deps(actions): Bump softprops/action-gh-release from 2 to 3#3
Open
dependabot[bot] wants to merge 1 commit into
Open
deps(actions): Bump softprops/action-gh-release from 2 to 3#3dependabot[bot] wants to merge 1 commit into
dependabot[bot] wants to merge 1 commit into
Conversation
Author
LabelsThe following labels could not be found: Please fix the above issues or remove invalid values from |
dependabot
Bot
force-pushed
the
dependabot/github_actions/softprops/action-gh-release-3
branch
from
May 7, 2026 05:12
08b650a to
5cc2bfd
Compare
Bumps [softprops/action-gh-release](https://github.com/softprops/action-gh-release) from 2 to 3. - [Release notes](https://github.com/softprops/action-gh-release/releases) - [Changelog](https://github.com/softprops/action-gh-release/blob/master/CHANGELOG.md) - [Commits](softprops/action-gh-release@v2...v3) --- updated-dependencies: - dependency-name: softprops/action-gh-release dependency-version: '3' dependency-type: direct:production update-type: version-update:semver-major ... Signed-off-by: dependabot[bot] <[email protected]>
dependabot
Bot
force-pushed
the
dependabot/github_actions/softprops/action-gh-release-3
branch
from
May 19, 2026 10:17
5cc2bfd to
89fe36c
Compare
physercoe
pushed a commit
that referenced
this pull request
May 23, 2026
…ilent because agy stopped writing last_conversations.json (v1.0.645-alpha)
Third W11 smoke pass: mobile feed shows nothing even when agy is
actively responding. Forensic walk of the on-host state:
- Conversation 5536eee5-… was actively writing transcript_full.jsonl
(242 KB, modified within seconds of the smoke).
- agy's cli.log shows it forwarding user messages to this conv id
("Forwarding user message to conversation 5536eee5-…").
- BUT ~/.gemini/antigravity-cli/cache/last_conversations.json was
last modified 2026-05-22 04:30 — yesterday's probe runs. No entry
for /home/ubuntu/hub-work/antigravity.
## What changed in agy's behaviour
The workdir got promoted to a "project" the first time the user
accepted the trust dialog — agy created
~/.gemini/config/projects/95fe7eb6-….json with name=workdir, plus a
workspace-local symlink at <workdir>/.antigravitycli/95fe7eb6-….json.
For project workdirs, agy stops writing last_conversations.json
entirely (the May-22 cache entries are all from /tmp probes that
weren't promoted). Our adapter was polling a file that would never
update for this workdir → 30-min timeout → no events → silent feed.
## Fix
Brain dir creation is a more reliable signal: agy creates
brain/<convId>/ at the instant it mints a conversation, regardless of
which persistence mechanism it uses for the workdir.
- WaitForConversation gains a `since time.Time` parameter; on each
poll tick it also checks for the newest brain/*/ dir with
mtime > since. Whichever signal (brain dir or legacy cache) fires
first wins. Pre-launch siblings on the same host are filtered out
by the time threshold.
- adapter.resolveAndRun captures `launchTime := time.Now()` at the
top so the brain-dir signal is always scoped to our spawn.
- New newestBrainSince() pure helper — what NewestBrainFallback
used to be (after-timeout last resort, opt-in) is now the
load-bearing primary path.
## Verification
- pathresolver_test.go (new): NewBrainDirSinceWinsWithStaleCache
exercises the exact smoke scenario (yesterday's cache + a pre-launch
sibling brain dir + our spawn appears mid-test → only our conv id is
returned). LegacyCacheStillWorks confirms the back-compat path for
casual workdirs. IgnoresOlderDirs confirms the time threshold.
- All 20 hub packages green; go vet clean; lint-docs + lint-templates
clean.
## Notes
ADR-035 still Proposed; this is W11 fix-up #3. The cache-write
asymmetry is upstream agy behaviour, not a regression on our side.
Sub-second concurrent-spawn disambiguation (cross-check transcript
first-step content) deferred — time threshold is sufficient for
single-user smoke and the typical case.
