agentgui 1.0.987 → 1.0.988

package/AGENTS.md

@@ -1,5 +1,9 @@
 # AgentGUI — Agent Notes
 
+## CRITICAL — ACP is managed ONLY by `lib/acp-sdk-manager.js` (2026-06-20) — twentieth 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 quality sweep (2026-06-19) — nineteenth run
 
 132-agent workflow wf_8183560b-e3d (12 lenses, 93 confirmed). Kit: thinking settled state, retry on non-last message, AT aria fixes (followup chips live guard, cwd focus restore, skip-link target, sessions toggle chevron, shell print styles, a11y-01 index.html), composer disabled visual state, files-modals focus trap re-query + stable aria-labelledby, sessions.js listbox+aria-selected. App: resume-transcript-load (loadResumeTranscript historical messages + spinner), ACP force-restart (unhealthy agent restart btn + WS handler), history tool_use rail=purple, shortcuts overlay role=dialog, streaming-active badge on chat rail tab, sortedAgents memoized, files filter 150ms debounce, _seen Set capped 5000, humanizeMs->fmtDuration, dead historySide() removed, pathBasename util, ARM_RESET_MS constant, refreshHistory guard, perf-002/006/007/008 memoization+cleanup, live-tokens accumulated, backend-change-mid-chat guard, transcript loading state, session-expiry onSessionExpired hook, server-500 stream error sanitized. Server: IPv4-mapped IPv6 normalization in rate-limit, image route streaming (no sync read), isWindows module-level, getAvailableAgents export removed, files-plugin+workflow-plugin confinement, acp-plugin shell injection fix, plugin-routes CSRF fix, asset-server JS injection fix. Full detail in rs-learn (recall "agentgui 19th run").
@@ -50,28 +54,11 @@ Lands the 10th run's deferred batch + the janks a live witness surfaced on this
 
 ## GUI screen-real-estate layer (2026-06-12) — fluid columns + drag-resize handles
 
-An additive viewport-density layer on top of the concurrent 11th/12th-run work (the user verdict: maximize screen real-estate, the SDK must read as a desktop application). New workflow `.claude/workflows/gui-screen-realestate.js` (6 density lenses: space-efficiency, viewport-utilization, information-density, adaptive-layout, application-chrome, finish-pending). 50 agents -> 17 confirmed gaps (`PUNCHLIST-DENSITY.md`). Kit pushed `anentrypoint-design` 97dc646.
-
-**The shell reclaims the viewport (genuinely-new, complementary to the 12th run's breakpoint column-yield).** Fixed `--ws-rail-w 232 / --ws-sessions-w 296 / --ws-pane-w 320` (≈848px chrome) became fluid clamps (`clamp(200px,16vw,260px)` etc) so chrome scales with the viewport. Kit `shell.js` renders keyboard+pointer `.ws-resizer` separators (`role=separator`, ArrowLeft/Right 16px, pointer-drag) between each track that write a clamped inline `--ws-<col>-w` (inline overrides the clamp base, pinning the width) persisted to `localStorage ds.ws.w.<col>` and seeded on mount via a `ref` (`seedWsWidths`, no app wiring). Sessions gained a desktop collapse toggle (`.ws-sessions-toggle` in the crumb; `.ws-desktop-toggle` CSS = inverse of the mobile drawer-toggle; `toggleWs` extended with `'sessions' -> ws-sessions-collapsed`). `.btn*` moved off `--r-pill` to `--r-1` (8/14 pad, 32 min-h, 44 coarse floor); `.app-status` is a thin 26px ellipsized strip (`--app-status-h:26px`); `.chat-msg-flat` centers at `--measure` (`max-width:var(--measure);margin-inline:auto`) + density-scaled spacing via the `--density` token + 84ch ultrawide tier; `.ds-dash-grid` densified (minmax 240, 220 wide tier).
-
-**Concurrent-writer reconciliation (load-bearing — both repos have active concurrent writers).** While this run was in flight, concurrent writers shipped the ENTIRE Files multi-select + density-thumbnail + move feature on BOTH `anentrypoint-design/main` (FileGrid `selected/onToggleSelect/onSelectAll/onClearSelection/density/onDensity/thumbUrl` + `BulkBar` export) AND `agentgui/main` (the 11th/12th runs: `onMark`/`density` wiring + `B.moveEntry` -> confined `POST /api/move` + bulk move). So this run's parallel multi-select/move/F3 work was DROPPED and theirs adopted; the only thing landed on agentgui is the re-vendored kit dist (the screen-real-estate layer above) + this workflow + docs. **Rule: when origin advances mid-run with the same feature, drop yours and adopt theirs, then re-apply only your genuinely-distinct work onto their current files — never force a parallel implementation; the agentgui `/api/move-confined` endpoint built this run was redundant with their `/api/move` and discarded.** The kit ships 4 build lints (`lint-tokens/glyphs/null-children/classes`) - all must pass. Witness localhost:3009/gm/ (3000 owned by another app): the vendored bundle is served fresh only after a SERVER restart (it pre-warms statics in memory) AND a fresh browser context (the persistent witness context caches the ES module - clear it or new-session), else you read a stale bundle. Witnessed 0 console errors chat/files, btn 10px, status 26px, 3 resizers, sessions toggle, fluid rail 256@1600, no h-scroll.
+Fluid-column + drag-resize density layer (workflow `gui-screen-realestate.js`, 50 agents -> 17 gaps). Fixed `--ws-*-w` chrome became fluid `clamp()`; kit `shell.js` `.ws-resizer` separators write persisted inline `--ws-<col>-w`; `.ws-sessions-toggle` desktop collapse; `.btn*` -> `--r-1`; `.app-status` 26px. **Concurrent-writer rule (load-bearing): when origin advances mid-run with the same feature, drop yours and adopt theirs, then re-apply only genuinely-distinct work — never a parallel impl.** Full detail in rs-learn (recall "agentgui screen-real-estate layer").
 
