pi-voice 0.1.1 → 0.3.3

package/out/main/index.js

@@ -2,16 +2,39 @@ import { systemPreferences, app, BrowserWindow, ipcMain } from "electron";
 import { fileURLToPath } from "node:url";
 import { uIOhook, UiohookKey } from "uiohook-napi";
 import { join } from "node:path";
+import { homedir } from "node:os";
+import pino from "pino";
 import { readFileSync, existsSync, mkdirSync, createWriteStream, unlinkSync, writeFileSync } from "node:fs";
-import { GoogleGenAI } from "@google/genai";
+import { z } from "zod";
 import OpenAI, { toFile } from "openai";
 import { WhisperFullParams, WhisperSamplingStrategy, Whisper } from "@napi-rs/whisper";
-import { homedir } from "node:os";
+import { GoogleGenAI } from "@google/genai";
 import { Readable } from "node:stream";
 import { finished } from "node:stream/promises";
 import { spawn } from "node:child_process";
 import { createAgentSession, SessionManager } from "@mariozechner/pi-coding-agent";
 import { createServer } from "node:net";
+function resolveLogPath() {
+  const envPath = process.env["PI_VOICE_LOG_PATH"];
+  if (envPath) return envPath;
+  const configHome = process.env["XDG_CONFIG_HOME"] || join(homedir(), ".config");
+  return join(configHome, "pi-voice", "daemon.log");
+}
+const logPath = resolveLogPath();
+const logger = pino(
+  {
+    level: "debug"
+  },
+  pino.multistream([
+    // Console output (human-readable via stdout)
+    { level: "debug", stream: process.stdout },
+    // File output (JSON, auto-creates parent directories)
+    {
+      level: "debug",
+      stream: pino.destination({ dest: logPath, mkdir: true, sync: false })
+    }
+  ])
+);
 function getReleaseCodes(binding) {
   const codes = [binding.keycode];
   if (binding.ctrl) {
@@ -67,14 +90,14 @@ class FnHook {
     });
     uIOhook.start();
     this.started = true;
-    console.log(`[FnHook] Started monitoring ${this.displayName}`);
+    logger.info({ key: this.displayName }, "Started monitoring key");
   }
   stop() {
     if (!this.started) return;
     uIOhook.stop();
     this.started = false;
     this.active = false;
-    console.log("[FnHook] Stopped monitoring");
+    logger.info("Stopped monitoring key");
   }
   get isFnDown() {
     return this.active;
@@ -216,50 +239,88 @@ function formatKeyDisplay(binding) {
 }
 const DEFAULT_KEY_STRING = "meta+shift+i";
 const DEFAULT_PROVIDER = "local";
-const VALID_PROVIDERS = ["local", "gemini", "openai"];
 function defaultConfig() {
+  const binding = parseKeyBinding(DEFAULT_KEY_STRING);
   return {
-    key: parseKeyBinding(DEFAULT_KEY_STRING),
-    keyDisplay: formatKeyDisplay(parseKeyBinding(DEFAULT_KEY_STRING)),
+    key: binding,
+    keyDisplay: formatKeyDisplay(binding),
     provider: DEFAULT_PROVIDER
   };
 }
+const configFileSchema = z.object({
+  key: z.string().refine(
+    (v) => {
+      try {
+        parseKeyBinding(v);
+        return true;
+      } catch {
+        return false;
+      }
+    },
+    { message: "Invalid key binding" }
+  ).optional().default(DEFAULT_KEY_STRING),
+  provider: z.enum(["local", "gemini", "openai"]).optional().default(DEFAULT_PROVIDER)
+});
+class ConfigError extends Error {
+  constructor(configPath, details) {
+    super(`Invalid config at ${configPath}:
+${details}`);
+    this.configPath = configPath;
+    this.details = details;
+    this.name = "ConfigError";
+  }
+}
 function loadConfig(cwd) {
   const configPath = join(cwd, ".pi", "pi-voice.json");
+  let raw;
   try {
-    const raw = readFileSync(configPath, "utf-8");
-    const json = JSON.parse(raw);
-    let binding;
-    let display;
-    if (typeof json.key === "string") {
-      binding = parseKeyBinding(json.key);
-      display = formatKeyDisplay(binding);
-    } else {
-      if (json.key !== void 0) {
-        console.warn(`[Config] "key" must be a string in ${configPath}, using default`);
-      }
-      binding = parseKeyBinding(DEFAULT_KEY_STRING);
-      display = formatKeyDisplay(binding);
-    }
-    let provider = DEFAULT_PROVIDER;
-    if (typeof json.provider === "string") {
-      const p = json.provider.toLowerCase();
-      if (VALID_PROVIDERS.includes(p)) {
-        provider = p;
-      } else {
-        console.warn(`[Config] Unknown provider "${json.provider}" in ${configPath}, using default "${DEFAULT_PROVIDER}"`);
