Update asr.py

asr.py

@@ -1,58 +1,59 @@
 import librosa
 import torch
 import numpy as np
-import langid
+import langid  # Language detection library
 from transformers import Wav2Vec2ForCTC, AutoProcessor
 
 ASR_SAMPLING_RATE = 16_000
-MODEL_ID = "facebook/mms-1b-all"  # Or your model ID
+MODEL_ID = "facebook/mms-1b-all"
 
-# Load MMS Model (outside the function, for efficiency)
-try:
-    processor = AutoProcessor.from_pretrained(MODEL_ID)
-    model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
-    model.eval()
-except Exception as e:
-    print(f"Error loading initial model: {e}")  # Handle initial model loading errors
-    exit(1)  # Or raise the exception if you prefer
+# Load MMS Model
+processor = AutoProcessor.from_pretrained(MODEL_ID)
+model = Wav2Vec2ForCTC.from_pretrained(MODEL_ID)
+model.eval()
 
 def detect_language(text):
+    """Detects language using langid (fast & lightweight)."""
     lang, _ = langid.classify(text)
-    return lang if lang in ["en", "sw"] else "en"
+    return lang if lang in ["en", "sw"] else "en"  # Default to English
 
 def transcribe_auto(audio_data=None):
     if not audio_data:
         return "<<ERROR: Empty Audio Input>>"
-
-    # ... (audio processing code remains the same) ...
-
-    try:  # Wrap the entire transcription process
-        # **Step 1: Transcribe without Language Detection**
-        with torch.no_grad():
-            outputs = model(**inputs).logits
-            ids = torch.argmax(outputs, dim=-1)[0]
-            raw_transcription = processor.decode(ids)
-
-        # **Step 2: Detect Language from Transcription**
-        detected_lang = detect_language(raw_transcription)
-        lang_code = "eng" if detected_lang == "en" else "swh"
-
-        # **Step 3: Reload Model with Correct Adapter (CRITICAL CHANGE)**
-        try:  # Wrap adapter loading
-            processor.tokenizer.set_target_lang(lang_code)
-            model.load_adapter(lang_code)  # This is the most likely source of errors
-        except Exception as adapter_error:  # Catch adapter loading errors
-            print(f"Error loading adapter for {detected_lang}: {adapter_error}")
-            return f"<<ERROR: Could not load adapter for {detected_lang}>>"  # Or raise
-
-        # **Step 4: Transcribe Again with Correct Adapter**
-        with torch.no_grad():
-            outputs = model(**inputs).logits
-            ids = torch.argmax(outputs, dim=-1)[0]
-            final_transcription = processor.decode(ids)
-
-        return f"Detected Language: {detected_lang.upper()}\n\nTranscription:\n{final_transcription}"
-
-    except Exception as overall_error:  # Catch any other errors during transcription
-        print(f"An error occurred during transcription: {overall_error}")
-        return f"<<ERROR: {overall_error}>>"
+    
+    # Process Microphone Input
+    if isinstance(audio_data, tuple):
+        sr, audio_samples = audio_data
+        audio_samples = (audio_samples / 32768.0).astype(np.float32)
+        if sr != ASR_SAMPLING_RATE:
+            audio_samples = librosa.resample(audio_samples, orig_sr=sr, target_sr=ASR_SAMPLING_RATE)
+    
+    # Process File Upload Input
+    else:
+        if not isinstance(audio_data, str):
+            return "<<ERROR: Invalid Audio Input>>"
+        audio_samples = librosa.load(audio_data, sr=ASR_SAMPLING_RATE, mono=True)[0]
+
+    inputs = processor(audio_samples, sampling_rate=ASR_SAMPLING_RATE, return_tensors="pt")
+
+    # **Step 1: Transcribe without Language Detection**
+    with torch.no_grad():
+        outputs = model(**inputs).logits
+        ids = torch.argmax(outputs, dim=-1)[0]
+        raw_transcription = processor.decode(ids)
+
+    # **Step 2: Detect Language from Transcription**
+    detected_lang = detect_language(raw_transcription)
+    lang_code = "eng" if detected_lang == "en" else "swh"
+
+    # **Step 3: Reload Model with Correct Adapter**
+    processor.tokenizer.set_target_lang(lang_code)
+    model.load_adapter(lang_code)
+
+    # **Step 4: Transcribe Again with Correct Adapter**
+    with torch.no_grad():
+        outputs = model(**inputs).logits
+        ids = torch.argmax(outputs, dim=-1)[0]
+        final_transcription = processor.decode(ids)
+
+    return f"Detected Language: {detected_lang.upper()}\n\nTranscription:\n{final_transcription}"