openvoiceui 1.0.0

package/setup-config.js

@@ -0,0 +1,262 @@
+#!/usr/bin/env node
+// setup-config.js — Generates all config files for OpenVoiceUI Pinokio install.
+// Reads API keys from PINOKIO_* environment variables (set by install.js env: block).
+// This avoids the broken json.set / pinokio-input.json flow entirely.
+
+const fs = require("fs");
+const path = require("path");
+const crypto = require("crypto");
+
+// ---------------------------------------------------------------------------
+// Helpers
+// ---------------------------------------------------------------------------
+
+function getKey(envName) {
+  // install.js passes keys as PINOKIO_<KEY_NAME> via env
+  const v = process.env["PINOKIO_" + envName] || "";
+  if (!v || v === "undefined" || v === "null" || v.startsWith("{{")) return "";
+  return v.trim();
+}
+
+const PORT = getKey("PORT") || "5001";
+const token = crypto.randomBytes(24).toString("hex");
+const secret = crypto.randomBytes(32).toString("hex");
+
+// ---------------------------------------------------------------------------
+// 1. Write openclaw.json (nested gateway/agents format for v2026.3.2+)
+// ---------------------------------------------------------------------------
+
+// Pick default model based on which keys the user provided
+let defaultModel = "groq/llama-3.3-70b-versatile"; // fallback — Groq is required
+if (getKey("ZAI_API_KEY"))       defaultModel = "z-ai/glm-4.7";
+if (getKey("ANTHROPIC_API_KEY")) defaultModel = "anthropic/claude-sonnet-4-5";
+if (getKey("OPENAI_API_KEY"))    defaultModel = "openai/gpt-4o";
+// Z.AI wins if multiple are set (best value)
+if (getKey("ZAI_API_KEY"))       defaultModel = "z-ai/glm-4.7";
+
+const openclawConfig = {
+  gateway: {
+    mode: "local",
+    port: 18791,
+    bind: "lan",
+    auth: { mode: "token", token: token },
+    trustedProxies: ["127.0.0.1", "172.16.0.0/12", "10.0.0.0/8"],
+    controlUi: {
+      allowInsecureAuth: true,
+      dangerouslyDisableDeviceAuth: true,
+    },
+  },
+  agents: {
+    defaults: {
+      model: defaultModel,
+      thinkingDefault: "off",
+      blockStreamingDefault: "on",
+      blockStreamingBreak: "text_end",
+      timeoutSeconds: 300,
+    },
+    list: [{ id: "main", default: true, workspace: "/root/.openclaw/workspace" }],
+  },
+};
+
+fs.mkdirSync("openclaw-data/workspace", { recursive: true });
+fs.writeFileSync(
+  "openclaw-data/openclaw.json",
+  JSON.stringify(openclawConfig, null, 2) + "\n"
+);
+console.log("  Wrote openclaw-data/openclaw.json");
+
+// ---------------------------------------------------------------------------
+// 2. Write .env
+// ---------------------------------------------------------------------------
+
+const envLines = [
+  "# OpenVoiceUI — generated by Pinokio installer",
+  `PORT=${PORT}`,
+  "DOMAIN=localhost",
+  `SECRET_KEY=${secret}`,
+  "",
+  "# OpenClaw Gateway",
+  "CLAWDBOT_GATEWAY_URL=ws://127.0.0.1:18791",
+  `CLAWDBOT_AUTH_TOKEN=${token}`,
+  "GATEWAY_SESSION_KEY=voice-main-1",
+  "",
+  "# AI Provider Keys",
+];
+
+const envKeyList = [
+  "ANTHROPIC_API_KEY", "OPENAI_API_KEY", "GEMINI_API_KEY",
+  "OPENROUTER_API_KEY", "MISTRAL_API_KEY", "XAI_API_KEY", "ZAI_API_KEY",
+  "CEREBRAS_API_KEY", "TOGETHER_API_KEY", "HF_TOKEN",
+  "MOONSHOT_API_KEY", "KIMI_API_KEY", "MINIMAX_API_KEY", "QIANFAN_API_KEY",
