream child is not running". A failed forward now sets a plain "Notes and clarifying questions are paused" status, cleared when Spek writes again; the final status leads with "Notes and clarifying questions are off for this meeting". Part of #18590 Signed-off-by: Andy Anderson <andy@clubanderson.com>
…8599) The spoken ack waited for the final transcript, which on Google Meet comes after the turn closes and the speaker's drafts are reconfirmed, so it landed about 6 s after she was asked (#18590 finding 4). Partial segments now reach the persona (ObservePartial, filtered before the queue); once a partial names her with a request of at least two words and voice is wanted, she speaks the ack at once and the final line gets the answer without a second ack. An unaddressed final clears the mark; Spek and open asks still see only final lines. Part of #18590 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 v5.292.0 entry. See src/docs/releases.md. Signed-off-by: hive-release-bot <actions@github.com>
…leave (#18594) * 🐛 Meetings: consent counts as posted only on a confirmed delivery; speak it or leave In a live Teams test the bot could not open the meeting chat (chat_send_failed reason=no_composer) yet hive set consent_posted_at, because Quorum's chat.post answered as soon as Vexa accepted the line. The consent notice is now posted with chat.post confirm:true (quorum v0.2.2, bumped in the image) and counts as posted only on delivery confirmed; an older Quorum that cannot say is treated as not posted. When chat cannot deliver it and the meeting has voice on, Betty speaks the notice (consent_via=voice); when neither works she leaves at once with state reason consent_undelivered and the Meetings panel says why. Transient failures (not admitted yet, rate limit, dropped call) are retried up to three times. Part of #18590 (finding 11). Signed-off-by: Andy Anderson <andy@clubanderson.com> * 🐛 Meetings: consent tests follow the consent-first rule TestConsentErrorRecordedAndThrottled encoded the old behaviour (chat platform_unavailable, no voice: stay and keep listening). The throttle is now exercised with a transient rate limit, and a new test pins the leave. Part of #18590 (finding 11). Signed-off-by: Andy Anderson <andy@clubanderson.com> --------- Signed-off-by: Andy Anderson <andy@clubanderson.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com> Signed-off-by: scanner <280983584+hivecommons-hive[bot]@users.noreply.github.com>
…he ingress page (#18601) The hosted ingress intercepts 502/503 and serves its generic 'hive unavailable' page, so a 502 the spoke answered with its own error read as an outage. On a request carrying the hub's X-Hive-Proxy-Auth proof, a JSON 502 from an /api/ handler now goes out as 424 (passed through by the ingress) with X-Hive-Status: 502. Unreachable spokes still get the generic page; 503 is unchanged because the UI reads spoke 503s by status. Refs #18598 Signed-off-by: Andy Anderson <andy@clubanderson.com>
…ne, docs (#18516) (#18603) - Settings → Repos → Reference repos: owner-only three-way control (Hive bot, falling back to me / Hive bot only / Me) with one provenance line; every write entry says which identity posts there and why ("posts as hivecommons-hive[bot]: App installed on <org>", "posts as <owner>: App not installed on <org>", "read-only: connect your GitHub account", "read-only: bot only, App not installed on <org>"). Owner-panel copy no longer says remote writes are never the App. - PUT /api/config/reference-repos/write-identity: validates, dry-runs the hand-written entries under the new value (refuses what boot would stop on), saves, re-resolves synchronously, audits config_remote_write_identity (sensitive). GET /api/config/reference-repos carries write_identity, write_identities and bot_login; entries and /api/status carry credential_identity and credential_note. - Repos card chip names the bot identity; unresolved chip restored. - Docs: two-lane section (whose name the writes carry), operator reference row, API reference, openapi. Settings-parity allowlist entry removed. Fixes #18516 Signed-off-by: Andy Anderson <andy@clubanderson.com>
