tesseract-agent 0.9.6 → 0.9.7

package/dist-bundle/cli.js

@@ -22752,13 +22752,13 @@ Add it to an EDL as an audio track (duckUnderSpeech:true under speech).`);
       return text6(`Speech generated: ${res.file.path}
 Overlay it in render_edl as an audio track (mute the original speaker's segments to replace their voice).`);
     }),
-    tool7("list_avatars", "List the HeyGen avatars available on the user's account (for virtual-presenter videos), with their avatar ids. Key-gated: needs a HeyGen API key. Pair an avatar_id with a HeyGen voice for generate_avatar_video.", {}, async () => {
+    tool7("list_avatars", "List the HeyGen avatars available for virtual-presenter videos, with their avatar ids. Works on the free bundled tier when signed in, or on the user's own HeyGen key if they set one. Pair an avatar_id with a HeyGen voice for generate_avatar_video.", {}, async () => {
       const res = await deps.listAvatars();
       if (!res.ok)
         return text6(`Could not list avatars: ${res.error ?? "unknown error"}`, true);
       return jsonText(res.avatars ?? []);
     }),
-    tool7("generate_avatar_video", "Generate a talking-presenter (avatar) video from a script (HeyGen \u2014 paid, needs a HeyGen key). Draft the script yourself; pick an avatar via list_avatars and a HeyGen voice_id. Takes 1\u20135 minutes; the finished video is displayed in the chat. Use for news-desk reads, explainers, spokesperson clips. ONLY use avatars/likenesses you are authorized to use.", {
+    tool7("generate_avatar_video", "Generate a talking-presenter (avatar) video from a script (HeyGen). Runs on the free bundled tier when signed in, or on the user's own HeyGen key if they set one \u2014 they never need to be asked for a key up front. Draft the script yourself; pick an avatar via list_avatars and a HeyGen voice_id. Takes 1\u20135 minutes; the finished video is displayed in the chat. Use for news-desk reads, explainers, spokesperson clips. ONLY use avatars/likenesses you are authorized to use.", {
       script: external_exports.string().describe("What the presenter says."),
       avatar_id: external_exports.string().optional().describe("HeyGen avatar id (from list_avatars). Provide this OR talking_photo_id."),
       talking_photo_id: external_exports.string().optional().describe("HeyGen talking-photo id (alternative to avatar_id)."),
@@ -24266,6 +24266,13 @@ ${JSON.stringify(r, null, 2)}`
       // ── image (image-generation/edit providers) ──
       { id: "image.list_providers", domain: "image", title: "List image providers", description: "Image providers with configured/default/model status.", input: external_exports.object({}), mutation: "read", availability: "host", binding: { kind: "ipc", channel: "imageEdit", verb: "list" } },
       { id: "image.set_provider_key", domain: "image", title: "Set image provider key", description: "Set (blank clears) an image provider's API key.", input: external_exports.object({ provider_id: IMAGE_PROVIDER, api_key: external_exports.string().describe("API key (secret); blank clears.") }), mutation: "write", availability: "host", binding: { kind: "ipc", channel: "imageEdit", verb: "setKey" } },
+      // Avatar videos are the one media provider whose key lives outside the image pool
+      // (HeyGen, its own vendor) — but it is set in the SAME Settings panel, so it belongs
+      // to the same domain rather than earning a domain of its own. `has_` returns a
+      // boolean, never the key: writing a secret in is a capability, reading one out is
+      // not (the same line `auth:token` sits on the wrong side of).