-      }
-    }
-    console.log(`[Config] Loaded config: key=${display}, provider=${provider} from ${configPath}`);
-    return { key: binding, keyDisplay: display, provider };
+    raw = readFileSync(configPath, "utf-8");
   } catch (err) {
     if (err.code === "ENOENT") {
-      console.log(`[Config] No config file found at ${configPath}, using default (${DEFAULT_KEY_STRING})`);
-    } else {
-      console.warn(`[Config] Failed to load ${configPath}, using default:`, err);
+      logger.info({ configPath }, "No config file found, using defaults");
+      return defaultConfig();
     }
-    return defaultConfig();
+    throw new ConfigError(configPath, `Failed to read file: ${err.message}`);
   }
+  let json;
+  try {
+    json = JSON.parse(raw);
+  } catch {
+    throw new ConfigError(configPath, "Invalid JSON syntax");
+  }
+  const result = configFileSchema.safeParse(json);
+  if (!result.success) {
+    const details = result.error.issues.map((issue) => {
+      const path = issue.path.length > 0 ? `"${issue.path.join(".")}"` : "(root)";
+      return `  - ${path}: ${issue.message}`;
+    }).join("\n");
+    throw new ConfigError(configPath, details);
+  }
+  const parsed = result.data;
+  const binding = parseKeyBinding(parsed.key);
+  const display = formatKeyDisplay(binding);
+  logger.info({ key: display, provider: parsed.provider, configPath }, "Loaded config");
+  return { key: binding, keyDisplay: display, provider: parsed.provider };
+}
+let geminiClient = null;
+function getGeminiClient() {
+  if (geminiClient) return geminiClient;
+  const forceVertexOff = process.env.GOOGLE_GENAI_USE_VERTEXAI === "false";
+  const project = process.env.GOOGLE_CLOUD_PROJECT;
+  const location = process.env.GOOGLE_CLOUD_LOCATION ?? "us-central1";
+  const apiKey = process.env.GEMINI_API_KEY ?? process.env.GOOGLE_API_KEY;
+  if (project && !forceVertexOff) {
+    logger.info({ project, location }, "Initializing Gemini client (Vertex AI)");
+    geminiClient = new GoogleGenAI({ vertexai: true, project, location });
+  } else if (apiKey) {
+    logger.info("Initializing Gemini client (API key)");
+    geminiClient = new GoogleGenAI({ apiKey });
+  } else {
+    throw new Error(
+      "Gemini provider requires either GOOGLE_CLOUD_PROJECT (for Vertex AI) or GEMINI_API_KEY / GOOGLE_API_KEY (for Gemini API)."
+    );
+  }
+  return geminiClient;
 }
 const DEFAULT_MODEL = "medium-q5_0";
 const HF_BASE_URL = "https://huggingface.co/ggerganov/whisper.cpp/resolve/main";
@@ -271,7 +332,7 @@ function modelFileName(model) {
 }
 async function downloadModel(model, destPath) {
   const url = `${HF_BASE_URL}/${modelFileName(model)}`;
-  console.log(`[Whisper] Downloading model "${model}" from ${url} ...`);
+  logger.info({ model, url }, "Downloading Whisper model");
   const response = await fetch(url, { method: "GET", redirect: "follow" });
   if (!response.ok) {
     throw new Error(
@@ -323,7 +384,7 @@ async function downloadModel(model, destPath) {
     process.stderr.write("\n");
     const { renameSync } = await import("node:fs");
     renameSync(tmpPath, destPath);
-    console.log(`[Whisper] Model saved to ${destPath}`);
+    logger.info({ destPath }, "Whisper model saved");
   } catch (err) {
     try {
       unlinkSync(tmpPath);
@@ -352,17 +413,6 @@ async function resolveModelPath() {
   await downloadModel(model, destPath);
   return destPath;
 }
-let geminiClient$1 = null;
-function getGeminiClient$1() {
-  if (geminiClient$1) return geminiClient$1;
-  const project = process.env.GOOGLE_CLOUD_PROJECT;
-  const location = process.env.GOOGLE_CLOUD_LOCATION ?? "us-central1";
-  if (!project) {
-    throw new Error("GOOGLE_CLOUD_PROJECT environment variable is required");
-  }
-  geminiClient$1 = new GoogleGenAI({ vertexai: true, project, location });
-  return geminiClient$1;
-}
 let openaiClient$1 = null;
 function getOpenAIClient$1() {
   if (openaiClient$1) return openaiClient$1;
@@ -380,15 +430,15 @@ async function getWhisperInstance() {
   if (whisperInitPromise) return whisperInitPromise;
   whisperInitPromise = (async () => {
     const modelPath = await resolveModelPath();
-    console.log("[STT:local] Loading Whisper model from", modelPath, "...");
+    logger.info({ modelPath }, "Loading Whisper model");
     whisperInstance = new Whisper(modelPath);
-    console.log("[STT:local] Whisper model loaded");
+    logger.info("Whisper model loaded");
     return whisperInstance;
   })();
   return whisperInitPromise;
 }
 async function transcribeGemini(audioBuffer) {
-  const client = getGeminiClient$1();
+  const client = getGeminiClient();
   const base64Audio = audioBuffer.toString("base64");
   const response = await client.models.generateContent({
     model: "gemini-2.5-flash",
@@ -447,20 +497,9 @@ async function transcribe(audioData, provider = "local") {
       text = await transcribeGemini(Buffer.from(audioData));
       break;
   }
-  console.log(`[STT:${provider}] Transcribed: "${text}"`);
+  logger.info({ provider, text }, "Transcribed");
   return text;
 }
-let geminiClient = null;
-function getGeminiClient() {
-  if (geminiClient) return geminiClient;
-  const project = process.env.GOOGLE_CLOUD_PROJECT;
-  const location = process.env.GOOGLE_CLOUD_LOCATION ?? "us-central1";
-  if (!project) {
-    throw new Error("GOOGLE_CLOUD_PROJECT environment variable is required");
-  }
-  geminiClient = new GoogleGenAI({ vertexai: true, project, location });
-  return geminiClient;
-}
 let openaiClient = null;
 function getOpenAIClient() {
   if (openaiClient) return openaiClient;
@@ -525,8 +564,9 @@ async function* synthesizeStreamGemini(text) {
     totalBytes += leftover.length;
     yield leftover;
   }
-  console.log(
-    `[TTS:gemini] Streamed ${totalBytes} bytes of PCM audio for "${text.substring(0, 50)}..."`
+  logger.info(
+    { provider: "gemini", totalBytes, text: text.substring(0, 50) },
+    "Streamed PCM audio"
   );
 }
 async function* synthesizeStreamOpenAI(text) {
@@ -549,8 +589,9 @@ async function* synthesizeStreamOpenAI(text) {
     yield chunk;
     offset = end;
   }
-  console.log(
-    `[TTS:openai] Streamed ${totalBytes} bytes of PCM audio for "${text.substring(0, 50)}..."`
+  logger.info(
+    { provider: "openai", totalBytes, text: text.substring(0, 50) },
+    "Streamed PCM audio"
   );
 }
 function speakLocal(text) {
@@ -571,8 +612,9 @@ function speakLocal(text) {
     });
     child.on("close", (code) => {
       if (code === 0) {
-        console.log(
-          `[TTS:local] Spoke "${text.substring(0, 50)}${text.length > 50 ? "..." : ""}"`
+        logger.info(
+          { provider: "local", text: text.substring(0, 50) },
+          "Spoke text"
         );
         resolve();
       } else {
@@ -603,13 +645,13 @@ function setSessionCwd(cwd) {
 }
 async function getOrCreateSession() {
   if (session) return session;
-  console.log(`[PiSession] Creating new agent session (cwd: ${sessionCwd})...`);
+  logger.info({ cwd: sessionCwd }, "Creating new agent session");
   const result = await createAgentSession({
     cwd: sessionCwd,
     sessionManager: SessionManager.inMemory()
   });
   session = result.session;
-  console.log("[PiSession] Session created");
+  logger.info("Agent session created");
   return session;
 }
 async function prompt(text, options) {
@@ -618,7 +660,7 @@ async function prompt(text, options) {
     if (event.type === "message_update" && event.assistantMessageEvent.type === "text_end") {
       const content = event.assistantMessageEvent.content.trim();
       if (content.length > 0) {
-        console.log(`[PiSession] Response: ${content}`);
+        logger.info({ content }, "Agent response");
         options?.onTextEnd?.(content);
       }
     }
@@ -633,7 +675,7 @@ function dispose() {
   if (session) {
     session.dispose();
     session = null;
-    console.log("[PiSession] Session disposed");
+    logger.info("Agent session disposed");
   }
 }
 const IPC = {
@@ -706,7 +748,7 @@ function startDaemonServer(handler) {
     });
   });
   server.listen(socketPath);
-  console.log(`[DaemonIPC] Listening on ${socketPath}`);
+  logger.info({ socketPath }, "DaemonIPC listening");
   return socketPath;
 }
 function stopDaemonServer() {
@@ -721,7 +763,7 @@ function stopDaemonServer() {
     } catch {
     }
   }
-  console.log("[DaemonIPC] Server stopped");
+  logger.info("DaemonIPC server stopped");
 }
 const workingCwd = process.env["PI_VOICE_CWD"] || process.cwd();
 let mainWindow = null;
@@ -730,7 +772,7 @@ let currentState = "idle";
 setSessionCwd(workingCwd);