+  "MODELSTUDIO_API_KEY", "XIAOMI_API_KEY", "VOLCANO_ENGINE_API_KEY",
+  "BYTEPLUS_API_KEY", "SYNTHETIC_API_KEY", "VENICE_API_KEY",
+  "OPENCODE_ZEN_API_KEY", "KILOCODE_API_KEY", "AI_GATEWAY_API_KEY",
+  "CLOUDFLARE_AI_GATEWAY_API_KEY", "LITELLM_API_KEY", "SUNO_API_KEY",
+  "RESEMBLE_API_KEY", "RESEMBLE_VOICE_UUID",
+];
+for (const k of envKeyList) envLines.push(`${k}=${getKey(k)}`);
+
+envLines.push(
+  "",
+  "# TTS",
+  `GROQ_API_KEY=${getKey("GROQ_API_KEY")}`,
+  "USE_GROQ=true",
+  "USE_GROQ_TTS=true",
+  "",
+  "# STT",
+  `DEEPGRAM_API_KEY=${getKey("DEEPGRAM_API_KEY")}`,
+  "",
+  "# Supertonic TTS",
+  "SUPERTONIC_API_URL=http://supertonic:8765"
+);
+
+fs.writeFileSync(".env", envLines.join("\n") + "\n");
+console.log("  Wrote .env");
+
+// ---------------------------------------------------------------------------
+// 3. Write auth-profiles.json (so OpenClaw has working providers on first start)
+// ---------------------------------------------------------------------------
+
+const providerMap = {
+  ANTHROPIC_API_KEY: "anthropic",
+  OPENAI_API_KEY: "openai",
+  GEMINI_API_KEY: "google",
+  OPENROUTER_API_KEY: "openrouter",
+  MISTRAL_API_KEY: "mistral",
+  XAI_API_KEY: "xai",
+  ZAI_API_KEY: "zai",
+  CEREBRAS_API_KEY: "cerebras",
+  TOGETHER_API_KEY: "together",
+  HF_TOKEN: "huggingface",
+  GROQ_API_KEY: "groq",
+  DEEPGRAM_API_KEY: "deepgram",
+  MOONSHOT_API_KEY: "moonshot",
+  KIMI_API_KEY: "kimi-coding",
+  MINIMAX_API_KEY: "minimax",
+  QIANFAN_API_KEY: "qianfan",
+  MODELSTUDIO_API_KEY: "modelstudio",
+  XIAOMI_API_KEY: "xiaomi",
+  VOLCANO_ENGINE_API_KEY: "volcano-engine",
+  BYTEPLUS_API_KEY: "byteplus",
+  SYNTHETIC_API_KEY: "synthetic",
+  VENICE_API_KEY: "venice",
+  OPENCODE_ZEN_API_KEY: "opencode",
+  KILOCODE_API_KEY: "kilocode",
+  AI_GATEWAY_API_KEY: "ai-gateway",
+  CLOUDFLARE_AI_GATEWAY_API_KEY: "cloudflare-ai-gateway",
+  LITELLM_API_KEY: "litellm",
+  SUNO_API_KEY: "suno",
+};
+
+const authProfiles = {};
+let keyCount = 0;
+for (const [envVar, providerId] of Object.entries(providerMap)) {
+  const key = getKey(envVar);
+  if (key) {
+    authProfiles[providerId] = [
+      {
+        id: "default",
+        type: "api_key",
+        key,
+        lastUsed: null,
+        disabled: false,
+        cooldown: null,
+      },
+    ];
+    keyCount++;
+  }
+}
+
+const agentDir = "openclaw-data/agents/main/agent";
+fs.mkdirSync(agentDir, { recursive: true });
+fs.writeFileSync(
+  path.join(agentDir, "auth-profiles.json"),
+  JSON.stringify(authProfiles, null, 2) + "\n"
+);
+console.log(`  Wrote auth-profiles.json (${keyCount} provider(s))`);
+
+// ---------------------------------------------------------------------------
+// 4. Pre-generate device identity and pre-pair it
+// ---------------------------------------------------------------------------
+// OpenClaw requires device pairing for WebSocket connections. Even with
+// dangerouslyDisableDeviceAuth:true, the gateway still requires pairing for
+// the WS protocol (that flag only affects the control UI).
+//
+// Fix: generate an Ed25519 keypair at install time, register the public key
+// in OpenClaw's devices/paired.json, and save the full identity so start.js
+// can inject it into the OpenVoiceUI container via docker exec.