 ## GUI Claude-Code-web parity sweep (2026-06-11) — tenth maximum-effort run
 
-Hunts the VISUAL + LAYOUT + MOTION design-language parity gap vs claude.ai/code (the user verdict: too primitive, layout jank). New workflow `.claude/workflows/gui-claude-code-parity.js` (6 design-language lenses: visual-language, layout-composition, chat-thread-craft, file-browser-fidelity, live-session-command, motion-microinteraction; hunt -> adversarial verify (kept-typography guard) -> plan). 66 agents -> 37 confirmed gaps (`PUNCHLIST-PARITY.md`). Pushed: kit `anentrypoint-design` 28387c3, agentgui 0a93fc6.
-
-**Chat reads like Claude Code, not a messenger.** AgentChat passes `flat:true` + `aicat:false` to `ChatMessage`; flat turns are FULL-WIDTH, avatar-less, capped at `--measure`, led by a `.chat-role` label (You / agent name), the assistant turn carrying a faint `color-mix(--fg 3%)` background (`.chat-msg-flat` in chat.css). The bubble/avatar layout is kept ONLY for the messenger demo. Composer is ONE bordered rounded-rect shell (`.chat-composer` border `var(--rule)` + `--r-2`, `:focus-within` ring) with a borderless textarea and an inline `--r-1` send (was a `--r-pill` stadium + 50% round button); `flex-wrap` + `.chat-composer-hint{order:5}` lay context/input/hint as rows. Tool calls render as a bordered status-toned card (`.chat-tool` de-frames the inherited `.chat-bubble`, head=icon·name·label·status-pill, running->accent / error->flame); the bundle shipped ZERO `.chat-tool` CSS before. `injectCodeCopy` reads the `<code>` `language-*` class -> a `.chat-code-lang` header tab on every rendered block. `::selection` is accent-tint, not mascot pink.
-
-**Layout has a real content gutter.** Kit `WorkspaceShell` gains `mainFlush`; `.ws-main` now has `padding: var(--space-4) var(--space-5)` (files/live/history/settings) and the chat tab opts out via `.ws-main--flush` (app sets `mainFlush: state.tab==='chat'` since the thread self-gutters at `--measure`). `.ws-crumb` gets `min-height:48px` + a matching left gutter so the top edge aligns with the rail head.
-
-**Live dashboard is a command center.** `SessionDashboard`: status-bucketed grid (Errored/Running/Idle/External `role=group` sections when `sort==='status'`, flat otherwise); header status breakdown `N running · M idle · K errors` (toned `.ds-dash-breakdown .seg.is-*`); a `.ds-dash-stream-disc` heartbeat (pulsing `status-dot-live` when connected); tri-state `role=checkbox` select-all + clear (`onSelectAll`/`onClearSelection`, app toggles `state.live.selected`); `is-armed` one-shot pulse on the armed stop button. `SessionCard`: cost/token stat (`session.cost`/`tokens`, app feeds the in-page chat's `totalCost` onto its own card), `is-new` arrival cue (`session.isNew`, app sets it for <3s-old sids), icon-led actions (open=primary), `is-stale` amber inset bar (was opacity .92). Files: `TYPE_ICON.dir='folder'` + new `file-image`/`link`/`folder-open` icons (shell.js); ConversationList loading skeleton rows. File preview: Prism highlight hook (`highlightAllUnder` ref) + `.ds-preview-gutter` line numbers + image fit/actual toggle + checkerboard.