Co-Authored-By: Claude Opus 4.7 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
May 23, 2026
…(v1.0.659-alpha) ADR-027 W11 fix-up wedge #3 — claude-code M4 boot regression caught on on-host smoke of v1.0.658: settings.local.json hooks fail validation with "Expected string, but received undefined" and claude refuses to launch. Root cause: ADR-027 W6 emitted hook entries of shape {"type": "mcp_tool", "tool": "mcp__termipod-host__hook_*", ...} which was a speculative design — claude-code's actual hook schema only supports {"type": "command", "command": <string>}. The `tool` field is unknown; `command` is missing; validator throws on first load. Hooks have been silently broken since v1.0.592; nobody noticed because most claude-code spawns run M2 stream-json, not M4 LocalLogTail. Fix (rebuild W6 properly): New host-runner subcommand `hook-fire` (internal/hookfire/) — a one-shot stdio bridge that wraps claude-code's hook contract over the existing UDS MCP gateway: - claude spawns `host-runner hook-fire --socket <uds> --event <Event>` - writes hook payload (single JSON object) to stdin, closes stdin - shim wraps as JSON-RPC `tools/call` with name=`hook_<event_snake>`, arguments=<payload> - dials per-spawn UDS, half-closes write side, reads one response line - extracts `result.content[0].text`, writes to stdout, exits 0 Failure semantics tuned for blocking hooks: transport blip → stdout "{}" + non-zero exit + stderr warning, same outcome as an uninstalled hook (claude defaults to "allow"). hooks_install.go rewrites appendTermipodMatcher to emit: {"matcher": "", "_termipod_managed": true, "hooks": [{"type": "command", "command": "host-runner hook-fire --socket '<sock>' --event <Event>", "timeout": <n>}]} `_termipod_managed` retained as the strip-then-append key — this lets v1.0.659 spawns self-heal stale workdirs that still hold the pre-v1.0.659 invalid mcp_tool entries. New test TestInstallClaudeHooks_SelfHealsStaleMcpToolEntries locks this. Side effect: the v1.0.657 turn.result emission in hookStop now actually fires (Stop hook reaches the adapter again). v1.0.657 symptom #4 fix becomes load-bearing for real. Tests: - internal/hookfire/run_test.go × 6 (round-trip, gateway error, dial failure, event coverage, CLI usage errors) - internal/hostrunner/hooks_install_test.go rewritten — schema validation, self-heal of stale workdirs, plus the 5 prior behaviour tests adapted to the new command-typed shape - launch_m4_locallogtail_test.go updated — substring assertions now check for "host-runner hook-fire" + "--event PreCompact" instead of the (now-absent) mcp__termipod-host__hook_pre_compact string Co-Authored-By: Claude Opus 4.7 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
May 24, 2026
…(W10, v1.0.683-alpha) ADR-030 Phase 1 W10 — re-address permission_prompt rows raised by steward-parented workers. When a worker has a same-project steward parent, the row lands assigned_tier='project-steward' + current_assignees_json=[<parent_id>] so the parent steward sees it in their inbox first. Otherwise the row stays team-wide-addressed (existing behaviour preserved for orphan workers, non-steward parents, and binding-drift cases). mcp_more.go::permissionPromptAddressee: - Single SQL JOIN with the strict three-clause predicate per ADR-030 pre-W1 decision #3: worker.parent_agent_id IS NOT NULL AND parent.kind LIKE 'steward.%' AND parent.project_id IS NOT NULL AND = worker.project_id - Two IS NOT NULL guards defend against SQL's NULL=NULL → NULL semantics (without them, two unbound rows would accidentally match) - Best-effort: DB errors log a warn + return '' so transient issues degrade to safe (team-wide) mcpPermissionPrompt: - Calls the helper before the INSERT - When steward_id returned: stamps assigned_tier='project-steward' + current_assignees_json=[<steward_id>] - Otherwise: leaves assignees='[]' and assigned_tier=NULL - INSERT widened to include assigned_tier (the ADR-030 W1 column previously written only by the propose path) Tests (6 in mcp_permission_prompt_addressing_test.go): - same-project steward parent → row addressed to parent - cross-project steward parent (binding drift) → stays team-wide (load-bearing regression for v1.0.605-class bug) - non-steward parent → stays team-wide - orphan worker (no parent) → stays team-wide - NULL project_ids both sides → stays team-wide (NULL=NULL guard) - direct helper test: ghost worker returns '' Existing TestDecide_PermissionPromptFansOutAttentionReply still passes — dispatchAttentionReply is untouched. The new addressing only affects which inbox surfaces the row first; the fan-back still routes by session_id. Test fixture note: Write tool is TierRoutine (auto-allow), so tests use Task (TierSignificant) so the gate doesn't short-circuit before the INSERT. ctx timeout 100ms so the call returns ctx-cancelled shortly after the INSERT lands. Plan W10 rewritten to document the helper shape, 5-conjunct JOIN, and 6-test coverage. New verify symbol anchor on permissionPromptAddressee → 23 anchors total. Co-Authored-By: Claude Opus 4.7 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
May 27, 2026
User-direction changes from review of v1 of the doc:
- §5.0 (new grounding section) captures two cross-host facts I had
to verify in the codebase before answering questions about the
fleet model:
* agents.project_id and agents.host_id are independent — two
agents in one project on two hosts is a supported config.
