negotium 0.5.0 → 0.5.2

package/dist/main.js

@@ -270,10 +270,13 @@ function resolveModelEnv(envKey, agentConst) {
 function resolveDefaultModel(agent, registryDefaultModel) {
   return agent === SESSION_AGENT && SESSION_MODEL ? SESSION_MODEL : registryDefaultModel;
 }
+function resolveWhisperBackend() {
+  return envText("WHISPER_BACKEND") === "mlx-whisper" ? "mlx-whisper" : "faster-whisper";
+}
 function codexAuthFilePath() {
   return process.env.NEGOTIUM_CODEX_AUTH_FILE || join(process.env.CODEX_HOME || join(homedir(), ".codex"), "auth.json");
 }
-var HOME, PROJECT_ROOT, STATE_DIR_ENV, STATE_DIR, WORKSPACE_DIR, TOPIC_WORKSPACE_DIR, SHARED_WIKI_DIR, CRON_WORKSPACE_DIR, BROWSER_DIR, BROWSER_PROFILES_DIR, BINARIES_DIR, SECRETS_DIR, DM_WORKSPACE_DIR, SESSION_WORKSPACE_DIR, CLAUDE_EXECUTABLE_ENV, CLAUDE_EXECUTABLE, DEFAULT_OUTPUT_LANGUAGE = "English", BROWSER_RS_VERSION = "v0.1.23", BROWSER_RS_MIN_SECURE_VERSION = "0.1.15", BROWSER_RS_VERSION_PROBE_TIMEOUT_MS = 15000, SNIPPETS_API_URL, BROWSER_RS_BIN, BASH_RS_VERSION = "v0.1.7", BASH_RS_BIN, PLAYWRIGHT_MCP_BIN, TSX_BIN, TSX_LOADER, TSCONFIG_PATH, SESSION_COMM_SERVER, TASK_SERVER, DECISION_SERVER, CANONICAL_MCP_PROXY_SERVER, WIKI_SERVER, TOKEN_STATS_SERVER, COMPACTION_LOG_SERVER, SYSTEM_HEALTH_SERVER, AGENT_HEALTH_SERVER, VAULT_SERVER, BG_BASH_BASE_PORT, BG_BASH_MAX_PORT, RUNTIME_MCP_SECRET, NODE_CONTROL_TOKEN, NODE_ID, VAULT_MASTER_KEY, NEGOTIUM_PORT, hostname, DATA_DIR, LOG_DIR, UPLOADS_DIR, VAULT_DIR, SESSIONS_DB, DEBUG_FILE, USERS_LOG_DIR, RUN_DIR, PROGRESS_DIR, DM_CMD_DIR, DM_RESP_DIR, SESSION_INBOX_DIR, SESSION_ASKS_DIR, BASHRS_SPILL_ROOT, PLAYWRIGHT_BASE_PORT, PLAYWRIGHT_MAX_PORT, PLAYWRIGHT_PORTS_DIR, ACTIVE_QUERY_STALE_MS, AGENTS_PROMPTS_DIR, RESOURCES_DIR, FILE_TAG_REGEX, MODEL_SONNET = "claude-sonnet-5", MODEL_OPUS = "claude-opus-5", MODEL_FABLE = "claude-fable-5", FALLBACK_AGENT, SESSION_AGENT, GATEWAY_AGENT, FALLBACK_MODEL, SESSION_MODEL, GATEWAY_MODEL, FFMPEG_BIN, FFPROBE_BIN, PYTHON_BIN, FASTER_WHISPER_WRAPPER, WHISPER_MODEL, TESSERACT_BIN, PDFTOTEXT_BIN, _envMaxTellDepth, MAX_TELL_DEPTH;