-
-**Motion (all `prefers-reduced-motion: no-preference` guarded).** `.ws-shell` eases `grid-template-columns` (rail/pane collapse no longer snaps; track count stays stable); `.agentchat-thread` `scroll-behavior:smooth`; `.ds-session-row` hover transition; `.ws-scrim` opacity fade (was display toggle); `.ds-alert` enter; copy->copied color flip; `.ds-dash-card.is-new` fade-in.
-
-**Load-bearing webjsx fix (caught by the live witness).** SessionCard passed raw `null`s positionally in its head/meta/actions/top-level children arrays; webjsx `applyDiff` crashes (`reading 'key'`) on ANY null among VElement siblings (not just keyed ones). This was a PRE-EXISTING latent bug (the unmodified kit crashed identically on this env's external sessions) that this env's data now triggered. Fix: `.filter(Boolean)` EVERY children array in SessionCard + SessionDashboard header/body, and `Btn` now spreads array `children`. The kit-wide rule: never pass a conditional `x ? h() : null` positionally — build the array and `.filter(Boolean)` it (the same pattern ConversationList already uses). Witness localhost:3009/gm/ (port 3000 was owned by another app): 0 console/page errors across chat/files/live; composer 14px radius; ws-main flush=0 on chat, 32px on files; folder icon distinct; no h-scroll. First `/gm/` request triggers the 30-90s ccsniff walk -> poll until 200 before witnessing.
-
-**Deferred (documented in `PUNCHLIST-PARITY.md`, rows pending):** Files multi-select + BulkBar + density-thumbnail grid (kit props + heavy app/server wiring), the `/api/move-confined` drag-to-move endpoint, a toolbar up-button + file-grid keys in the SHORTCUTS array, and F3 panel/card elevation unification. These are lower-leverage than the shipped visual/layout/motion/command-center batch.
-
+Visual/layout/motion parity vs claude.ai/code (workflow `gui-claude-code-parity.js`, 66 agents -> 37 gaps). Flat full-width chat turns (`.chat-msg-flat` at `--measure`), single bordered composer rounded-rect, `.chat-tool` status-toned cards, `WorkspaceShell mainFlush`, command-center `SessionDashboard`, reduced-motion-guarded transitions. **Load-bearing webjsx rule: never pass a conditional `x?h():null` positionally — build the array and `.filter(Boolean)` it; ANY null among VElement siblings crashes `applyDiff` (reading 'key'); `Btn` spreads array children.** Full detail in rs-learn (recall "agentgui 10th run parity").
 ## GUI logic+predictability sweep (2026-06-10) — ninth maximum-effort run
 
 Audits whether everything OPERATES logically and predictably (not coverage, not cohesion). New workflow `.claude/workflows/gui-logic-predictability.js` (6 lenses: interaction-lifecycle, cross-surface-consistency, information-architecture, realtime-truth, input-ergonomics, scale-robustness; hunt -> adversarial verify (kept-typography guard) -> plan). 78 agents -> 67 confirmed findings (`PUNCHLIST-LOGIC.md`). ALL implemented across kit/app/server.

package/lib/acp-sdk-manager.js

@@ -68,6 +68,22 @@ function startProcess(tool) {
 
   log(tool.id + ' started port ' + tool.port + ' pid ' + proc.pid);
 