+      { id: "image.set_avatar_key", domain: "image", title: "Set avatar-video (HeyGen) key", description: "Set (blank clears) the HeyGen API key used for avatar/presenter videos. Without one, avatar videos fall back to the free Tesseract proxy tier for signed-in users.", input: external_exports.object({ api_key: external_exports.string().describe("HeyGen API key (secret); blank clears.") }), mutation: "write", availability: "host", binding: { kind: "ipc", channel: "heygen", verb: "setKey" } },
+      { id: "image.has_avatar_key", domain: "image", title: "Avatar-video key configured?", description: "Whether a personal HeyGen key is set. Never returns the key itself.", input: external_exports.object({}), mutation: "read", availability: "host", binding: { kind: "ipc", channel: "heygen", verb: "hasKey" } },
       { id: "image.remove_provider_key", domain: "image", title: "Remove image provider key", description: "Remove all credentials for an image provider.", input: external_exports.object({ provider_id: IMAGE_PROVIDER }), mutation: "write", availability: "host", binding: { kind: "ipc", channel: "imageEdit", verb: "removeKey" } },
       { id: "image.set_default_provider", domain: "image", title: "Set default image provider", description: "Set the default image provider.", input: external_exports.object({ provider_id: IMAGE_PROVIDER }), mutation: "write", availability: "host", binding: { kind: "ipc", channel: "imageEdit", verb: "setDefault" } },
       { id: "image.set_provider_model", domain: "image", title: "Set image provider model", description: "Set the model for an image provider.", input: external_exports.object({ provider_id: IMAGE_PROVIDER, model: external_exports.string() }), mutation: "write", availability: "host", binding: { kind: "ipc", channel: "imageEdit", verb: "setModel" } },
@@ -66328,16 +66335,17 @@ async function listVoicesElevenLabs(cfg2, opts) {
 }
 
 // ../../packages/video-core/dist/heygen.js
-var BASE_URL2 = "https://api.heygen.com";
+var HEYGEN_DIRECT_BASE_URL = "https://api.heygen.com";
 var DEFAULT_DIM = { width: 1280, height: 720 };
 var HG_REQUEST_TIMEOUT_MS = 6e4;
 var HG_DOWNLOAD_TIMEOUT_MS = 18e4;
-async function hgFetch(path6, apiKey, init, signal) {
+async function hgFetch(path6, cfg2, init, signal) {
+  const base = (cfg2.baseUrl || HEYGEN_DIRECT_BASE_URL).replace(/\/+$/, "");
   let res;
   try {
-    res = await fetch(`${BASE_URL2}${path6}`, {
+    res = await fetch(`${base}${path6}`, {
       method: init.method,
-      headers: { "X-Api-Key": apiKey, ...init.body != null ? { "Content-Type": "application/json" } : {} },
+      headers: { "X-Api-Key": cfg2.apiKey, ...init.body != null ? { "Content-Type": "application/json" } : {} },
       ...init.body != null ? { body: JSON.stringify(init.body) } : {},
       signal: signal ?? AbortSignal.timeout(HG_REQUEST_TIMEOUT_MS)
     });
@@ -66354,7 +66362,7 @@ async function hgFetch(path6, apiKey, init, signal) {
   return json2;
 }
 async function listAvatarsHeyGen(cfg2, signal) {
-  const j = await hgFetch("/v2/avatars", cfg2.apiKey, { method: "GET" }, signal);
+  const j = await hgFetch("/v2/avatars", cfg2, { method: "GET" }, signal);
   return (j.data?.avatars ?? []).map((a) => ({ avatarId: a.avatar_id, name: a.avatar_name ?? a.avatar_id }));
 }
 async function submitAvatarVideoHeyGen(spec, cfg2, signal) {
@@ -66371,14 +66379,14 @@ async function submitAvatarVideoHeyGen(spec, cfg2, signal) {
     ],
     dimension: { width: spec.width ?? DEFAULT_DIM.width, height: spec.height ?? DEFAULT_DIM.height }
   };
-  const j = await hgFetch("/v2/video/generate", cfg2.apiKey, { method: "POST", body }, signal);
+  const j = await hgFetch("/v2/video/generate", cfg2, { method: "POST", body }, signal);
   const id = j.data?.video_id;
   if (!id)
     throw new Error("HeyGen returned no video_id.");
   return id;
 }
 async function pollAvatarVideoHeyGen(videoId, cfg2, signal) {
-  const j = await hgFetch(`/v1/video_status.get?video_id=${encodeURIComponent(videoId)}`, cfg2.apiKey, { method: "GET" }, signal);
+  const j = await hgFetch(`/v1/video_status.get?video_id=${encodeURIComponent(videoId)}`, cfg2, { method: "GET" }, signal);
   const d = j.data ?? {};
   const error51 = typeof d.error === "string" ? d.error : d.error?.message;
   return { status: d.status ?? "processing", videoUrl: d.video_url, error: error51 };
@@ -67662,13 +67670,13 @@ ${inputs.prompt}` : "");
 
 // ../../packages/host-core/dist/video-transcribe.js
 var DEFAULT_MODEL5 = "gemini-3.5-flash";
-var BASE_URL3 = "https://generativelanguage.googleapis.com/v1beta";
+var BASE_URL2 = "https://generativelanguage.googleapis.com/v1beta";
 var CHUNK_PROMPT = `Transcribe this audio VERBATIM with speaker diarization.
