support Espeak and Piper speach engines and notifications only for long-running tasks

qwen3-tts-server/setup.sh

@@ -41,7 +41,7 @@ fi
 echo "  Python: $py_version"
 
 # ── Virtual environment ──────────────────────────────────────────────────────
-if [[ -d venv ]] && [[ ! -x venv/bin/python3 ]] && [[ ! -x venv/bin/python ]]; then
+if [[ -d venv ]] && { [[ ! -x venv/bin/python3 ]] && [[ ! -x venv/bin/python ]]; } || [[ ! -e venv/bin/pip ]]; then
     warn "Existing virtual environment looks stale — recreating it."
     rm -rf venv
 fi

src/audio.js

@@ -237,7 +237,9 @@ function getPlaybackCommand(platform, volume, cachePath) {
     return { cmd: "ffplay", args: ["-nodisp", "-autoexit", "-loglevel", "quiet", cachePath] };
   }
   if (platform === "linux") {
-    return { cmd: "ffplay", args: ["-nodisp", "-autoexit", "-loglevel", "quiet", cachePath] };
+    try { execSync("command -v ffplay", { stdio: "ignore" }); return { cmd: "ffplay", args: ["-nodisp", "-autoexit", "-loglevel", "quiet", cachePath] }; } catch {}
+    try { execSync("command -v paplay", { stdio: "ignore" }); return { cmd: "paplay", args: [cachePath] }; } catch {}
+    try { execSync("command -v pw-play", { stdio: "ignore" }); return { cmd: "pw-play", args: [cachePath] }; } catch {}
   }
   return null;
 }
@@ -344,6 +346,35 @@ function downloadChatterbox(phrase, cachePath, config, voicePath, ttsParams) {
   });
 }
 
+function downloadEspeak(phrase, cachePath) {
+  return new Promise((resolve) => {
+    const proc = spawn("espeak-ng", [
+      "-w", cachePath,
+      "-s", "150",
+      phrase,
+    ], { stdio: ["ignore", "ignore", "ignore"] });
+    proc.on("close", () => resolve());
+    proc.on("error", () => resolve());
+  });
+}
+
+function downloadPiper(phrase, cachePath, config) {
+  return new Promise((resolve) => {
+    const piperBin = config.piper_binary || process.env.PIPER_BIN || "";
+    const piperModel = config.piper_model || "";
+    if (!piperModel) return resolve();
+    const args = ["-m", piperModel, "-f", cachePath];
+    const extraArgs = config.piper_args || [];
+    const proc = spawn(piperBin || "piper", [...args, ...extraArgs], {
+      stdio: ["pipe", "ignore", "ignore"],
+    });
+    proc.stdin.write(phrase);
+    proc.stdin.end();
+    proc.on("close", () => resolve());
+    proc.on("error", () => resolve());
+  });
+}
+
 function downloadQwen(phrase, cachePath, config, voiceId) {
   return new Promise((resolve) => {
     const qwenUrl = config.qwen_tts_url || "http://localhost:8100";
@@ -396,6 +427,12 @@ function downloadQwen(phrase, cachePath, config, voiceId) {
 }
 
 async function downloadToCache(phrase, cachePath, config, voicePath, ttsParams, refText) {
+  if (config.tts_backend === "espeak") {
+    return downloadEspeak(phrase, cachePath);
+  }
+  if (config.tts_backend === "piper") {
+    return downloadPiper(phrase, cachePath, config);
+  }
   if (config.tts_backend === "qwen") {
     const voiceId = await registerVoiceWithQwen(config, voicePath, refText);
     return downloadQwen(phrase, cachePath, config, voiceId);

src/llm.js

@@ -190,6 +190,7 @@ function generatePhraseLocal(context, config, style, llmTemperature, examples) {
     const model = local.model || "default";
     const maxTokens = local.max_tokens || 50;
     const timeout = local.timeout || 15000;
+    const apiKey = local.api_key || config.llm_api_key || process.env.ANTHROPIC_AUTH_TOKEN || process.env.API_KEY || "";
 
     const messages = [
       { role: "system", content: buildSystemPrompt(style, "status-report", examples) },
@@ -213,14 +214,17 @@ function generatePhraseLocal(context, config, style, llmTemperature, examples) {
     const isHttps = url.protocol === "https:";
     const reqFn = isHttps ? httpsRequest : httpRequest;
 