+  // A missing/unspawnable binary surfaces as an async 'error' event, NOT the
+  // synchronous throw the try/catch above guards (witnessed under Bun: an
+  // uninstalled ACP CLI emits ENOENT asynchronously and, without this listener,
+  // escapes as an uncaught exception that kills the whole server). Funnel it
+  // into the same CRASHED + backoff path as a normal exit so an uninstalled
+  // on-demand agent degrades to "not healthy", never a process crash.
+  proc.on('error', (err) => {
+    processes.delete(tool.id);
+    if (shuttingDown) return;
+    log(tool.id + ' spawn error: ' + err.message);
+    acpMachine.send(tool.id, { type: 'CRASHED' });
+    const snap = acpMachine.snapshot(tool.id);
+    if (snap?.value === 'stopped') { log(tool.id + ' max restarts reached'); return; }
+    scheduleRestart(tool);
+  });
+
   proc.on('close', (code) => {
     processes.delete(tool.id);
     if (shuttingDown) return;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentgui",
-  "version": "1.0.987",
+  "version": "1.0.988",
   "description": "Multi-agent ACP client with real-time communication",
   "type": "module",
   "main": "electron/main.js",

package/lib/plugins/acp-plugin.js

@@ -1,110 +0,0 @@
-// ACP plugin - OpenCode, Gemini, Kilo, Codex startup and health checks
-
-import { spawn, execFileSync } from 'child_process';
-import path from 'path';
-import fs from 'fs';
-
-// Resolve a binary name to an absolute path (mirrors lib/claude-runner-direct.js pattern).
-// Returns the resolved path, or null if not found.
-function resolveBinaryPath(name) {
-  try {
-    const which = process.platform === 'win32' ? 'where' : 'which';
-    const result = execFileSync(which, [name], { encoding: 'utf8', stdio: ['ignore', 'pipe', 'ignore'] });
-    const first = result.trim().split(/\r?\n/)[0];
-    return first || null;
-  } catch {
-    return null;
-  }
-}
-
-export default {
-  name: 'acp',
-  version: '1.0.0',
-  dependencies: [],
-
-  async init(config, plugins) {
-    const toolProcesses = new Map();
-    const healthCheckIntervals = new Map();
-    const restartCounts = new Map();
-    const acpPorts = new Map();
-
-    // Each spec uses argv (binary + args) rather than a shell command string so
-    // spawn runs with shell:false — no shell injection risk even if port/name
-    // values are later sourced from config. Mirrors lib/claude-runner-direct.js.
-    const toolSpecs = [
-      { name: 'opencode', port: 18100, argv: ['opencode', 'acp', '--port', '18100'] },
-      { name: 'gemini',   port: 18101, argv: ['gemini',   'acp', '--port', '18101'] },
-      { name: 'kilo',     port: 18102, argv: ['kilo',     'acp', '--port', '18102'] },
-      { name: 'codex',    port: 18103, argv: ['codex',    'acp', '--port', '18103'] },
-    ];
-
-    const startTool = async (spec) => {
-      try {
-        const [bin, ...args] = spec.argv;
-        // Resolve to an absolute path so spawn never relies on PATH expansion
-        // inside a shell; shell:false is the default for spawn() when called
-        // this way, matching the project-wide rule in AGENTS.md.
-        const resolvedBin = resolveBinaryPath(bin) || bin;
-        const proc = spawn(resolvedBin, args, { shell: false });
-        toolProcesses.set(spec.name, proc);
-        acpPorts.set(spec.name, spec.port);
-        restartCounts.set(spec.name, 0);
-
-        // Health check every 30s
-        const interval = setInterval(() => {
-          if (proc.killed) {
-            clearInterval(interval);
-            healthCheckIntervals.delete(spec.name);
-          }
-        }, 30000);
-        healthCheckIntervals.set(spec.name, interval);
-      } catch (e) {
-        console.error(`[ACP] Failed to start ${spec.name}:`, e.message);
-      }
-    };
-
-    // Start all ACP tools
-    for (const spec of toolSpecs) {
-      await startTool(spec);
-    }
-
-    return {
-      routes: [
-        {
-          method: 'GET',
-          path: '/api/acp/status',
-          handler: (req, res) => {
-            const status = {};
-            for (const [name, proc] of toolProcesses) {
-              status[name] = {
-                running: !proc.killed,
-                port: acpPorts.get(name),
-                pid: proc.pid,
-                restarts: restartCounts.get(name) || 0,
-              };
-            }
-            res.json({ tools: status });
-          },
-        },
-      ],
-      wsHandlers: {},
-      api: {
-        getStatus: () => Object.fromEntries(acpPorts),
-      },
-      stop: async () => {
-        for (const [name, interval] of healthCheckIntervals) {
-          clearInterval(interval);
-        }
-        for (const [name, proc] of toolProcesses) {
-          if (proc && !proc.killed) proc.kill();
-        }
-      },
-    };
-  },
-
-  async reload(state) {
-    return state;
-  },
-
-  async stop() {},
-};