pi-harness-runtime 1.1.337 → 1.1.339

package/packages/voice-api/src/transcriber.ts

@@ -0,0 +1,136 @@
+/**
+ * Voice Transcriber - Main interface for speech-to-text
+ */
+
+import type { TranscriptionResult, TranscribeOptions, VoiceApiConfig } from './types.js';
+import { WhisperProvider, createOpenAIProvider, createReplicateProvider } from './provider.js';
+
+/**
+ * Voice Transcriber
+ * 
+ * Main interface for transcribing voice messages to text.
+ * Supports multiple providers (OpenAI, Replicate).
+ */
+export class VoiceTranscriber {
+  private providers: Map<string, WhisperProvider> = new Map();
+  private defaultProvider: string = '';
+
+  constructor(config: VoiceApiConfig = {}) {
+    // Initialize OpenAI provider
+    if (config.openaiApiKey) {
+      const provider = createOpenAIProvider(config.openaiApiKey);
+      this.providers.set('openai', provider);
+      this.defaultProvider = 'openai';
+    }
+
+    // Initialize Replicate provider
+    if (config.replicateApiToken) {
+      const provider = createReplicateProvider(config.replicateApiToken);
+      this.providers.set('replicate', provider);
+      if (!this.defaultProvider) {
+        this.defaultProvider = 'replicate';
+      }
+    }
+
+    // Set default provider from config
+    if (config.provider && this.providers.has(config.provider)) {
+      this.defaultProvider = config.provider;
+    }
+  }
+
+  /**
+   * Check if transcriber is configured
+   */
+  isConfigured(): boolean {
+    return this.providers.size > 0;
+  }
+
+  /**
+   * Get configured providers
+   */
+  getProviders(): string[] {
+    return Array.from(this.providers.keys());
+  }
+
+  /**
+   * Get default provider
+   */
+  getDefaultProvider(): string | null {
+    return this.defaultProvider || null;
+  }
+
+  /**
+   * Transcribe audio to text
+   */
+  async transcribe(options: TranscribeOptions): Promise<TranscriptionResult> {
+    const providerName = options.provider || this.defaultProvider;
+    
+    if (!providerName) {
+      return {
+        text: '',
+        timestamp: new Date().toISOString(),
+        provider: 'openai',
+        error: 'No provider configured',
+      };
+    }
+
+    const provider = this.providers.get(providerName);
+    if (!provider) {
+      return {
+        text: '',
+        timestamp: new Date().toISOString(),
+        provider: providerName as 'openai' | 'replicate',
+        error: `Provider ${providerName} not configured`,
+      };
+    }
+
+    return provider.transcribe(options);
+  }
+
+  /**
+   * Add a provider
+   */
+  addProvider(name: string, provider: WhisperProvider): void {
+    this.providers.set(name, provider);
+    if (!this.defaultProvider) {
+      this.defaultProvider = name;
+    }
+  }
+
+  /**
+   * Remove a provider
+   */
+  removeProvider(name: string): boolean {
+    const deleted = this.providers.delete(name);
+    if (this.defaultProvider === name) {
+      this.defaultProvider = this.providers.keys().next().value || '';
+    }
+    return deleted;
+  }
+
+  /**
+   * Set default provider
+   */
+  setDefaultProvider(name: string): boolean {
+    if (this.providers.has(name)) {
+      this.defaultProvider = name;
+      return true;
+    }
+    return false;
+  }
+}
+
+/**
+ * Create a VoiceTranscriber from environment variables
+ */
+export function createFromEnv(): VoiceTranscriber {
+  // SAFETY: globalThis.process is always available in Node.js
+  const env = (globalThis as unknown as { process: { env: Record<string, string> } }).process;
+  const envVars = env?.env ?? {};
+  
+  return new VoiceTranscriber({
+    openaiApiKey: envVars.OPENAI_API_KEY,
+    replicateApiToken: envVars.REPLICATE_API_TOKEN,
+    provider: (envVars.VOICE_PROVIDER as 'openai' | 'replicate') || 'openai',
+  });
+}

