Update app.py

app.py

@@ -1,39 +1,14 @@
 import gradio as gr
-import torch
-from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
-import soundfile as sf
 import openai
 from deep_translator import GoogleTranslator
 import os
 
-# Load the Wav2Vec2 model and processor for Tamil
-model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-large-xlsr-53-tamil")
-processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-large-xlsr-53-tamil")
-
 # Initialize OpenAI API
 openai.api_key = os.environ.get("OPENAI_API_KEY")
 
-def transcribe_audio(audio_path):
-    # Load audio file
-    audio_input, sample_rate = sf.read(audio_path)
-    
-    # Resample if necessary
-    if sample_rate != 16000:
-        print(f"Resampling from {sample_rate} to 16000 Hz")
-        audio_input = librosa.resample(audio_input, sample_rate, 16000)
-    
-    # Preprocess the audio
-    input_values = processor(audio_input, return_tensors="pt", sampling_rate=16000).input_values
-    
-    # Perform inference
-    with torch.no_grad():
-        logits = model(input_values).logits
-
-    # Decode the output
-    predicted_ids = torch.argmax(logits, dim=-1)
-    transcription = processor.batch_decode(predicted_ids)[0]
-    
-    return transcription
+def translate_text(text, target_lang):
+    translator = GoogleTranslator(source='auto', target=target_lang)
+    return translator.translate(text)
 
 def get_llm_response(prompt):
     response = openai.Completion.create(
@@ -43,16 +18,9 @@ def get_llm_response(prompt):
     )
     return response.choices[0].text.strip()
 
-def translate_text(text, target_lang):
-    translator = GoogleTranslator(source='auto', target=target_lang)
-    return translator.translate(text)
-
-def process_voice_input(audio_path, output_lang):
-    # Transcribe audio
-    transcription = transcribe_audio(audio_path)
-    
-    # Translate transcription to English
-    english_query = translate_text(transcription, "en")
+def process_input(input_text, output_lang):
+    # Translate input to English
+    english_query = translate_text(input_text, "en")
     
     # Get response from LLM
     llm_response = get_llm_response(english_query)
@@ -60,23 +28,23 @@ def process_voice_input(audio_path, output_lang):
     # Translate LLM response to desired output language
     final_response = translate_text(llm_response, output_lang)
     
-    return transcription, english_query, llm_response, final_response
+    return input_text, english_query, llm_response, final_response
 
 # Gradio interface
 iface = gr.Interface(
-    fn=process_voice_input,
+    fn=process_input,
     inputs=[
-        gr.Audio(source="microphone", type="filepath"),
+        gr.Textbox(label="Input (Tamil or English)"),
         gr.Radio(["ta", "en"], label="Output Language")
     ],
     outputs=[
-        gr.Textbox(label="Transcribed Tamil"),
+        gr.Textbox(label="Original Input"),
         gr.Textbox(label="Translated English Query"),
         gr.Textbox(label="LLM Response (English)"),
         gr.Textbox(label="Final Response (Tamil/English)")
     ],
-    title="Tamil Voice Assistant",
-    description="Speak in Tamil, get responses in Tamil or English!"
+    title="Tamil-English Assistant",
+    description="Enter text in Tamil or English, get responses in Tamil or English!"
 )
 
 iface.launch()