 function setState(state, message) {
   currentState = state;
-  console.log(`[Main] State: ${state}${message ? ` - ${message}` : ""}`);
+  logger.info({ state, message }, "State changed");
 }
 function createWindow() {
   mainWindow = new BrowserWindow({
@@ -764,7 +806,7 @@ function setupIpcHandlers(provider) {
   ipcMain.on(IPC.RECORDING_DATA, async (_event, data) => {
     if (currentState !== "recording") return;
     if (data.byteLength < 1e3) {
-      console.log("[Main] Recording too short, ignoring");
+      logger.info("Recording too short, ignoring");
       setState("idle", "Recording too short");
       return;
     }
@@ -830,7 +872,7 @@ function setupIpcHandlers(provider) {
       }
     } catch (err) {
       const msg = err instanceof Error ? err.message : String(err);
-      console.error("[Main] Pipeline error:", msg);
+      logger.error({ err: msg }, "Pipeline error");
       setState("error", msg);
       setTimeout(() => {
         if (currentState === "error") setState("idle");
@@ -838,7 +880,7 @@ function setupIpcHandlers(provider) {
     }
   });
   ipcMain.on(IPC.RECORDING_ERROR, (_event, error) => {
-    console.error("[Main] Recording error:", error);
+    logger.error({ err: error }, "Recording error");
     setState("error", error);
     setTimeout(() => {
       if (currentState === "error") setState("idle");
@@ -856,8 +898,9 @@ function setupFnHook(config) {
     {
       onFnDown: () => {
         if (currentState !== "idle") {
-          console.log(
-            `[Main] ${config.keyDisplay} pressed but state is ${currentState}, ignoring`
+          logger.info(
+            { key: config.keyDisplay, state: currentState },
+            "Key pressed but not idle, ignoring"
           );
           return;
         }
@@ -876,7 +919,7 @@ function setupFnHook(config) {
     fnHook.start();
   } catch (err) {
     const msg = err instanceof Error ? err.message : String(err);
-    console.error("[Main] FnHook error:", msg);
+    logger.error({ err: msg }, "FnHook error");
     setState("error", msg);
   }
 }
@@ -900,7 +943,7 @@ function handleDaemonCommand(command) {
   }
 }
 function gracefulShutdown() {
-  console.log("[Main] Shutting down...");
+  logger.info("Shutting down...");
   fnHook?.stop();
   dispose();
   stopDaemonServer();
@@ -912,17 +955,28 @@ process.on("SIGTERM", () => {
 });
 const gotLock = app.requestSingleInstanceLock();
 if (!gotLock) {
-  console.log("[Main] Another instance is already running. Exiting.");
+  logger.warn("Another instance is already running. Exiting.");
   app.quit();
 }
 app.whenReady().then(async () => {
-  const config = loadConfig(workingCwd);
+  let config;
+  try {
+    config = loadConfig(workingCwd);
+  } catch (err) {
+    if (err instanceof ConfigError) {
+      logger.error({ err: err.message }, "Config error");
+    } else {
+      logger.error({ err: err instanceof Error ? err.message : String(err) }, "Failed to load config");
+    }
+    app.quit();
+    return;
+  }
   if (config.provider === "local") {
     try {
       await resolveModelPath();
     } catch (err) {
       const msg = err instanceof Error ? err.message : String(err);
-      console.error("[Main] Failed to prepare Whisper model:", msg);
+      logger.error({ err: msg }, "Failed to prepare Whisper model");
       app.quit();
       return;
     }
@@ -932,7 +986,7 @@ app.whenReady().then(async () => {
   setupFnHook(config);
   startDaemonServer(handleDaemonCommand);
   saveRuntimeState(workingCwd);
-  console.log(`[Main] pi-voice daemon started (cwd: ${workingCwd})`);
+  logger.info({ cwd: workingCwd }, "pi-voice daemon started");
 });
 app.on("window-all-closed", () => {
 });

package/package.json

@@ -1,9 +1,13 @@
 {
   "name": "pi-voice",
-  "version": "0.1.1",
+  "version": "0.3.3",
   "description": "Voice interface for pi coding agent",
   "author": "Yuku Kotani",
   "license": "MIT",
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/yukukotani/pi-voice"
+  },
   "type": "module",
   "main": "./out/main/index.js",
   "bin": {
@@ -37,7 +41,9 @@
     "@napi-rs/whisper": "^0.0.4",
     "electron": "^40.2.1",
     "openai": "^6.10.0",
-    "uiohook-napi": "^1.5.4"
+    "pino": "^10.3.1",
+    "uiohook-napi": "^1.5.4",
+    "zod": "^4.3.6"
   },
   "trustedDependencies": [
     "electron",