@tiflis-io/tiflis-code-workstation 0.3.9 → 0.3.10

package/dist/main.js

@@ -94,7 +94,7 @@ var EnvSchema = z.object({
   // ─────────────────────────────────────────────────────────────
   // Speech-to-Text (STT) Configuration
   // ─────────────────────────────────────────────────────────────
-  STT_PROVIDER: z.enum(["openai", "elevenlabs", "deepgram"]).default("openai"),
+  STT_PROVIDER: z.enum(["openai", "elevenlabs", "deepgram", "local"]).default("openai"),
   STT_API_KEY: z.string().optional(),
   STT_MODEL: z.string().default("whisper-1"),
   STT_BASE_URL: z.string().url().optional(),
@@ -102,7 +102,7 @@ var EnvSchema = z.object({
   // ─────────────────────────────────────────────────────────────
   // Text-to-Speech (TTS) Configuration
   // ─────────────────────────────────────────────────────────────
-  TTS_PROVIDER: z.enum(["openai", "elevenlabs"]).default("openai"),
+  TTS_PROVIDER: z.enum(["openai", "elevenlabs", "local"]).default("openai"),
   TTS_API_KEY: z.string().optional(),
   TTS_MODEL: z.string().default("tts-1"),
   TTS_VOICE: z.string().default("alloy"),
@@ -8379,7 +8379,18 @@ var STTService = class {
     this.logger.debug({ audioSize: audioBuffer.length, format }, "Transcribing audio");
     const startTime = Date.now();
     try {
-      const result = this.config.provider === "openai" ? await this.transcribeOpenAI(audioBuffer, format, signal) : await this.transcribeElevenLabs(audioBuffer, format, signal);
+      let result;
+      switch (this.config.provider) {
+        case "openai":
+        case "local":
+          result = await this.transcribeOpenAI(audioBuffer, format, signal);
+          break;
+        case "elevenlabs":
+          result = await this.transcribeElevenLabs(audioBuffer, format, signal);
+          break;
+        default:
+          throw new Error(`Unsupported STT provider: ${this.config.provider}`);
+      }
       const elapsed = Date.now() - startTime;
       this.logger.info(
         {
@@ -8491,6 +8502,9 @@ var STTService = class {
    * Check if the service is configured and ready.
    */
   isConfigured() {
+    if (this.config.provider === "local") {
+      return Boolean(this.config.baseUrl && this.config.model);
+    }
     return Boolean(this.config.apiKey && this.config.model);
   }
   /**
@@ -8507,20 +8521,28 @@ var STTService = class {
 function createSTTService(env, logger) {
   const provider = (env.STT_PROVIDER ?? "openai").toLowerCase();
   const apiKey = env.STT_API_KEY;
-  if (!apiKey) {
+  const baseUrl = env.STT_BASE_URL;
+  if (provider === "local") {
+    if (!baseUrl) {
+      logger.warn("STT_BASE_URL not configured for local provider, STT service disabled");
+      return null;
+    }
+  } else if (!apiKey) {
     logger.warn("STT_API_KEY not configured, STT service disabled");
     return null;
   }
   const defaults = {
     openai: { model: "whisper-1" },
-    elevenlabs: { model: "scribe_v1" }
+    elevenlabs: { model: "scribe_v1" },
+    deepgram: { model: "nova-2" },
+    local: { model: "large-v3" }
   };
   const config2 = {
     provider,
-    apiKey,
+    apiKey: apiKey ?? "",
     model: env.STT_MODEL ?? defaults[provider].model,
     language: env.STT_LANGUAGE,
-    baseUrl: env.STT_BASE_URL
+    baseUrl
   };
   return new STTService(config2, logger);
 }
@@ -8540,7 +8562,18 @@ var TTSService = class {
     this.logger.debug({ textLength: text2.length }, "Synthesizing speech");
     const startTime = Date.now();
     try {
-      const result = this.config.provider === "openai" ? await this.synthesizeOpenAI(text2) : await this.synthesizeElevenLabs(text2);
+      let result;
+      switch (this.config.provider) {
+        case "openai":
+        case "local":
+          result = await this.synthesizeOpenAI(text2);
+          break;
+        case "elevenlabs":
+          result = await this.synthesizeElevenLabs(text2);
+          break;
+        default:
+          throw new Error(`Unsupported TTS provider: ${this.config.provider}`);
+      }
       const elapsed = Date.now() - startTime;
       this.logger.info(
         { textLength: text2.length, audioSize: result.audio.length, elapsedMs: elapsed },
@@ -8617,6 +8650,9 @@ var TTSService = class {
    * Check if the service is configured and ready.
    */
   isConfigured() {
+    if (this.config.provider === "local") {
+      return Boolean(this.config.baseUrl && this.config.voice);
+    }
     return Boolean(this.config.apiKey && this.config.model && this.config.voice);
   }
   /**
@@ -8633,20 +8669,27 @@ var TTSService = class {
 function createTTSService(env, logger) {
   const provider = (env.TTS_PROVIDER ?? "openai").toLowerCase();
   const apiKey = env.TTS_API_KEY;
-  if (!apiKey) {
+  const baseUrl = env.TTS_BASE_URL;
+  if (provider === "local") {
+    if (!baseUrl) {
+      logger.warn("TTS_BASE_URL not configured for local provider, TTS service disabled");
+      return null;
+    }
+  } else if (!apiKey) {
     logger.warn("TTS_API_KEY not configured, TTS service disabled");
     return null;
   }
   const defaults = {
     openai: { model: "tts-1", voice: "alloy" },
-    elevenlabs: { model: "eleven_flash_v2_5", voice: "21m00Tcm4TlvDq8ikWAM" }
+    elevenlabs: { model: "eleven_flash_v2_5", voice: "21m00Tcm4TlvDq8ikWAM" },
+    local: { model: "kokoro", voice: "af_heart" }
   };
   const config2 = {
     provider,
-    apiKey,
+    apiKey: apiKey ?? "",
     model: env.TTS_MODEL ?? defaults[provider].model,
     voice: env.TTS_VOICE ?? defaults[provider].voice,
-    baseUrl: env.TTS_BASE_URL
+    baseUrl
   };
   return new TTSService(config2, logger);
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tiflis-io/tiflis-code-workstation",
-  "version": "0.3.9",
+  "version": "0.3.10",
   "description": "Workstation server for tiflis-code - manages agent sessions and terminal access",
   "author": "Roman Barinov <rbarinov@gmail.com>",
   "license": "FSL-1.1-NC",