+
+// Generate Ed25519 keypair
+const { publicKey, privateKey } = crypto.generateKeyPairSync("ed25519");
+
+// Raw 32-byte public key — extract via JWK.x (most reliable cross-platform)
+const devJwk = publicKey.export({ format: "jwk" });
+const rawPub = Buffer.from(devJwk.x, "base64url");
+
+// deviceId = SHA256(raw public key) — matches Python's hashlib.sha256(raw_pub).hexdigest()
+const deviceId = crypto.createHash("sha256").update(rawPub).digest("hex");
+
+// PEM formats — OpenVoiceUI Python client uses PEM for signing
+const pubPem = publicKey.export({ type: "spki", format: "pem" });
+const privPem = privateKey.export({ type: "pkcs8", format: "pem" });
+
+// base64url of raw Ed25519 bytes — this is what the gateway compares during WS handshake
+const pubB64url = rawPub.toString("base64url");
+
+// Save full identity for injection into OpenVoiceUI container at start time
+// (Python client reads PEM format for Ed25519 signing)
+const deviceIdentity = {
+  deviceId: deviceId,
+  publicKeyPem: pubPem,
+  privateKeyPem: privPem,
+};
+fs.writeFileSync(
+  "openclaw-data/pre-paired-device.json",
+  JSON.stringify(deviceIdentity, null, 2) + "\n"
+);
+console.log(`  Generated device identity: ${deviceId.slice(0, 16)}...`);
+
+// Pre-register in OpenClaw's devices/paired.json so it accepts this device.
+// Format must match what approveDevicePairing() writes:
+//   - publicKey: base64url of raw Ed25519 bytes (NOT PEM — gateway compares literally)
+//   - role/roles/scopes/approvedScopes: authorization metadata
+//   - tokens: per-role auth tokens for verifyDeviceToken() calls
+fs.mkdirSync("openclaw-data/devices", { recursive: true });
+const nowMs = Date.now();
+const pairingToken = crypto.randomBytes(32).toString("hex");
+const pairedDevices = {};
+pairedDevices[deviceId] = {
+  deviceId: deviceId,
+  publicKey: pubB64url,
+  displayName: "pinokio-openvoiceui",
+  platform: "linux",
+  clientId: "cli",
+  clientMode: "cli",
+  role: "operator",
+  roles: ["operator"],
+  scopes: ["operator.read", "operator.write"],
+  approvedScopes: ["operator.read", "operator.write"],
+  tokens: {
+    operator: {
+      token: pairingToken,
+      role: "operator",
+      scopes: ["operator.read", "operator.write"],
+      createdAtMs: nowMs,
+    },
+  },
+  createdAtMs: nowMs,
+  approvedAtMs: nowMs,
+};
+fs.writeFileSync(
+  "openclaw-data/devices/paired.json",
+  JSON.stringify(pairedDevices, null, 2) + "\n"
+);
+// Clear pending.json — stale entries with silent:false permanently block auto-approval
+fs.writeFileSync("openclaw-data/devices/pending.json", "{}\n");
+console.log("  Wrote devices/paired.json (pre-paired)");
+
+// ---------------------------------------------------------------------------
+// Done
+// ---------------------------------------------------------------------------
+
+console.log("\n  Configuration complete!\n");

package/src/adapters/ClawdBotAdapter.js

@@ -0,0 +1,264 @@
+/**
+ * ClawdBotAdapter — Multi-Agent Framework adapter for ClawdBot / OpenClaw (P6-T2)
+ *
+ * Wraps the existing VoiceSession (P3-T7) and exposes it through the EventBridge
+ * contract so the app shell can treat ClawdBot as a swappable agent adapter.
+ *
+ * Internally:
+ *   - VoiceSession handles STT (Web Speech API), HTTP streaming to /api/conversation,
+ *     TTS playback (TTSPlayer), and canvas/music command parsing.
+ *   - This adapter translates between EventBus events (VoiceSession's internal bus)
+ *     and EventBridge AgentEvents (the shell's canonical event vocabulary).
+ *
+ * Ref: future-dev-plans/17-MULTI-AGENT-FRAMEWORK.md — "ClawdBot Adapter" section
+ *
+ * Adapter contract:
+ *   init(bridge, config)  — called when this mode is selected
+ *   start()               — called when user clicks the call button
+ *   stop()                — called when user clicks stop
+ *   destroy()             — called when switching to a different adapter
+ *
+ * Config shape:
+ *   {
+ *     serverUrl:  string,   // e.g. 'http://localhost:5001'
+ *     sessionKey: string,   // e.g. 'voice-main-3'  (passed to VoiceSession)
+ *     musicPlayer: object,  // optional — MusicPlayer instance from shell
+ *   }
+ */
+
+import { AgentEvents, AgentActions } from '../core/EventBridge.js';
+import { eventBus } from '../core/EventBus.js';
+import { VoiceSession } from '../core/VoiceSession.js';
+
+const ClawdBotAdapter = {
+    // ── Identity & capabilities ───────────────────────────────────────────────
+
+    name: 'ClawdBot (OpenClaw)',
+
+    /**
+     * Feature flags: app shell shows/hides UI elements based on this array.