+var HOME, PROJECT_ROOT, STATE_DIR_ENV, STATE_DIR, WORKSPACE_DIR, TOPIC_WORKSPACE_DIR, SHARED_WIKI_DIR, CRON_WORKSPACE_DIR, BROWSER_DIR, BROWSER_PROFILES_DIR, BINARIES_DIR, SECRETS_DIR, DM_WORKSPACE_DIR, SESSION_WORKSPACE_DIR, CLAUDE_EXECUTABLE_ENV, CLAUDE_EXECUTABLE, DEFAULT_OUTPUT_LANGUAGE = "English", BROWSER_RS_VERSION = "v0.1.23", BROWSER_RS_MIN_SECURE_VERSION = "0.1.15", BROWSER_RS_VERSION_PROBE_TIMEOUT_MS = 15000, SNIPPETS_API_URL, BROWSER_RS_BIN, BASH_RS_VERSION = "v0.1.7", BASH_RS_BIN, PLAYWRIGHT_MCP_BIN, TSX_BIN, TSX_LOADER, TSCONFIG_PATH, SESSION_COMM_SERVER, TASK_SERVER, DECISION_SERVER, CANONICAL_MCP_PROXY_SERVER, WIKI_SERVER, TOKEN_STATS_SERVER, COMPACTION_LOG_SERVER, SYSTEM_HEALTH_SERVER, AGENT_HEALTH_SERVER, VAULT_SERVER, BG_BASH_BASE_PORT, BG_BASH_MAX_PORT, RUNTIME_MCP_SECRET, NODE_CONTROL_TOKEN, NODE_ID, VAULT_MASTER_KEY, NEGOTIUM_PORT, hostname, DATA_DIR, LOG_DIR, UPLOADS_DIR, VAULT_DIR, SESSIONS_DB, DEBUG_FILE, USERS_LOG_DIR, RUN_DIR, PROGRESS_DIR, DM_CMD_DIR, DM_RESP_DIR, SESSION_INBOX_DIR, SESSION_ASKS_DIR, BASHRS_SPILL_ROOT, PLAYWRIGHT_BASE_PORT, PLAYWRIGHT_MAX_PORT, PLAYWRIGHT_PORTS_DIR, ACTIVE_QUERY_STALE_MS, AGENTS_PROMPTS_DIR, RESOURCES_DIR, FILE_TAG_REGEX, MODEL_SONNET = "claude-sonnet-5", MODEL_OPUS = "claude-opus-5", MODEL_FABLE = "claude-fable-5", FALLBACK_AGENT, SESSION_AGENT, GATEWAY_AGENT, FALLBACK_MODEL, SESSION_MODEL, GATEWAY_MODEL, FFMPEG_BIN, FFPROBE_BIN, PYTHON_BIN, FASTER_WHISPER_WRAPPER, WHISPER_MODEL, TESSERACT_BIN, PDFTOTEXT_BIN, WHISPER_BACKEND, MLX_WHISPER_BIN, WHISPER_MODEL_MLX, _envMaxTellDepth, MAX_TELL_DEPTH;
 var init_config = __esm(() => {
   init_config_helpers();
   init_logger();
@@ -375,6 +378,9 @@ var init_config = __esm(() => {
   WHISPER_MODEL = envText("WHISPER_MODEL_FILE") ?? "turbo";
   TESSERACT_BIN = envText("TESSERACT_BIN") ?? "tesseract";
   PDFTOTEXT_BIN = envText("PDFTOTEXT_BIN") ?? "pdftotext";
+  WHISPER_BACKEND = resolveWhisperBackend();
+  MLX_WHISPER_BIN = envText("MLX_WHISPER_BIN");
+  WHISPER_MODEL_MLX = envText("WHISPER_MODEL_MLX") ?? "mlx-community/whisper-large-v3-turbo";
   _envMaxTellDepth = Number.parseInt(process.env.MAX_TELL_DEPTH ?? "", 10);
   MAX_TELL_DEPTH = Number.isInteger(_envMaxTellDepth) && _envMaxTellDepth > 0 ? _envMaxTellDepth : 20;
 });
@@ -1868,7 +1874,7 @@ var exports_version = {};
 __export(exports_version, {
   NEGOTIUM_VERSION: () => NEGOTIUM_VERSION
 });
-var NEGOTIUM_VERSION = "0.5.0";
+var NEGOTIUM_VERSION = "0.5.2";
 
 // ../../packages/core/src/agents/codex-native-multi-agent.ts
 import { spawn } from "child_process";
@@ -21383,8 +21389,14 @@ async function extractFromImage(filePath) {
 }
 function isTranscriptionConfigured(opts = {}) {
   const ffmpeg = opts.ffmpegBin ?? FFMPEG_BIN;
+  if (!ffmpeg)
+    return false;
+  const backend = opts.backend ?? WHISPER_BACKEND;