* Cross-host file sharing works today for ≤25 MiB via the hub
blob store. >25 MiB has a design slot (host-served URIs per
blueprint §4) but no host-runner serve-bytes endpoint yet.
Both make A1 (mailbox three-state lifecycle) more urgent than the
abstract pitch suggested.
- OAuth-enabled MCP moves from Tier A to Tier C (deferred post-MVP)
— multi-week spec work for tier-3 ergonomics that bring-your-own-
bearer already covers.
- Auto-compaction preserving task state (former B4) moves to Tier C
— claude-code and codex both ship native /compact surfaces;
layering a hub carry-forward is not urgent until a real
user-visible incident lands.
- Tier A renumbered: A1 mailbox, A2 fallback chains (was A3),
A3 doctor (was A4).
- §6 frontier-signal #3 softened correspondingly — long-session
survival is a shared frontier, not the strongest near-term wedge.
- TL;DR rewritten to match.
No version bump (doc-only).
Co-Authored-By: Claude Opus 4.7 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
Jun 1, 2026
…t (transcript P2 step 1, v1.0.771-alpha) P2 step 1 of docs/plans/agent-transcript-debug-and-header-parity.md. Tester feedback #3: the same agent transcript wears different faces across surfaces. Introduce one shared header so they can't drift. - New widgets/session_header.dart: SessionHeader owns title+subtitle+ optional session chip + a dedicated `View ▾` switcher + caller actions /⋮ + optional ×. Hybrid LayoutBuilder chip placement (inline when it fits, row 2 when tight; title ellipsizes first, controls fixed). - Dense SessionInitChip variant (`dense: true`): permission_mode → a colored shield glyph (full label on tooltip + drawer); t/mcp counts → drawer only. ~halves the chip width for the header's tight budget. - Project-agent sheet adopts SessionHeader: the hand-rolled header Row + separate chip row + engine·host line collapse into one header (host is the subtitle); the visible 4-tab TabBar becomes a View▾-driven IndexedStack (Feed/Pane/Journal/Insights), preserving per-view scroll + state. SessionChatScreen adoption is step 2. Co-Authored-By: Claude Opus 4.8 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
Jun 1, 2026
…cript P2 complete, v1.0.773-alpha) P2 step 3 (final) of docs/plans/agent-transcript-debug-and-header-parity.md. Closes tester feedback #3: the session-detail surface was a feed-only full screen while the project-agent sheet had four tabs. Both now render the same SessionHeader and reach the same four views via `View ▾`. - SessionChatScreen: the Material AppBar becomes a shared SessionHeader (back button + @host subtitle + dense chip + View▾ + propose-inbox/ scope/mode-model pills + session-actions ⋮). Body becomes an IndexedStack: rich AgentFeed (Feed) + shared AgentPaneView / AgentJournalView / InsightsPanel — the views it previously lacked. SafeArea supplies the status-bar inset the AppBar gave. - SessionHeader gains an optional `leading` slot (full-screen back button; the sheet keeps the × on the right). Co-Authored-By: Claude Opus 4.8 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
Jun 5, 2026
…roposer (#3, #4) The W8 ADR-030 refactor moved template install onto the propose-kind registry (apply_template_install) but left templates_propose still raising the legacy `template_proposal` attention kind, and rewired the decide handler to dispatch ONLY when kind=="propose". The result: - #4: approving a template_proposal was a silent no-op — the template never installed, so it looked like a separate "install" step was still needed. - #3: template_proposal was excluded from the decide fan-back, so the proposing steward got no feedback on approve/reject. Restore the W8-intended `alias_legacy` routing without disturbing the mobile template_proposal review UI: - decide handler: load pending_payload_json and, for a template_proposal approve, dispatch through the same template.install Apply the propose path uses (pending_payload already carries {category,name,blob_sha256}). - decide fan-back: add template_proposal to the allowlist and report change_kind="template.install" so the steward's session learns what landed. - mcpTemplatesPropose: stamp the proposer's session_id on the attention row so the fan-back has a delivery target (parity with mcpPropose). The existing TestMCP_TemplatesPropose_ApproveInstalls called installProposedTemplate directly — mocking the very step that was broken, which is why this shipped green. Rewrote it to drive the real /decide endpoint and assert both the install (executed.path + file content) and the fan-back turn (decision + change_kind). Co-Authored-By: Claude Opus 4.8 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
Jun 5, 2026