+     * Ref: doc 17 — "capability-driven UI shell"
+     */
+    capabilities: [
+        'canvas',           // [CANVAS:...] commands parsed and emitted
+        'vps_control',      // agent can run server-side tools
+        'file_ops',         // agent can read/write files on VPS
+        'code_execution',   // agent can execute code on VPS
+        'dj_soundboard',    // future: soundboard integration
+        'music_sync',       // [MUSIC_PLAY/STOP/NEXT] commands parsed and emitted
+        'camera',           // webcam + Gemini vision + face recognition
+    ],
+
+    // ── Private state ─────────────────────────────────────────────────────────
+
+    _bridge:       null,   // EventBridge singleton
+    _session:      null,   // VoiceSession instance
+    _config:       null,   // adapter config passed to init()
+    _eventUnsubs:  [],     // eventBus.on() cleanup functions
+    _bridgeUnsubs: [],     // bridge.on() cleanup functions (belt-and-suspenders)
+
+    // ── Lifecycle ─────────────────────────────────────────────────────────────
+
+    /**
+     * Initialize the adapter.
+     * Called by AgentOrchestrator when this mode is selected.
+     *
+     * @param {import('../core/EventBridge.js').EventBridge} bridge
+     * @param {object} config
+     */
+    async init(bridge, config) {
+        this._bridge = bridge;
+        this._config = config;
+
+        // Create VoiceSession — the existing ClawdBot conversation engine
+        this._session = new VoiceSession({
+            serverUrl:   config.serverUrl  || '',
+            musicPlayer: config.musicPlayer || null,
+        });
+
+        // Translate VoiceSession's internal EventBus events → EventBridge AgentEvents
+        this._wireSessionEvents();
+
+        // Listen for UI→Agent actions emitted on the bridge
+        this._bridgeUnsubs.push(
+            bridge.on(AgentActions.SEND_MESSAGE,   (d) => this._session.sendMessage(d.text))
+        );
+        this._bridgeUnsubs.push(
+            bridge.on(AgentActions.FORCE_MESSAGE,  (d) => this._session.sendMessage(d.text))
+        );
+        this._bridgeUnsubs.push(
+            bridge.on(AgentActions.END_SESSION,    ()  => this.stop())
+        );
+        // CONTEXT_UPDATE: inject background context as a silent system message
+        this._bridgeUnsubs.push(
+            bridge.on(AgentActions.CONTEXT_UPDATE, (d) => this._session.sendMessage(`[CONTEXT: ${d.text}]`))
+        );
+    },
+
+    /**
+     * Start the conversation.
+     * Called when the user clicks the call button.
+     */
+    async start() {
+        if (!this._session) return;
+        // VoiceSession.start() emits session:start on success, which we forward as CONNECTED
+        await this._session.start();
+    },
+
+    /**
+     * Stop the conversation.
+     * Called when the user clicks stop.
+     */
+    async stop() {
+        if (!this._session) return;
+        this._session.stop();
+        // session:stop handler emits DISCONNECTED via the wired event
+    },
+
+    /**
+     * Full teardown — called when switching to a different adapter.
+     * MUST release all resources.
+     */
+    async destroy() {
+        // Stop conversation first
+        this._session?.stop();
+
+        // Destroy VoiceSession's AudioContext
+        this._session?.destroy();
+        this._session = null;
+
+        // Clean up EventBus subscriptions (prevents stale handlers from a dead session)
+        this._eventUnsubs.forEach(fn => fn());
+        this._eventUnsubs = [];
+
+        // Bridge unsubs: AgentOrchestrator calls bridge.clearAll() anyway,
+        // but we clean up explicitly for correctness.
+        this._bridgeUnsubs.forEach(fn => fn());
+        this._bridgeUnsubs = [];
+
+        this._bridge = null;
+        this._config = null;
+    },
+
+    // ── Private: VoiceSession → EventBridge wiring ───────────────────────────
+
+    /**
+     * Subscribe to all EventBus events emitted by VoiceSession and translate
+     * them into EventBridge AgentEvents for the app shell.