release: v5.292.0
Merge origin/v5 (db0afc2, #18595) into v6 (3d13bc1, v6.40.1). Carries #18388 (needs-triage label loop fix) among 60 v5 commits. Conflicts resolved by hand: v6's split config layout gets v5's config.go changes ported into the matching files; v6-only sandbox kick-observer wiring, layout sections, hold-guard and autofix Dependabot handling are kept alongside v5's fixes. CHANGELOG union and static/index.html regeneration follow v6-topup-autoresolve.sh; TRUSTED_SCRIPTS_TREE re-pinned to the merged src/scripts tree. Signed-off-by: Andy Anderson <andy@clubanderson.com>
Second merge for the v5 commits that landed while the first was being resolved (v5.292.0: #18594, #18599, #18601, #18603, #18612). audit.go gains config_remote_write_identity; server.go wraps v6's request-log/REST-caller chain in v5's hosted gateway passthrough; api-reference citations re-pinned; CHANGELOG union and index.html regeneration as before. 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 |
v5's sidebar-collaborate-group test (#18562) lists the Dashboard group without v6's runs-section link. Signed-off-by: Andy Anderson <andy@clubanderson.com>
Signed-off-by: Andy Anderson <andy@clubanderson.com>
|
Hive status — updated in place; edits do not notify. Full history is in the Activity Monitor. ℹ️ Sentinel: informational, author is trusted (owner), merge is not blocked. Head
No @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. Evidence bundle `#186200bb7ac5` sha256:696e730e67d4c9d3afa9954651d1e875e4a24693f2b034eebe372995efede471 — download from the dashboard or `hivectl review evidence #18620`.merged by Hive trusted-author auto-merge: author |
…okIssue signatures Signed-off-by: Andy Anderson <andy@clubanderson.com>
Mutation testingHead ⏱️ The 900s mutation budget ran out, so 21 package(s) were not fully mutated and the results below do not cover them:
❌ gremlins did not produce a report for 2 package(s), so no verdict can be derived from this run (see the job log):
gremlins found 5 surviving mutant(s) in changed files: the code ran under tests, but changing it did not fail any test.
27 mutant(s) in changed files are not reached by any test (NOT COVERED). A maintainer may apply the
ℹ️ 4 mutant(s) on added lines were SKIPPED because gremlins v0.6.0 only honours the first
Verdict: |
This PR merges
origin/v5intov6(3d13bc1, v6.40.1) as two signed merge commits. The first carries v5 through db0afc2 (#18595). The second carries five more v5 commits through b2a1c14 (#18619, v5.292.0) that landed while the first was being resolved. The v5 range topped up is17ecea2b3..b2a1c149f(67 commits). After it lands,origin/v5is an ancestor ofv6again (checked withgit merge-base --is-ancestor). The automatedv6-topup.ymlrun has failed on every v5 push since the last top-up because of conflicts outside the resolver's known patterns, so I carried it by hand.cc @Danathar: this carries #18388 (needs-triage loop fix) to v6.
What it carries
Everything on v5 since #18508, including:
needs-triageonce, then yield it to the repo's own triage tooling (merge commit 2ae44fb).holdretirement: the hive appliesready-for-review/decision-neededand neverhold(🐛 hive never applies hold itself: gates use ready-for-review / decision-needed #18542, 🐛 needs-sha keeps hive off an issue until the reporter adds a SHA #18563).pr.labeledtrigger (✨ triggers: emit pr.labeled so adding the review opt-in label kicks the reviewer promptly #18539), the Collaborate sidebar group (✨ Dashboard: Collaborate sidebar group (Inbox, Calendar, Meetings, Sessions) #18523, 🐛 Dashboard: keep the Collaborate sidebar group and place it after Dashboard #18562), and the Meetings, Inbox and Betty updates.Conflicts and how they were resolved