package/packages/voice-api/src/types.ts

@@ -0,0 +1,79 @@
+/**
+ * Voice API Types
+ */
+
+/**
+ * Transcription result from Whisper
+ */
+export interface TranscriptionResult {
+  /** The transcribed text */
+  text: string;
+  /** Language detected (ISO 639-1) */
+  language?: string;
+  /** Duration in seconds */
+  duration?: number;
+  /** Confidence score (0-1) */
+  confidence?: number;
+  /** Timestamp of transcription */
+  timestamp: string;
+  /** Provider used (openai/replicate) */
+  provider: 'openai' | 'replicate';
+  /** Error message if failed */
+  error?: string;
+}
+
+/**
+ * Voice message from messaging platforms
+ */
+export interface VoiceMessage {
+  /** Unique message ID */
+  id: string;
+  /** Platform source (telegram/line) */
+  platform: 'telegram' | 'line';
+  /** Chat/user ID */
+  chatId: string;
+  /** Audio file URL or path */
+  audioUrl: string;
+  /** Duration in seconds */
+  duration?: number;
+  /** MIME type */
+  mimeType: string;
+  /** File size in bytes */
+  fileSize?: number;
+  /** Raw platform-specific data */
+  raw?: unknown;
+}
+
+/**
+ * Voice API configuration
+ */
+export interface VoiceApiConfig {
+  /** OpenAI API key (for Whisper) */
+  openaiApiKey?: string;
+  /** Replicate API token (alternative) */
+  replicateApiToken?: string;
+  /** Default provider */
+  provider?: 'openai' | 'replicate';
+  /** Model to use */
+  model?: string;
+  /** Language hint (ISO 639-1) */
+  language?: string;
+  /** Temperature for generation */
+  temperature?: number;
+}
+
+/**
+ * Transcribe options
+ */
+export interface TranscribeOptions {
+  /** Audio URL or file path (required if no audioBuffer) */
+  audioUrl?: string;
+  /** Audio as Buffer (Uint8Array for cross-platform) */
+  audioBuffer?: Uint8Array;
+  /** Language hint */
+  language?: string;
+  /** Model override */
+  model?: string;
+  /** Provider override */
+  provider?: 'openai' | 'replicate';
+}

package/packages/webmcp-tester/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pi-harness/webmcp-tester",
-  "version": "1.1.337",
+  "version": "1.1.339",
   "description": "WebMCP-powered E2E testing for pi-harness - natural language test scenarios",
   "type": "module",
   "main": "./dist/index.js",

package/packages/wiki/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pi-harness/wiki",
-  "version": "1.1.337",
+  "version": "1.1.339",
   "description": "LLM Wiki integration for pi-harness-runtime",
   "type": "module",
   "main": "./dist/index.js",

package/packages/workflow-events/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pi/workflow-events",
-  "version": "1.1.337",
+  "version": "1.1.339",
   "type": "module",
   "main": "src/index.ts",
   "exports": {

package/packages/workspace-scanner/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pi/workspace-scanner",
-  "version": "1.1.337",
+  "version": "1.1.339",
   "type": "module",
   "main": "src/index.ts",
   "exports": {

package/packages/worktree/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@pi-harness/worktree",
-  "version": "1.1.337",
+  "version": "1.1.339",
   "description": "pi-harness-runtime — worktree",
   "type": "module",
   "main": "./dist/index.js",

package/packages/write-review/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@moocoding/write-review",
-  "version": "1.1.337",
+  "version": "1.1.339",
   "description": "Two-agent write-review loop with blackboard coordination",
   "type": "module",
   "main": "src/index.ts",

