Update app.py

app.py

@@ -1,5 +1,5 @@
 from fastapi import FastAPI, HTTPException, UploadFile, File
-from fastapi.responses import StreamingResponse, JSONResponse
+from fastapi.responses import StreamingResponse
 from pydantic import BaseModel
 import torch
 from transformers import (
@@ -25,7 +25,7 @@ app = FastAPI(title="Kinyarwanda Engine", version="1.0")
 MODEL_PATH = "/app/models/suno-bark"
 SENTIMENT_MODEL_PATH = "/app/models/sentiment"
 SAMPLE_RATE = 24000
-ASR_MODEL_PATH = "jonatasgrosman/wav2vec2-large-xlsr-53-Kinyarwanda"
+ASR_MODEL_PATH = "lucio/wav2vec2-large-xlsr-kinyarwanda"
 
 # Ensure working directory for audio
 AUDIO_DIR = "/tmp/audio"
@@ -107,6 +107,12 @@ def speech_to_text(audio_file: UploadFile = File(...)):
         audio_bytes = audio_file.file.read()
         audio, sample_rate = sf.read(BytesIO(audio_bytes))
 
+        # Resample if necessary
+        if sample_rate != 16000:
+            import librosa
+            audio = librosa.resample(audio, orig_sr=sample_rate, target_sr=16000)
+            sample_rate = 16000
+
         inputs = asr_processor(audio, sampling_rate=sample_rate, return_tensors="pt", padding=True).input_values.to(device)
 
         with torch.no_grad():