…P1–P3) Bundles the unreleased hub work since v1.0.803: - ADR-044 P1: agent MCP affordance for deliverables/criteria/phase - ADR-044 P2: governed propose verbs (deliverable.create + criteria.*) - ADR-044 P3: AC-driven auto-advance; retire propose phase.advance - #20: per-phase deliverable/criteria hydration - #3/#4: template-proposal approve installs + fans back Re-verified the two governed-actions contract docs (ADR-030 + the MVP rollout plan) against the retired phase.advance kind + the four new ADR-044 kinds; framework unchanged. Co-Authored-By: Claude Opus 4.8 <[email protected]>
This was referenced Jun 10, 2026
physercoe
added a commit
that referenced
this pull request
Jun 27, 2026
…118 §4) (#296) Only the operator stop path (stopSessionInternal) finalized the run digest at terminal time. A crash/failure (PATCH status=crashed|failed) and a session-less terminate fell through without folding, so the FIRST Insight open on those runs paid the full O(n) backfill (#118 bottleneck #3/#4) on the user's read path. Converge both missing terminal transitions through finalizeDigestOutcome (fold current if stale, then O(1) outcome stamp) — same hook the stop path already uses. Chose finalizeDigestOutcome's idempotent recompute over the background worker's incremental foldDirtyAgent to avoid a double-fold race with that worker. - handlePatchAgent crashed/failed branch -> finalizeDigestOutcome - applyAgentTerminationEffects session-less branch -> finalizeDigestOutcome - test: a crashed agent's digest is sealed (watermark==max, outcome stamped) right after the PATCH, so no read-time backfill is owed Co-authored-by: Ubuntu <[email protected]> Co-authored-by: Claude Opus 4.8 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
Jul 12, 2026
#3 — Clicking Meta (or Enrich, which switches to the Meta tab) blanked the whole app and forced a relaunch. Root cause: RefMeta declared a prop named `ref`, which React reserves and never passes to a function component, so `ref` was undefined and the first `ref.` access threw. Two fixes: - rename the prop to `reference` (aliased to a local `ref`) — the instance; - add a surface-level ErrorBoundary (ui/ErrorBoundary.tsx) around the AppShell surface switch — the class. A surface throw now shows a recoverable panel (Try again / Reload) with the shell chrome intact, instead of a blank app requiring relaunch. #5 — draw.war download failed on-device ('error sending request for url'). The asset exists (v30.3.6, 52.9 MB) so it's a client-side transport failure (GitHub release assets 302 to githubusercontent, throttled on some networks). Add an offline fallback: drawio_install_file picks a locally-downloaded draw.war via the native dialog and extracts it (shared install_war_bytes helper; no network). Also harden the download (explicit redirect policy + User-Agent + clearer error). New "Install from file…" button in the DiagramEditor install screen. Co-Authored-By: Claude Opus 4.8 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
Jul 12, 2026
links, +notes fix, and an assistant tab (#1-#5) #1 The PDF outline/TOC panel is now resizable (drag handle, width persisted to termipod.read.pdfTocW). #2 Search now highlights the exact matched substrings: each match is wrapped in a <mark> inside the transparent text-layer span, so a translucent box paints right over the matched word (was a coarse whole-span tint that rarely showed). #3 Internal PDF links (refs, figures, ToC destinations) are now clickable and jump within the document — annotations with a `dest` (not just a `url`) are captured and routed through goToDest; external links still open the in-app browser tab. Internal links get a distinct green hover. #4 +notes did nothing because clicking the toolbar button collapsed the text selection before onClick read it. Fixed: snapshot the selection on the button's mousedown (+ preventDefault) and use that snapshot. #5 The assistant is now a tab alongside Info/Read/Notes/Cite/Meta in the reader (embedded Inspector only), so it can be used while reading the PDF; its context is the paper + current notes, and Insert appends to notes. Co-Authored-By: Claude Opus 4.8 <[email protected]>
physercoe
pushed a commit
that referenced
this pull request
Jul 13, 2026
…o-style, #3 pt1) The PDF outline panel becomes a tabbed navigator: Outline (the document's bookmarks, as before) and Thumbnails (every page rendered small, lazily via IntersectionObserver; click to jump, active page outlined + scrolled into view). The panel toggle (☰) now opens whenever a PDF is loaded — thumbnails work even when a PDF has no embedded outline — defaulting to Outline when one exists. Annotation tools (highlight/underline/note-comment, then draw/area-select) are the next slice — they need a persistence model, tracked separately. Co-Authored-By: Claude Opus 4.8 <[email protected]>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bumps softprops/action-gh-release from 2 to 3.
Release notes
Sourced from softprops/action-gh-release's releases.
... (truncated)
Changelog
Sourced from softprops/action-gh-release's changelog.
Commits
b430933release: cut v3.0.0 for Node 24 upgrade (#670)c2e35e0chore(deps): bump the npm group across 1 directory with 7 updates (#783)