+    const headers = {
+      "Content-Type": "application/json",
+      "Content-Length": Buffer.byteLength(payload),
+    };
+    if (apiKey) headers.Authorization = `Bearer ${apiKey}`;
+
     const req = reqFn(
       url,
       {
         method: "POST",
-        headers: {
-          "Content-Type": "application/json",
-          "Content-Length": Buffer.byteLength(payload),
-        },
+        headers,
         timeout,
       },
       (res) => {

src/voxlert.js

@@ -7,17 +7,20 @@
  */
 
 import { resolvePrefix, DEFAULT_PREFIX } from "./prefix.js";
-import { appendFileSync, mkdirSync } from "fs";
+import { appendFileSync, mkdirSync, readFileSync, writeFileSync, existsSync } from "fs";
 import { hostname } from "os";
 import { execSync } from "child_process";
 import { loadConfig, EVENT_MAP, CONTEXTUAL_EVENTS, FALLBACK_PHRASES } from "./config.js";
 import { extractContext, generatePhrase } from "./llm.js";
 import { speakPhrase } from "./audio.js";
 import { showOverlay } from "./overlay.js";
 import { loadPack } from "./packs.js";
+import { join } from "path";
 import { STATE_DIR, LOG_FILE, HOOK_DEBUG_LOG } from "./paths.js";
 import { appendLog } from "./activity-log.js";
 
+const TASK_START_FILE = join(STATE_DIR, "task-start.json");
+
 function debugLog(msg, data) {
   try {
     mkdirSync(STATE_DIR, { recursive: true });
@@ -43,6 +46,31 @@ function logFallback(eventName, reason, detail) {
   }
 }
 
+function getTaskStartTime() {
+  try {
+    if (existsSync(TASK_START_FILE)) {
+      const data = JSON.parse(readFileSync(TASK_START_FILE, "utf-8"));
+      return data.timestamp || 0;
+    }
+  } catch {}
+  return 0;
+}
+
+function setTaskStartTime() {
+  try {
+    mkdirSync(STATE_DIR, { recursive: true });
+    writeFileSync(TASK_START_FILE, JSON.stringify({ timestamp: Date.now() }));
+  } catch {}
+}
+
+function shouldAnnounceStop(config) {
+  const minDuration = (config.min_task_duration_ms ?? 30000);
+  const start = getTaskStartTime();
+  if (!start) return true; // no start recorded, allow
+  const elapsed = Date.now() - start;
+  return elapsed >= minDuration;
+}
+
 /**
  * Auto-detect tmux pane context from the environment.
  * Returns an object with { pane_id, session_name, window_index } when
@@ -102,6 +130,19 @@ export async function processHookEvent(eventData) {
     return;
   }
 
+  // Track task start time on user prompt submit (always, regardless of category filter)
+  if (eventName === "UserPromptSubmit") {
+    setTaskStartTime();
+    debugLog("processHookEvent task start recorded", { source });
+  }
+
+  // Skip short events for all categories (always, regardless of category filter)
+  if (!shouldAnnounceStop(config)) {
+    const elapsed = Date.now() - getTaskStartTime();
+    debugLog("processHookEvent skip: short task", { source, eventName, elapsed_ms: elapsed });
+    return;
+  }
+
   // Check source-specific overrides (config.sources.<source>.enabled / .categories)
   const sourceOverride = (config.sources || {})[source] || {};
   if (sourceOverride.enabled === false) {