+     */
+    _wireSessionEvents() {
+        const b = this._bridge;
+        const push = (unsub) => this._eventUnsubs.push(unsub);
+
+        // ── Connection lifecycle ──────────────────────────────────────────────
+
+        push(eventBus.on('session:start', () => {
+            b.emit(AgentEvents.CONNECTED);
+            b.emit(AgentEvents.MOOD, { mood: 'happy' });
+        }));
+
+        push(eventBus.on('session:stop', () => {
+            b.emit(AgentEvents.DISCONNECTED);
+            b.emit(AgentEvents.MOOD, { mood: 'neutral' });
+        }));
+
+        push(eventBus.on('session:reset', (d) => {
+            // Server-side session was reset (e.g. context overflow)
+            console.info('[ClawdBotAdapter] Session reset:', d.old, '→', d.new);
+        }));
+
+        // ── Conversation state ────────────────────────────────────────────────
+
+        push(eventBus.on('session:thinking', () => {
+            b.emit(AgentEvents.STATE_CHANGED, { state: 'thinking' });
+            b.emit(AgentEvents.MOOD,          { mood: 'thinking' });
+        }));
+
+        push(eventBus.on('session:listening', () => {
+            b.emit(AgentEvents.STATE_CHANGED, { state: 'listening' });
+        }));
+
+        // ── Content ───────────────────────────────────────────────────────────
+
+        // Final messages (user speech → text, assistant response)
+        push(eventBus.on('session:message', (d) => {
+            b.emit(AgentEvents.MESSAGE, {
+                role:  d.role,
+                text:  d.text,
+                final: true,
+            });
+        }));
+
+        // Streaming text deltas (assistant thinking → partial transcript display)
+        push(eventBus.on('session:streaming', (d) => {
+            b.emit(AgentEvents.TRANSCRIPT, { text: d.text, partial: true });
+        }));
+
+        // ── Audio / TTS ───────────────────────────────────────────────────────
+
+        push(eventBus.on('tts:start', () => {
+            b.emit(AgentEvents.TTS_PLAYING);
+            b.emit(AgentEvents.STATE_CHANGED, { state: 'speaking' });
+        }));
+
+        push(eventBus.on('tts:stop', () => {
+            b.emit(AgentEvents.TTS_STOPPED);
+            b.emit(AgentEvents.STATE_CHANGED, { state: 'listening' });
+        }));
+
+        // ── Errors ────────────────────────────────────────────────────────────
+
+        push(eventBus.on('session:error', (d) => {
+            b.emit(AgentEvents.ERROR, { message: d.message });
+            b.emit(AgentEvents.MOOD,  { mood: 'sad' });
+        }));
+
+        // ── Tool calls ────────────────────────────────────────────────────────
+
+        push(eventBus.on('session:tool', (d) => {
+            b.emit(AgentEvents.TOOL_CALLED, { name: d.name, params: {}, result: null });
+        }));
+
+        // ── Emotion / mood ────────────────────────────────────────────────────
+
+        // Server-provided emotion state (ADR-004: mood + intensity + directives)
+        push(eventBus.on('session:emotion', (d) => {
+            if (d.mood) b.emit(AgentEvents.MOOD, { mood: d.mood });
+        }));
+
+        // ── Canvas commands ───────────────────────────────────────────────────
+
+        // [CANVAS_MENU] → open canvas page picker
+        push(eventBus.on('cmd:canvas_menu', () => {
+            b.emit(AgentEvents.CANVAS_CMD, { action: 'menu' });
+        }));
+
+        // [CANVAS:pagename] → present a specific canvas page
+        push(eventBus.on('cmd:canvas_page', (d) => {
+            b.emit(AgentEvents.CANVAS_CMD, { action: 'present', url: d.page });
+        }));
+
+        // ── Music commands ────────────────────────────────────────────────────
+
+        // [MUSIC_PLAY] or [MUSIC_PLAY:trackname]
+        push(eventBus.on('cmd:music_play', (d) => {
+            b.emit(AgentEvents.MUSIC_PLAY, {
+                action: 'play',
+                track:  d.track || null,
+            });
+        }));
+
+        // [MUSIC_STOP]
+        push(eventBus.on('cmd:music_stop', () => {
+            b.emit(AgentEvents.MUSIC_PLAY, { action: 'stop' });
+        }));
+
+        // [MUSIC_NEXT]
+        push(eventBus.on('cmd:music_next', () => {
+            b.emit(AgentEvents.MUSIC_PLAY, { action: 'skip' });
+        }));
+    },
+};
+
+export default ClawdBotAdapter;
+export { ClawdBotAdapter };