groove-dev 0.27.94 → 0.27.97

package/README.md

@@ -1,8 +1,8 @@
-# groove
+# Groove OS 🌀
 
-**Orchestrate your AI coding agents. Stop losing context.**
+**Orchestrate your AI coding agents over localhost.**
 
-The open-source orchestration layer for AI coding tools. Spawn teams of agents, coordinate their work, and never lose context — with a full GUI dashboard.
+The open-source desktop OS for AI coding tools. Spawn multi-agent teams, eliminate cold-starts via our "Journalist" state-tracker, and manage complex workflows with a beautiful GUI dashboard. Your code never touches a centralized server.
 
 [![npm](https://img.shields.io/npm/v/groove-dev)](https://www.npmjs.com/package/groove-dev)
 [![License](https://img.shields.io/badge/license-FSL--1.1--Apache--2.0-blue)](LICENSE)
@@ -19,6 +19,8 @@ No terminal needed. Download, install, open a project folder.
 
 Everything is bundled — daemon, GUI, auto-updates. No Node.js required.
 
+**Zero Dependencies:** The Groove Desktop App is completely self-contained. You do not need to install Node.js, Python, or manage terminal dependencies. Just download, open, and spawn a team.
+
 ### Developer Install
 
 ```bash
@@ -91,8 +93,6 @@ groove auto-detects monorepo workspaces (npm, pnpm, lerna) and lets you spawn ea
 
 ## Remote Access
 
-*Included with Groove Pro.*
-
 Run groove on a VPS and manage your agents from anywhere. No ports exposed to the internet. No tokens. No custom auth code. Zero attack surface.
 
 ### How It Works
@@ -154,8 +154,6 @@ This is by design. Direct exposure requires custom auth, rate limiting, TLS mana
 
 ### Federation
 
-*Included with Groove Pro.*
-
 Pair groove daemons across machines with Ed25519 key exchange. The security layer is built — cross-server agent coordination (typed contracts, federated registry) is coming soon.
 
 ```bash
@@ -185,7 +183,7 @@ groove audit           # view recent entries
 groove audit -n 50     # last 50 entries
 ```
 
-```
+```log
 2:14:32 PM  agent.spawn          id=a1 role=backend provider=claude-code
 2:14:35 PM  agent.spawn          id=a2 role=frontend provider=claude-code
 2:33:12 PM  agent.rotate         oldId=a1 newId=a3 role=backend
@@ -217,9 +215,11 @@ Append-only, `0600` permissions, auto-rotates at 5MB. When team auth is added, e
 | Provider | Auth | Models |
 |----------|------|--------|
 | **Claude Code** | Subscription | Opus 4.6, Sonnet 4.6, Haiku 4.5 |
-| **Codex** | API Key | o3, o4-mini, GPT-4.1, GPT-4.1 Mini, GPT-4.1 Nano |
-| **Gemini CLI** | API Key | 3.1 Pro, 3 Flash, 3.1 Flash Lite, 2.5 Pro, 2.5 Flash |
-| **Ollama** | Local | Any |
+| **Codex** | API Key | GPT-5.5, o3, o4-mini |
+| **Gemini CLI** | API Key | 3.1 Pro, 3 Flash, 2.5 Pro, 2.5 Flash |
+| **Grok** | API Key | Grok 4, 4.1 Fast, Code Fast |
+| **Local Models** | Local | Qwen, DeepSeek, Llama, Mistral, Phi, Codestral |
+| **Groove Network** | P2P | Federated inference |
 
 groove is a process manager — it spawns actual AI tool binaries. It never proxies API calls, never touches OAuth tokens, never impersonates any client. Your AI tools talk directly to their servers.
 
@@ -229,46 +229,19 @@ Works in any terminal, any IDE, any OS. Technical and non-technical users alike.
 
 Open the dashboard after starting the daemon (local or remote):
 
-- **Agent Tree** — visual node graph with Bezier spline connections, role badges, live status
-- **File Editor** — CodeMirror 6 with syntax highlighting, file tree, tabs, media viewer, and embedded terminal
-- **Chat** — instruct agents, query without disrupting, continue completed agents, streaming text
-- **Command Center** — gauge charts, live telemetry, token savings, model routing, adaptive thresholds
-- **Quick Launch** — planner recommends team, one-click to spawn all
-- **Skills Store** — app-store marketplace for agent skills with ratings and verification
-- **PM Review Log** — full audit trail of AI Project Manager decisions
-- **Team Management** — save, load, export, import agent configurations
-
-## Plans
-
-groove's core orchestration is free. Unlimited agents, teams, context rotation, Layer 7 memory, model routing, dashboard, all providers, skills marketplace, and scheduling all ship in the free tier. No agent caps.
-
-Pro and Team add the infrastructure for users who outgrow a single machine — remote access, federation, and multi-device sync.
-
-| Feature | Free | Pro | Team |
-|---------|:----:|:---:|:----:|
-| Unlimited agents | ✓ | ✓ | ✓ |
-| Unlimited teams | ✓ | ✓ | ✓ |
-| Context rotation (adaptive) | ✓ | ✓ | ✓ |
-| Layer 7 memory | ✓ | ✓ | ✓ |
-| Journalist (zero cold-start) | ✓ | ✓ | ✓ |
-| AI Project Manager | ✓ | ✓ | ✓ |
-| Adaptive model routing | ✓ | ✓ | ✓ |
-| Dashboard + telemetry | ✓ | ✓ | ✓ |
-| All providers (Claude, Codex, Gemini, Ollama) | ✓ | ✓ | ✓ |
-| Skills marketplace | ✓ | ✓ | ✓ |
-| Quick Launch | ✓ | ✓ | ✓ |
-| Workspaces | ✓ | ✓ | ✓ |
-| Scheduling | ✓ | ✓ | ✓ |
-| Gateways | ✓ | ✓ | ✓ |
-| Remote tunnels (SSH / Tailscale) | — | ✓ | ✓ |
-| Federation (daemon-to-daemon) | — | ✓ | ✓ |
-| Cloud team management | — | — | ✓ |
-| Centralized config | — | — | ✓ |
-| Seat management | — | — | ✓ |
-
-- **Free** — one machine, one person, everything works.
-- **Pro** — multiple machines, one person, everything connects.
-- **Team** — multiple machines, multiple people, everything is managed.
+- **Agent Tree** — React Flow with slim 200px nodes, animated Bezier edges, minimap, welcome onboarding
+- **File Editor** — CodeMirror 6, 7 language modes, file tree, tabs, media viewer, embedded terminal
+- **Chat** — bubble UI with mode pills, streaming text, markdown rendering, continue completed agents
+- **Dashboard** — KPI strip with sparklines, token/cost charts, fleet panel, savings breakdown, per-team burn
+- **Quick Launch** — planner recommends team, one-click spawn all agents
+- **Skills Marketplace** — NFT-style skill cards, category bar, search, ratings, integrations tab
+- **Team Management** — create, rename, delete teams, inherited defaults
+- **Terminal** — xterm.js embedded terminal with resize/fullscreen
+- **Command Palette** — Cmd+K fuzzy search with dynamic agent commands
+
+## Pricing
+
+Everything is free and open source. Unlimited agents, teams, context rotation, Layer 7 memory, model routing, dashboard, all providers, federation, remote access, skills marketplace, and scheduling. No tiers, no upgrades, no gates.
 
 ## Adaptive Model Routing
 
@@ -294,7 +267,7 @@ groove routes tasks to the cheapest model that can handle them. Planners get Opu
     └──────────────────────┬───────────────────────┘
                            │
     ┌──────────────────────▼───────────────────────┐
-    │   Claude Code · Codex · Gemini CLI · Ollama  │
+    │  Claude · Codex · Gemini · Grok · Local · P2P│
     └──────────────────────────────────────────────┘
 ```
 

package/node_modules/@groove-dev/cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@groove-dev/cli",
-  "version": "0.27.94",
+  "version": "0.27.97",
   "description": "GROOVE CLI — manage AI coding agents from your terminal",
   "license": "FSL-1.1-Apache-2.0",
   "type": "module",

package/node_modules/@groove-dev/daemon/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@groove-dev/daemon",
-  "version": "0.27.94",
+  "version": "0.27.97",
   "description": "GROOVE daemon — agent orchestration engine",
   "license": "FSL-1.1-Apache-2.0",
   "type": "module",

package/node_modules/@groove-dev/daemon/src/api.js

@@ -3694,18 +3694,26 @@ Keep responses concise. Help them think, don't lecture them about the system the
       const baseDir = daemon.config?.defaultWorkingDir || daemon.projectDir;
 
       if (mode === 'plan-first') {
-        // Spawn a headless planner to generate per-agent prompts, then auto-launch
+        const rolesList = roles.map(r => r.role || r.name || r).join(', ');
+        const providerNote = teamProvider ? ` (provider: ${teamProvider})` : '';
+        let plannerPrompt;
+        if (task) {
+          plannerPrompt = `The user wants these agents: ${rolesList}${providerNote}. Task: ${task}`;
+        } else {
+          plannerPrompt = '';
+        }
         const plannerConfig = validateAgentConfig({
           role: 'planner',
-          prompt: task || 'Analyze the codebase and create a plan for the team.',
+          prompt: plannerPrompt,
           provider: teamProvider,
           model: teamModel,
           workingDir: baseDir,
         });
         plannerConfig.teamId = defaultTeamId;
+        plannerConfig.teamBuilderRoles = roles.map(r => ({ role: r.role || r, provider: r.provider || null }));
         const planner = await daemon.processes.spawn(plannerConfig);
         daemon.audit.log('team-builder.plan-first', { plannerId: planner.id, roles: roles.length });
-        return res.status(202).json({ mode: 'plan-first', plannerId: planner.id, message: 'Planner spawned — team will launch when plan is ready' });
+        return res.status(202).json({ mode: 'plan-first', plannerId: planner.id, message: 'Planner spawned — waiting for user instructions' });
       }
 
       const spawned = [];
@@ -4157,17 +4165,17 @@ Keep responses concise. Help them think, don't lecture them about the system the
   // --- Federation ---
 
   // Federation status (v1 — includes whitelist, connections, ambassadors)
-  app.get('/api/federation', proOnly, (req, res) => {
+  app.get('/api/federation', (req, res) => {
     res.json(daemon.federation.getStatus());
   });
 
   // List peers
-  app.get('/api/federation/peers', proOnly, (req, res) => {
+  app.get('/api/federation/peers', (req, res) => {
     res.json(daemon.federation.getPeers());
   });
 
   // Unpair a peer
-  app.delete('/api/federation/peers/:id', proOnly, (req, res) => {
+  app.delete('/api/federation/peers/:id', (req, res) => {
     try {
       daemon.federation.unpair(req.params.id);
       res.json({ ok: true });
@@ -4177,7 +4185,7 @@ Keep responses concise. Help them think, don't lecture them about the system the
   });
 
   // Initiate pairing with a remote daemon
-  app.post('/api/federation/initiate', proOnly, async (req, res) => {
+  app.post('/api/federation/initiate', async (req, res) => {
     try {
       const { remoteUrl } = req.body;
       if (!remoteUrl || typeof remoteUrl !== 'string') {
@@ -4192,11 +4200,11 @@ Keep responses concise. Help them think, don't lecture them about the system the
 
   // --- Federation v1: Whitelist ---
 
-  app.get('/api/federation/whitelist', proOnly, (req, res) => {
+  app.get('/api/federation/whitelist', (req, res) => {
     res.json(daemon.federation.whitelist?.list() || []);
   });
 
-  app.post('/api/federation/whitelist', proOnly, (req, res) => {
+  app.post('/api/federation/whitelist', (req, res) => {
     try {
       const { ip, port, name } = req.body;
       if (!ip || typeof ip !== 'string') {
@@ -4210,7 +4218,7 @@ Keep responses concise. Help them think, don't lecture them about the system the
     }
   });
 
-  app.delete('/api/federation/whitelist/:ip', proOnly, (req, res) => {
+  app.delete('/api/federation/whitelist/:ip', (req, res) => {
     try {
       daemon.federation.whitelist.remove(req.params.ip);
       daemon.broadcast({ type: 'federation:whitelist', data: daemon.federation.whitelist.list() });
@@ -4248,7 +4256,7 @@ Keep responses concise. Help them think, don't lecture them about the system the
 
   // --- Federation v1: Connections ---
 
-  app.get('/api/federation/connections', proOnly, (req, res) => {
+  app.get('/api/federation/connections', (req, res) => {
     res.json(daemon.federation.connections?.getStatus() || []);
   });
 
@@ -4271,13 +4279,13 @@ Keep responses concise. Help them think, don't lecture them about the system the
     }
   });
 
-  app.get('/api/federation/pouch/log', proOnly, (req, res) => {
+  app.get('/api/federation/pouch/log', (req, res) => {
     const limit = Math.min(parseInt(req.query.limit, 10) || 50, 200);
     res.json(daemon.federation.ambassadors?.getPouchLog(limit) || []);
   });
 