+  if (backend === "mlx-whisper") {
+    return Boolean(opts.mlxWhisperBin ?? MLX_WHISPER_BIN);
+  }
   const wrapper = opts.wrapperPath ?? FASTER_WHISPER_WRAPPER;
-  return Boolean(ffmpeg) && existsSync21(wrapper);
+  return existsSync21(wrapper);
 }
 async function transcribeAudio(filePath, opts = {}) {
   if (!isTranscriptionConfigured(opts))
@@ -21394,6 +21406,7 @@ async function transcribeAudio(filePath, opts = {}) {
 }
 async function extractFromAudio(filePath, opts = {}) {
   const tmpDir = `${filePath}_whisper_tmp`;
+  const backend = opts.backend ?? WHISPER_BACKEND;
   try {
     mkdirSync21(tmpDir, { recursive: true });
     const mp3Path = join31(tmpDir, "audio.mp3");
@@ -21411,18 +21424,26 @@ async function extractFromAudio(filePath, opts = {}) {
       "16000",
       mp3Path
     ], { timeout: 60000 });
-    await execFileAsync(opts.pythonBin ?? PYTHON_BIN, [
-      opts.wrapperPath ?? FASTER_WHISPER_WRAPPER,
+    const whisperArgs = [
       mp3Path,
       "--model",
-      opts.model ?? WHISPER_MODEL,
+      opts.model ?? (backend === "mlx-whisper" ? WHISPER_MODEL_MLX : WHISPER_MODEL),
       "--language",
       opts.language ?? "ko",
       "--output-dir",
       tmpDir,
       "--output-format",
       "txt"
-    ], { timeout: 120000 });
+    ];
+    if (backend === "mlx-whisper") {
+      const mlxWhisperBin = opts.mlxWhisperBin ?? MLX_WHISPER_BIN;
+      if (!mlxWhisperBin) {
+        return { text: null, method: "whisper", error: "MLX_WHISPER_BIN not configured" };
+      }
+      await execFileAsync(mlxWhisperBin, whisperArgs, { timeout: 300000 });
+    } else {
+      await execFileAsync(opts.pythonBin ?? PYTHON_BIN, [opts.wrapperPath ?? FASTER_WHISPER_WRAPPER, ...whisperArgs], { timeout: 120000 });
+    }
     const txtPath = join31(tmpDir, "audio.txt");
     const text2 = existsSync21(txtPath) ? readFileSync17(txtPath, "utf-8").trim() : null;
     if (!text2) {
@@ -21430,7 +21451,7 @@ async function extractFromAudio(filePath, opts = {}) {
     }
     return { text: text2, method: "whisper" };
   } catch (e) {
-    logger.debug({ err: e, filePath }, "whisper extraction failed");
+    logger.debug({ err: e, filePath, backend }, "whisper extraction failed");
     return {
       text: null,
       method: "whisper",
@@ -47639,4 +47660,4 @@ switch (command) {
   }
 }
 
-//# debugId=A120DD1C769C1B1164756E2164756E21
+//# debugId=A98143381B45D73164756E2164756E21