feat: use turbo model and return all probs

language_detector.py

@@ -27,6 +27,7 @@ class LanguageDetector:
         Returns:
             str: Detected language code (e.g., "en", "fr", etc.)
             float: Confidence score
+            dict: All language probabilities
         """
         try:
             # Load and preprocess audio
@@ -41,7 +42,7 @@ class LanguageDetector:
             detected_lang = max(probs, key=probs.get)
             confidence = probs[detected_lang]
             
-            return detected_lang, confidence
+            return detected_lang, confidence, probs
             
         except Exception as e:
             logger.error(f"Error in language detection: {e}")
@@ -57,6 +58,7 @@ class LanguageDetector:
         Returns:
             str: Detected language code (e.g., "en", "fr", etc.)
             float: Confidence score
+            dict: All language probabilities
         """
         try:
             # Convert bytes to numpy array using librosa
@@ -77,7 +79,7 @@ class LanguageDetector:
             detected_lang = max(probs, key=probs.get)
             confidence = probs[detected_lang]
             
-            return detected_lang, confidence
+            return detected_lang, confidence, probs
             
         except Exception as e:
             logger.error(f"Error in language detection: {e}")

main.py

@@ -31,7 +31,7 @@ language_detector = None
 async def lifespan(app: FastAPI):
     global kit, language_detector
     kit = WhisperLiveKit()
-    language_detector = LanguageDetector(model_name="medium")
+    language_detector = LanguageDetector(model_name="turbo")
     yield
 
 app = FastAPI(lifespan=lifespan)
@@ -68,14 +68,15 @@ async def detect_language(file: UploadFile = File(...)):
         
         # Use the language detector with the saved file
         if language_detector:
-            detected_lang, confidence = language_detector.detect_language_from_file(file_path)
+            detected_lang, confidence, probs = language_detector.detect_language_from_file(file_path)
             
             # Clean up - remove the temporary file
             os.remove(file_path)
             
             return JSONResponse({
                 "language": detected_lang,
-                "confidence": float(confidence)
+                "confidence": float(confidence),
+                "probabilities": {lang: float(prob) for lang, prob in probs.items()}
             })
         else:
             return JSONResponse(