 Return ONLY a JSON array; each element: {"start": <seconds from the START OF THIS AUDIO, number>, "end": <seconds, number>, "speaker": "<S1|S2|S3\u2026, stable within this audio>", "text": "<utterance>"}.
 Rules: keep the original language(s); split on speaker changes and natural sentence boundaries (segments \u2264 15s); timecodes must be accurate to the second; no commentary, no markdown.`;
 var REQUEST_TIMEOUT_MS3 = 18e4;
 async function geminiJson(opts) {
-  const res = await fetch(`${BASE_URL3}/models/${encodeURIComponent(opts.model)}:generateContent`, {
+  const res = await fetch(`${BASE_URL2}/models/${encodeURIComponent(opts.model)}:generateContent`, {
     method: "POST",
     headers: { "Content-Type": "application/json", "x-goog-api-key": opts.apiKey },
     body: JSON.stringify({
@@ -67774,14 +67782,14 @@ ${compact2}`;
 
 // ../../packages/host-core/dist/video-vision.js
 import { readFileSync as readFileSync19 } from "node:fs";
-var BASE_URL4 = "https://generativelanguage.googleapis.com/v1beta";
+var BASE_URL3 = "https://generativelanguage.googleapis.com/v1beta";
 var MODEL = "gemini-3.5-flash";
 var TIMEOUT_MS = 6e4;
 async function reviewFrameGemini(framePath, intent, apiKey) {
   try {
     const b642 = readFileSync19(framePath).toString("base64");
     const prompt = `This is a frame from an AI-generated video clip meant to show: "${intent}". Judge ONLY for obvious defects: melted/warped faces or hands, garbled text, broken anatomy, severe artifacts. Reply with STRICT JSON: {"acceptable": <true|false>, "note": "<\u226412 words, only if not acceptable>"}.`;
-    const res = await fetch(`${BASE_URL4}/models/${MODEL}:generateContent`, {
+    const res = await fetch(`${BASE_URL3}/models/${MODEL}:generateContent`, {
       method: "POST",
       headers: { "Content-Type": "application/json", "x-goog-api-key": apiKey },
       body: JSON.stringify({
@@ -67809,9 +67817,19 @@ function googleKey() {
 function elevenLabsKey() {
   return readDaemonAgentSettings().voiceKeys?.elevenlabs?.trim() || null;
 }
-function heygenKey() {
-  return readDaemonAgentSettings().heygenApiKey?.trim() || readDaemonKeyFile("heygen.key") || process.env.HEYGEN_API_KEY?.trim() || null;
+function resolveAvatarCreds(snap, keyFile, envKey2) {
+  const fromSnapshot = (snap.heygen?.baseUrl ? void 0 : snap.heygen?.apiKey?.trim()) || snap.heygenApiKey?.trim();
+  const personal = fromSnapshot || keyFile?.trim() || envKey2?.trim();
+  if (personal)
+    return { apiKey: personal };
+  if (snap.heygen?.baseUrl && snap.heygen.apiKey)
+    return { apiKey: snap.heygen.apiKey, baseUrl: snap.heygen.baseUrl };
+  return null;
+}
+function heygenCreds() {
+  return resolveAvatarCreds(readDaemonAgentSettings(), readDaemonKeyFile("heygen.key"), process.env.HEYGEN_API_KEY ?? null);
 }
+var NO_AVATAR_CREDS = "Avatar videos need either a HeyGen API key (add it in Settings, or set HEYGEN_API_KEY) or a signed-in Tesseract account for the free tier.";
 function readDaemonKeyFile(name) {
   try {
     const home = tesseractHome();
@@ -67984,21 +68002,21 @@ function createDaemonVideoEditRunner() {
       }
     },
     async listAvatars() {
-      const apiKey = heygenKey();
-      if (!apiKey)
-        return { ok: false, error: "Avatar videos need a HeyGen API key (set HEYGEN_API_KEY or add it in settings)." };
+      const creds3 = heygenCreds();
+      if (!creds3)
+        return { ok: false, error: NO_AVATAR_CREDS };
       try {
-        return { ok: true, avatars: await listAvatarsHeyGen({ apiKey }) };
+        return { ok: true, avatars: await listAvatarsHeyGen(creds3) };
       } catch (e) {
         return { ok: false, error: err(e) };
       }
     },
     async generateAvatarVideo(req, outDir, onProgress) {
-      const apiKey = heygenKey();
-      if (!apiKey)