   // Send a pouch message to a peer (local agents/GUI call this)
-  app.post('/api/federation/pouch/send', proOnly, async (req, res) => {
+  app.post('/api/federation/pouch/send', async (req, res) => {
     try {
       const { peerId, contract } = req.body;
       if (!peerId || !contract) {
@@ -4323,7 +4331,7 @@ Keep responses concise. Help them think, don't lecture them about the system the
     }
   });
 
-  app.post('/api/federation/contract/send', proOnly, async (req, res) => {
+  app.post('/api/federation/contract/send', async (req, res) => {
     try {
       const { peerId, contract } = req.body;
       if (!peerId || !contract) {

package/node_modules/@groove-dev/daemon/src/introducer.js

@@ -48,6 +48,17 @@ export class Introducer {
       }
     }
 
+    if (newAgent.teamBuilderRoles && newAgent.teamBuilderRoles.length > 0) {
+      lines.push('');
+      lines.push('## Team Builder Pre-Selection');
+      lines.push('');
+      const roleDescs = newAgent.teamBuilderRoles.map(r => {
+        return r.provider ? `${r.role} (provider: ${r.provider})` : r.role;
+      });
+      lines.push(`The user selected these roles in the Team Builder UI: ${roleDescs.join(', ')}.`);
+      lines.push('When the user gives you a task, create a plan using EXACTLY these roles. Do not redesign the team composition.');
+    }
+
     lines.push('');
 
     if (others.length === 0) {
@@ -199,12 +210,16 @@ export class Introducer {
     // CLAUDE.md parity — non-Claude providers don't read CLAUDE.md natively,
     // so inject its project content (minus the GROOVE section) into introContext
     if (newAgent.provider && newAgent.provider !== 'claude-code') {
-      const claudeMdContent = this._loadClaudeMd(newAgent.workingDir);
-      if (claudeMdContent) {
-        lines.push('');
-        lines.push('## Project Context (from CLAUDE.md)');
-        lines.push('');
-        lines.push(claudeMdContent);
+      if (newAgent.role === 'planner') {
+        // Planners don't need full project context — codebase structure is injected separately
+      } else {
+        const claudeMdContent = this._loadClaudeMd(newAgent.workingDir);
+        if (claudeMdContent) {
+          lines.push('');
+          lines.push('## Project Context (from CLAUDE.md)');
+          lines.push('');
+          lines.push(claudeMdContent);
+        }
       }
     }
 

package/node_modules/@groove-dev/daemon/src/process.js

@@ -315,7 +315,7 @@ function sanitizeFilename(name) {
 
 export function wrapWithRoleReminder(role, message) {
   if (role === 'planner' && !message.startsWith('ROLE REMINDER:')) {
-    return 'ROLE REMINDER: You are a PLANNING ONLY agent. Do NOT write code, edit files, or use Edit/Write/Bash tools. Route this task to your team by writing .groove/recommended-team.json.\n\nUser message: ' + message;
+    return 'ROLE REMINDER: You are a PLANNING ONLY agent. Do NOT write code, edit source files, or run build commands. Your ONLY file write should be .groove/recommended-team.json.\n\nUser message: ' + message;
   }
   return message;
 }
@@ -1336,7 +1336,23 @@ For normal file edits within your scope, proceed without review.
       if (existsSync(targetPath)) return;
 
       const log = readFileSync(logPath, 'utf8');
-      const match = log.match(/\{[\s\S]*?"agents"\s*:\s*\[[\s\S]*?\]\s*\}/);
+
+      // Extract text from JSON log events (Codex agent_message, Claude content blocks)
+      const textParts = [];
+      for (const line of log.split('\n')) {
+        try {
+          const evt = JSON.parse(line.trim());
+          if (evt.item?.type === 'agent_message' && evt.item?.text) textParts.push(evt.item.text);
+          if (evt.type === 'assistant' && evt.message?.content) {
+            for (const block of evt.message.content) {
+              if (block.type === 'text') textParts.push(block.text);
+            }
+          }
+        } catch { textParts.push(line); }
+      }
+      const fullText = textParts.join('\n');
+
+      const match = fullText.match(/\{[\s\S]*?"agents"\s*:\s*\[[\s\S]*?\]\s*\}/);
       if (!match) return;
 
       let parsed;

package/node_modules/@groove-dev/daemon/src/providers/gemini.js

@@ -147,11 +147,8 @@ export class GeminiProvider extends Provider {
     }
 
     switch (event.type) {
-      case 'agent_start':
-        return { type: 'activity', subtype: 'assistant', sessionId: event.streamId, data: [{ type: 'text', text: '' }] };
-
-      case 'session_update':
-        return null;
+      case 'init':
+        return { type: 'activity', subtype: 'assistant', sessionId: event.session_id, data: [{ type: 'text', text: '' }] };
 
       case 'message': {
         if (event.role === 'user') return null;
@@ -165,38 +162,39 @@ export class GeminiProvider extends Provider {
         return { type: 'activity', subtype: 'assistant', data: blocks };
       }
 