src/pkg/config/config.goReposandWorkflowConfig.Dispatchwent intoconfig.go,OwnerGitHubPromptReposintovariables_config.go, theAgentsMdMaxBytescomment andRemoteWriteIdentityintoproject_config.go, andReleaseFeedsand theReadyForReview/DecisionNeededgate labels with their helpers intogovernor_config.go.PerRepo,AdvisoryPerRepoEnvVarandPerRepoEnabledwent intoadvisory_config.go,ResolveLiteLLMInferenceKeyWithSourceintogateway_config.go, and the owner-approvedGitHubPromptAllowedintoload.go.crux/config/agent_config.gosandbox_runtime.go. I added v5'sRelayImage/RelayRunAsUsertoSandboxJobConfigthere. v5's newsecret_file_reporting.goredeclaredSecretFileRoots, which v6 already defines inagent_secret_files.gowith the same body, so I dropped v5's copy.crux/agent/manager_sandbox.go,sandbox_executor.go,crux/sandbox/sandbox.gorunSandboxKick(agent, gen, ...), theonStarted/OnStartedacceptance hook and the kick-observer archive notify. Added v5'sprReader/targetReaderplumbing,sandboxKickActive = false,creditSandboxPRandRelays.cmd/hive/github_client_config.gorecordPRTerminalObservedAuditand added v5's topic-claim release in front of it.cmd/hive/main_helpers.gopkg/github/automerge/automerge_held_runs.go(+test)required-check-never-ran-autofix, no comment posted, handed to a person). v5's #18441 posts nothing either. The rest of #18441 (dependabotUpdateBranchAllowed,resolveConflictingDependabotPR) merged in.automerge_sweep_trusted.go!laneBehindandreject(...), and added v5's conflicting-Dependabot resolver in front of them. v5's update-branch comment is ported.automerge_stage_human.godecisionActionsmoved tostagelabels. I ported v5's change tostagelabels/human.go: a self-imposed gate label (ready-for-review/decision-needed) gets the same release step as the hive's ownhold, naming the label.pkg/github/client.gohold(), since the SHA sweep now appliesneeds-sha(#18563).advisory.go,sentinel_sweep.gomarkerBodyplus v5's comment.notification_write_sites.go,docs/notifications.mdhandleOneIssueRequestFrom.pkg/scheduler/*DeferredExternalandtaskMCPURL/agentBackend/laneDepthsfrom v6;AppAuthored/HeadPushedAt/Issues,topicClaimsandnewReleasesfrom v5).templates.gokeeps v6's guidance gate and callsbuildCIFailingListForAgent. Inkickmessage.go, v6's refs-only listing also gets v5's busy-branch annotation.kick_priming.goimports are a union.layout.js,static_index_test.goanddashboard_section_reorder_9062_test.go, v6's order (runs-section,search-section) gets v5'scalendar-sectionandmeetings-sectionafterinbox-section.sidebar.htmlkeeps v6's Search link, and the Inbox, Calendar and Meetings links move to v5's Collaborate group.api_agent_config.gokeeps both.api_repo_item_decide_delegated.gokeeps v6'sholdLabels(it now includes gate labels).audit.goaddsconfig_remote_write_identity. Inserver.go, v5'swithHostedGatewayPassthroughwraps v6's request-log and REST-caller chain.maxDashboardInternalImportsgoes from 73 to 74 (pkg/repoagents).dashboard_import_allowlist.txtandsettings_parity_allowlist.txtare unions.validate.go,crux_aliases.goci-maintainer/sec-checkholdgated, both copies)--needs-decisionpark line and applied v5's "holdorready-for-review" wording.check-api-reference-citations.sh --fix. The stale copy of theGET /api/config/reference-reposrow was dropped in favor of v5's text. The prose docs (cel-triggers,operator-reference,sweeps-and-operations,dashboard-glossary,workflow-and-agents,agent-configuration) keep v6's text with v5's additions.check-docs-citations.pyreports clean after--fixand one hand re-pin inscale-envelope.md.CHANGELOG.md,static/index.html, pinindex.htmlis rebuilt fromstatic-src/manifest.txt.TRUSTED_SCRIPTS_TREEalready equals the mergedsrc/scriptstree (46005db3, unchanged from v6).Nothing was built or tested locally.
gofmt -lis clean on every hand-edited Go file. CI is the gate.🤖 Generated with Claude Code