-        return { ok: false, error: "Avatar videos need a HeyGen API key (set HEYGEN_API_KEY or add it in settings)." };
+      const creds3 = heygenCreds();
+      if (!creds3)
+        return { ok: false, error: NO_AVATAR_CREDS };
       try {
-        const videoId = await submitAvatarVideoHeyGen({ avatarId: req.avatarId, talkingPhotoId: req.talkingPhotoId, text: req.text, voiceId: req.voiceId, width: req.width, height: req.height, background: req.background }, { apiKey });
+        const videoId = await submitAvatarVideoHeyGen({ avatarId: req.avatarId, talkingPhotoId: req.talkingPhotoId, text: req.text, voiceId: req.voiceId, width: req.width, height: req.height, background: req.background }, creds3);
         onProgress?.("Avatar video submitted \u2014 rendering\u2026");
         const started = Date.now();
         for (; ; ) {
@@ -68007,7 +68025,7 @@ function createDaemonVideoEditRunner() {
           await new Promise((r) => setTimeout(r, HEYGEN_POLL_MS));
           let poll;
           try {
-            poll = await pollAvatarVideoHeyGen(videoId, { apiKey });
+            poll = await pollAvatarVideoHeyGen(videoId, creds3);
           } catch {
             onProgress?.(`Rendering avatar video\u2026 (${Math.round((Date.now() - started) / 1e3)}s)`);
             continue;
@@ -81930,25 +81948,38 @@ async function runDaemon(opts = {}) {
   }
   async function takeoverMachine() {
     const rivals = /* @__PURE__ */ new Set();
+    let lockPid = 0, statusPid = 0;
+    const otherPids = [];
     try {
-      const lp = Number(readFileSync30(pidfile(), "utf-8").trim());
-      if (lp > 0)
-        rivals.add(lp);
+      lockPid = Number(readFileSync30(pidfile(), "utf-8").trim());
+      if (lockPid > 0)
+        rivals.add(lockPid);
     } catch {
     }
     try {
-      const sp = Number(JSON.parse(readFileSync30(statusFile(), "utf-8")).pid ?? 0);
-      if (sp > 0)
-        rivals.add(sp);
+      statusPid = Number(JSON.parse(readFileSync30(statusFile(), "utf-8")).pid ?? 0);
+      if (statusPid > 0)
+        rivals.add(statusPid);
     } catch {
     }
-    for (const p of otherDaemonPids())
+    for (const p of otherDaemonPids()) {
+      otherPids.push(p);
       rivals.add(p);
+    }
+    try {
+      log9(`takeover: me=${process.pid} lockPid=${lockPid || "-"} statusPid=${statusPid || "-"} otherDaemonPids=[${otherPids.join(",") || "-"}]`);
+    } catch {
+    }
     rivals.delete(process.pid);
     const live = [...rivals].filter(isAlive2);
     if (live.length) {
       for (const pid of live) {
-        log9(`takeover: killing rival daemon pid=${pid} (latest-wins single-instance)`);
+        let cmd = "";
+        try {
+          cmd = execFileSync4("ps", ["-o", "command=", "-p", String(pid)], { encoding: "utf8", timeout: 2e3 }).trim().slice(0, 200);
+        } catch {
+        }
+        log9(`takeover: killing rival daemon pid=${pid} (latest-wins single-instance) cmd=[${cmd}]`);
         try {
           process.kill(pid, "SIGTERM");
         } catch {
@@ -82007,9 +82038,30 @@ async function runDaemon(opts = {}) {
   const creds3 = readCredentials();
   const deviceId2 = creds3?.deviceId;
   const daemonStartedAt = Date.now();
+  try {
+    const pp = process;
+    log9(`LAUNCH-CTX pid=${pp.pid} ppid=${pp.ppid} exec=${pp.execPath}`);
+    log9(`LAUNCH-CTX argv=${JSON.stringify(pp.argv)}`);
+    log9(`LAUNCH-CTX env: GUI=${process.env.CCB_DAEMON_GUI ?? "-"} FORCE_GUI=${process.env.CCB_FORCE_DAEMON_GUI ?? "-"} RUN_AS_NODE=${process.env.ELECTRON_RUN_AS_NODE ?? "-"} UNMANAGED=${process.env.CCB_DAEMON_UNMANAGED ?? "-"} ACTIVE_LINK=${process.env.CCB_ACTIVE_LINK ?? "-"} RDESK=${process.env.CCB_RDESK_ROLE ?? "-"}`);
+  } catch {
+  }
+  const snapshotProcs = (tag) => {
+    try {
+      const raw = execFileSync4("ps", ["-axww", "-o", "pid=,ppid=,pcpu=,etime=,command="], { encoding: "utf8", timeout: 3e3 });
