agentgui 1.0.1001 → 1.0.1002

package/AGENTS.md

@@ -1,8 +1,18 @@
 # AgentGUI — Agent Notes
 
-## CRITICAL — ACP is managed ONLY by `lib/acp-sdk-manager.js` (2026-06-20) — twentieth run
+## GUI cohesion sweep (2026-07-02) — twenty-first run
 
-A destructive quality run found a second, redundant ACP path: `lib/plugins/acp-plugin.js` eager-spawned all four ACP tools at boot on **conflicting/mismatched ports** (it mapped gemini=18101 / kilo=18102 / codex=18103, but the canonical `acp-sdk-manager.js` uses opencode=18100 / kilo=18101 / codex=18102), implemented no restart/health logic despite a `/api/acp/status` route that promised `restarts`, and **crashed server boot** with `[FATAL] Uncaught exception: Executable not found in $PATH: "codex"` whenever an ACP binary was missing. Its route was consumed by nobody (the client `acpStatusFor` reads `state.health.acp` from `getACPStatus()` → `acp-sdk-manager.js`); no plugin declared an `acp` dependency. **Deleted the file.** Rule: ACP lifecycle (on-demand start, health via `/provider`, restart-backoff, status) lives in `lib/acp-sdk-manager.js` alone — never add a second eager-spawn ACP manager. Also: `spawn()` for a missing binary surfaces ENOENT as an **async `'error'` event** under Bun (it escapes a synchronous try/catch), so every `spawn()` callsite needs a `proc.on('error', …)` handler. `acp-sdk-manager.js startProcess` was the only remaining callsite missing it (every other — terminal.js, claude-runner-direct.js, gm-agent-configs.js, claude-runner-acp.js — already had one); added `proc.on('error')` routing into the CRASHED + `scheduleRestart` backoff path. Witnessed: clean boot (0 FATAL), `[PLUGINS] Loaded extensions` no longer lists `acp`, `/api/acp/status` → 404, tests 46/46, browser localhost:3009/gm/ 0 console errors. Pushed this run.
+`gui-cohesion` workflow wf_6b9f1890-320 (25 agents, 12 confirmed). Kit (`../design`): `WorkspaceRail` items gain an optional `rail` attention-dot tone (a background surface, e.g. an errored Live session, can flag its nav item even while the user is elsewhere) — new `.ws-rail-item-flag` CSS. `WorkspaceShell` extends the pane-track `stableFrame` treatment to the sessions column (`keepSessionsTrack`, mirrors `keepPaneTrack`) so navigating chat/history/files/live/settings no longer reflows the 4-column grid to 3 and back. Mobile `ws-sessions`/`ws-pane` drawers now trap Tab/Shift+Tab focus (exported `trapTab` from `overlay-primitives.js`, reused in `shell.js`'s `toggleWsDrawer`) — previously Tab could walk focus out into the scrim. `ConversationList` rows accept a richer optional `status` ('error'|'stale'|'running'|'stopping') mirroring `SessionCard`'s `STATUS_DISC`, so a session pinned to the rail's "Running" group shows stuck-vs-busy. `ContextPane` gains a `recentFiles`/`onOpenFile` "recent files" panel. Also picked up inherited uncommitted 19th/20th-run work sitting dirty in the kit tree (agent-chat/chat/files-modals/files a11y+dropzone fixes) — commit and push before starting new work if `git status --porcelain` is non-empty at PLAN entry.
+
+App (`site/app/js/app.js`): new `computeLiveSessions()` is the single source of truth for the live session set (owned `chat.active` rows + external SSE-observed sessions) — shared by the `WorkspaceRail` Live badge count/error-flame tone AND `liveMain()`'s dashboard, closing a prior gap where the nav badge undercounted external sessions the dashboard showed. Sessions rail now renders on the Live tab too (previously only chat/history), so the "jump between conversations" rail and "all sessions at once" oversight view coexist. Follow-up chip "Open X in files" relabeled "Explain X" — it only ever called `sendChat`, never navigated, so the label lied about its behavior. `recentSessionFiles()` scans Edit/Write/MultiEdit/Read/NotebookEdit tool-call args for `file_path`/`path`/`notebook_path`, deduped, most-recent-first, feeding `ContextPane`. "disconnected" -> "offline" in two remaining spots (live-stream reconnect banner, ACP agent status) to match the connected/connecting/offline vocabulary used everywhere else. `refreshActive()` now skips its network poll while `document.hidden` (resumes immediately on `visibilitychange`) instead of polling every ~3.3s from backgrounded tabs.
+
+Server (`lib/http-handler.js`): `/health` now reports `corsWildcardOpen` (true when `CORS_ORIGIN=*` and no `PASSWORD`); the settings panel renders a persistent error `Alert` from it instead of relying on the console-only boot warning an operator not tailing logs would never see.
+
+Witnessed: kit build (`node scripts/build.mjs`) — lint-tokens/lint-glyphs/lint-null-children/lint-classes all OK; `test.js` 28/28; browser localhost:3010/gm/ chat/live/settings tabs, 0 console errors, sessions column confirmed present on Live tab post-fix. Kit pushed `anentrypoint-design` `6cfd115` (rebased onto concurrent-writer `fe50db5`, dist regenerated rather than hand-merging the conflicting minified bundle). agentgui pushed `3410178`. Both repos' CI (Test/Publish/Deploy GH Pages) green.
+
+## CRITICAL — ACP is managed ONLY by `lib/acp-sdk-manager.js`
+
+Twentieth run deleted a redundant eager-spawn `lib/plugins/acp-plugin.js` (conflicting ports, no restart logic, crashed boot on a missing binary). Rule: ACP lifecycle lives in `lib/acp-sdk-manager.js` alone; every `spawn()` callsite needs a `proc.on('error', …)` handler (ENOENT surfaces async under Bun). Full detail in rs-learn (recall "agentgui 20th run ACP redundant manager").
 
 ## GUI quality sweep (2026-06-19) — nineteenth run
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.1001",
+  "version": "1.0.1002",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "electron/main.js",