-      case 'tool_request': {
-        const toolName = (event.name || '').includes('shell') || (event.name || '').includes('exec')
-          ? 'Bash' : event.name || 'Tool';
-        const input = event.name === 'Bash' || (event.name || '').includes('shell')
-          ? { command: typeof event.args === 'string' ? event.args : JSON.stringify(event.args || {}) }
-          : (event.args || {});
+      case 'tool_use': {
+        const name = event.tool_name || '';
+        const toolName = name.includes('shell') || name.includes('exec')
+          ? 'Bash' : name || 'Tool';
+        const input = toolName === 'Bash'
+          ? { command: event.parameters?.command || (typeof event.parameters === 'string' ? event.parameters : JSON.stringify(event.parameters || {})) }
+          : (event.parameters || {});
         return {
           type: 'activity', subtype: 'assistant',
-          data: [{ type: 'tool_use', id: event.requestId || 'tool', name: toolName, input }],
+          data: [{ type: 'tool_use', id: event.tool_id || 'tool', name: toolName, input }],
         };
       }
 
-      case 'tool_response': {
-        const rawContent = event.content;
-        const contentParts = Array.isArray(rawContent) ? rawContent : (typeof rawContent === 'string' ? [{ text: rawContent }] : rawContent ? [rawContent] : []);
-        const content = contentParts.map((p) => p.text || '').join('').slice(0, 2000);
-        const toolName = (event.name || '').includes('shell') || (event.name || '').includes('exec')
-          ? 'Bash' : event.name || 'Tool';
+      case 'tool_result': {
+        const content = typeof event.output === 'string' ? event.output.slice(0, 2000) : '';
+        const toolId = event.tool_id || '';
+        const isShell = toolId.includes('shell') || toolId.includes('exec');
+        const toolName = isShell ? 'Bash' : 'Tool';
         return {
           type: 'activity', subtype: 'assistant',
           data: [
-            { type: 'tool_use', id: event.requestId || 'tool', name: toolName, input: {} },
+            { type: 'tool_use', id: toolId || 'tool', name: toolName, input: {} },
             ...(content ? [{ type: 'text', text: content }] : []),
           ],
         };
       }
 
-      case 'usage': {
-        const inputTokens = event.inputTokens || 0;
-        const outputTokens = event.outputTokens || 0;
-        const cachedTokens = event.cachedTokens || 0;
-        const totalTokens = inputTokens + outputTokens;
+      case 'result': {
+        const stats = event.stats || {};
+        const inputTokens = stats.input_tokens || 0;
+        const outputTokens = stats.output_tokens || 0;
+        const cachedTokens = stats.cached || 0;
+        const totalTokens = stats.total_tokens || (inputTokens + outputTokens);
 
         const model = GeminiProvider.models.find((m) => m.id === this._currentModel);
         const pricing = model?.pricing;
@@ -211,7 +209,8 @@ export class GeminiProvider extends Provider {
         }
 
         return {
-          type: 'activity', subtype: 'assistant',
+          type: 'result',
+          subtype: 'assistant',
           data: [{ type: 'text', text: '' }],
           tokensUsed: totalTokens,
           inputTokens,
@@ -220,12 +219,12 @@ export class GeminiProvider extends Provider {
           contextUsage: inputTokens / maxContext,
           estimatedCostUsd,
           costSource: pricing ? 'calculated' : 'estimated',
+          cost: estimatedCostUsd,
+          duration: stats.duration_ms,
+          turns: stats.tool_calls || 0,
         };
       }
 
-      case 'agent_end':
-        return { type: 'activity', subtype: 'assistant', data: [{ type: 'text', text: '' }] };
-
       case 'error':
         return { type: 'activity', subtype: 'assistant', data: [{ type: 'text', text: `Error: ${event.message || 'unknown'}` }] };