+      const rows = raw.split("\n").filter((l) => /[Tt]esseract|host-daemon/.test(l) && !/\bgrep\b|ps -axww/.test(l));
+      log9(`PROC-SNAPSHOT [${tag}] me=${process.pid} ppid=${process.ppid} \u2014 ${rows.length} tesseract proc(s):`);
+      for (const r of rows.slice(0, 40))
+        log9(`  PROC ${r.trim().slice(0, 240)}`);
+    } catch (e) {
+      try {
+        log9(`PROC-SNAPSHOT [${tag}] ps failed: ${e.message}`);
+      } catch {
+      }
+    }
+  };
   process.on("exit", (code) => {
     try {
-      log9(`PROCESS EXIT code=${code} uptimeMs=${Date.now() - daemonStartedAt}`);
+      log9(`PROCESS EXIT code=${code} uptimeMs=${Date.now() - daemonStartedAt} ppid=${process.ppid}`);
     } catch {
     }
   });
@@ -82017,6 +82069,8 @@ async function runDaemon(opts = {}) {
     try {
       log9(`SHUTDOWN reason=${reason} code=${code} uptimeMs=${Date.now() - daemonStartedAt} ppid=${process.ppid} pid=${process.pid}`);
       log9(`SHUTDOWN caller: ${(new Error("shutdown").stack ?? "").split("\n").slice(2, 7).map((l) => l.trim()).join("  <-  ")}`);
+      if (reason.startsWith("signal"))
+        snapshotProcs(reason);
     } catch {
     }
     if (timer) {
@@ -82047,6 +82101,16 @@ async function runDaemon(opts = {}) {
   }
   process.on("SIGTERM", () => shutdown(0, "signal:SIGTERM"));
   process.on("SIGINT", () => shutdown(0, "signal:SIGINT"));
+  try {
+    process.on("SIGHUP", () => {
+      try {
+        log9("SIGHUP received \u2014 IGNORED (kept alive)");
+        snapshotProcs("SIGHUP");
+      } catch {
+      }
+    });
+  } catch {
+  }
   process.on("uncaughtException", (e) => {
     try {
       log9(`UNCAUGHT (kept alive): ${e instanceof Error ? e.stack ?? e.message : String(e)}`);
@@ -82067,6 +82131,7 @@ async function runDaemon(opts = {}) {
       eapp.on("before-quit", (e) => {
         try {
           log9("APP before-quit \u2014 PREVENTED (daemon never self-quits via Electron)");
+          snapshotProcs("before-quit");
         } catch {
         }
         try {
@@ -82074,15 +82139,41 @@ async function runDaemon(opts = {}) {
         } catch {
         }
       });
+      eapp.on("will-quit", (e) => {
+        try {
+          log9(`APP will-quit \u2014 PREVENTED (uptime ${Date.now() - daemonStartedAt}ms)`);
+          e?.preventDefault?.();
+        } catch {
+        }
+      });
+      eapp.on("quit", (_e, code) => {
+        try {
+          log9(`APP quit event code=${JSON.stringify(code)} (uptime ${Date.now() - daemonStartedAt}ms)`);
+        } catch {
+        }
+      });
+      eapp.on("window-all-closed", () => {
+        try {
+          log9("APP window-all-closed (kept alive \u2014 no auto-quit)");
+        } catch {
+        }
+      });
       eapp.on("render-process-gone", (_e, d) => {
         try {
-          log9(`render-process-gone (kept alive): ${JSON.stringify(d)}`);
+          log9(`render-process-gone (kept alive) uptime=${Date.now() - daemonStartedAt}ms: ${JSON.stringify(d)}`);
+          snapshotProcs("render-gone");
         } catch {
         }
       });
       eapp.on("child-process-gone", (_e, d) => {
         try {
-          log9(`child-process-gone (kept alive): ${JSON.stringify(d)}`);
+          log9(`child-process-gone (kept alive) uptime=${Date.now() - daemonStartedAt}ms: ${JSON.stringify(d)}`);
+        } catch {
+        }
+      });
+      eapp.on("gpu-process-crashed", (_e, killed) => {
+        try {
+          log9(`gpu-process-crashed killed=${JSON.stringify(killed)} uptime=${Date.now() - daemonStartedAt}ms`);
         } catch {
         }
       });

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tesseract-agent",
-  "version": "0.9.6",
+  "version": "0.9.7",
   "license": "UNLICENSED",
   "publishConfig": {
     "access": "public",