package/scripts/voice-api-example.ts

@@ -0,0 +1,112 @@
+/**
+ * Voice API Example - Telegram Voice to Command
+ * 
+ * This script demonstrates how to:
+ * 1. Receive voice message from Telegram
+ * 2. Transcribe with Whisper API
+ * 3. Execute as pi-harness command
+ * 
+ * Run:
+ *   OPENAI_API_KEY=sk-xxx bun run scripts/voice-api-example.ts
+ */
+
+import * as fs from "node:fs";
+import * as path from "node:path";
+import { VoiceTranscriber } from "../packages/voice-api/dist/transcriber.js";
+import { TelegramVoiceHandler } from "../packages/voice-api/dist/telegram.js";
+
+// Load API keys from key files
+function loadKey(filename: string): string | null {
+  const keysDir = path.join(process.env.HOME || "/home/frappe", ".pi-harness-runtime", "keys");
+  const filepath = path.join(keysDir, filename);
+  try {
+    return fs.readFileSync(filepath, "utf8").trim();
+  } catch {
+    return null;
+  }
+}
+
+async function main() {
+  console.log("=".repeat(60));
+  console.log("Voice API Example - Telegram Voice to Command");
+  console.log("=".repeat(60));
+
+  // Load API keys
+  const openaiApiKey = process.env.OPENAI_API_KEY || loadKey("openai-api-key.txt");
+  const telegramBotToken = loadKey("telegram-bot-token.txt");
+
+  if (!openaiApiKey) {
+    console.log("❌ OpenAI API key not found");
+    console.log("   Set OPENAI_API_KEY or create ~/.pi-harness-runtime/keys/openai-api-key.txt");
+    process.exit(1);
+  }
+
+  console.log("✓ OpenAI API key loaded");
+
+  // Create transcriber
+  const transcriber = new VoiceTranscriber({ openaiApiKey });
+  console.log("✓ VoiceTranscriber created");
+
+  if (!transcriber.isConfigured()) {
+    console.log("❌ Transcriber not configured");
+    process.exit(1);
+  }
+
+  console.log(`✓ Providers: ${transcriber.getProviders().join(", ")}`);
+  console.log(`✓ Default: ${transcriber.getDefaultProvider()}`);
+
+  // If Telegram token provided, create handler
+  if (telegramBotToken) {
+    console.log("✓ Telegram bot token loaded");
+
+    const handler = new TelegramVoiceHandler({
+      botToken: telegramBotToken,
+      transcriber,
+      onTranscribed: async (chatId, text, result) => {
+        console.log("\n" + "=".repeat(60));
+        console.log("Voice transcribed:");
+        console.log(`  Chat: ${chatId}`);
+        console.log(`  Text: ${text}`);
+        console.log(`  Provider: ${result.provider}`);
+        console.log(`  Language: ${result.language || "auto"}`);
+        console.log("=".repeat(60));
+        
+        // Here you would execute the command in pi-harness
+        console.log("\n[pi-harness] Would execute:");
+        console.log(`  ${text}`);
+      },
+      onError: (chatId, error) => {
+        console.log(`❌ Error for chat ${chatId}: ${error.message}`);
+      },
+    });
+
+    console.log("✓ TelegramVoiceHandler created");
+    console.log("\nListening for Telegram voice messages...");
+    console.log("(This is a demo - actual webhook server not started)");
+  } else {
+    console.log("⚠ Telegram bot token not found");
+    console.log("   Create ~/.pi-harness-runtime/keys/telegram-bot-token.txt for Telegram integration");
+  }
+
+  // Demo: transcribe a local audio file
+  console.log("\n" + "-".repeat(60));
+  console.log("Demo: Transcribing test audio file...");
+  
+  // Note: In production, this would be triggered by actual Telegram/LINE webhooks
+  // For testing, you can manually transcribe an audio file:
+  //
+  // const result = await transcriber.transcribe({
+  //   audioUrl: "https://example.com/audio.ogg",
+  //   language: "en",
+  // });
+  // console.log("Transcribed:", result.text);
+
+  console.log("\n" + "=".repeat(60));
+  console.log("Ready for voice commands!");
+  console.log("=".repeat(60));
+}
+
+main().catch(err => {
+  console.error("❌ Error:", err);
+  process.exit(1);
+});