Update app.py

app.py

@@ -3,17 +3,27 @@ import requests
 import json
 import os
 from transformers import pipeline
-
-# Load a speech-to-text model
-# Using a smaller model like 'distil-whisper/distil-small.en' for efficiency
-# You might need to install 'pip install transformers torch soundfile librosa'
-asr_pipeline = pipeline("automatic-speech-recognition", model="distil-whisper/distil-small.en")
+from huggingface_hub import login
 
 # Hugging Face API details for Mistral
 API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
-HF_TOKEN = os.getenv("HF_TOKEN")
+HF_TOKEN = os.getenv("HF_TOKEN")  # Retrieve the token from environment variables
 headers = {"Authorization": f"Bearer {HF_TOKEN}"}
 
+# Log in to Hugging Face Hub programmatically
+if HF_TOKEN:
+    login(token=HF_TOKEN, add_to_git_credential=False)
+else:
+    print("Warning: HF_TOKEN environment variable not set. Some models may not be accessible.")
+
+# Load a speech-to-text model with authentication
+# Pass the token explicitly to the pipeline function
+asr_pipeline = pipeline(
+    "automatic-speech-recognition",
+    model="distil-whisper/distil-small.en",
+    use_auth_token=HF_TOKEN
+)
+
 # Prompt template
 PROMPT_TEMPLATE = """
 You are an AI translation assistant for a real-time universal translator.
@@ -65,7 +75,7 @@ def translate(text, source_lang, target_lang, emotion):
         }
 
     try:
-        raw_text = output[0]["generated_text"]
+        raw_text = output["generated_text"]
         parsed = json.loads(raw_text.strip())
     except Exception as e:
         parsed = {
@@ -81,6 +91,7 @@ def gradio_interface(audio, text, source_lang, target_lang, emotion):
     if audio is not None:
         try:
             # Transcribe the audio file using the ASR pipeline
+            # No token needed here as it's passed on initialization
             transcribed_text = asr_pipeline(audio)["text"]
             # If there's also text input, combine them
             if text: