Debugging

app.py

@@ -1,42 +1,35 @@
 import gradio as gr
-print("Gradio version:", gr.__version__)
-
 import torch
 from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
 from IndicTransToolkit.processor import IndicProcessor
-import gradio as gr
 import requests
 from datetime import datetime
 import tempfile
 from gtts import gTTS
 import os
 
-# Supabase configuration
-SUPABASE_URL = "https://gptmdbhzblfybdnohqnh.supabase.co"
-SUPABASE_API_KEY = "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJzdXBhYmFzZSIsInJlZiI6ImdwdG1kYmh6YmxmeWJkbm9ocW5oIiwicm9sZSI6ImFub24iLCJpYXQiOjE3NDc0NjY1NDgsImV4cCI6MjA2MzA0MjU0OH0.CfWArts6Kd_x7Wj0a_nAyGJfrFt8F7Wdy_MdYDj9e7U"
-SUPABASE_TABLE = "translations"
-
 # Device configuration
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 
+# Supabase configuration
+SUPABASE_URL = "https://gptmdbhzblfybdnohqnh.supabase.co"
+SUPABASE_API_KEY = "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9..."
+
 # Load translation models
 model_en_to_indic = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/indictrans2-en-indic-1B", trust_remote_code=True).to(DEVICE)
 tokenizer_en_to_indic = AutoTokenizer.from_pretrained("ai4bharat/indictrans2-en-indic-1B", trust_remote_code=True)
 model_indic_to_en = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/indictrans2-indic-en-1B", trust_remote_code=True).to(DEVICE)
 tokenizer_indic_to_en = AutoTokenizer.from_pretrained("ai4bharat/indictrans2-indic-en-1B", trust_remote_code=True)
-ip = IndicProcessor(inference=True)
 
-# Whisper STT pipeline (keep as is)
+ip = IndicProcessor(inference=True)
 asr = pipeline("automatic-speech-recognition", model="openai/whisper-small")
 
 
-# Save to Supabase
 def save_to_supabase(input_text, output_text, direction):
     if not input_text.strip() or not output_text.strip():
         return "Nothing to save."
 
     table_name = "translations" if direction == "en_to_ks" else "ks_to_en_translations"
-
     payload = {
         "timestamp": datetime.utcnow().isoformat(),
         "input_text": input_text,
@@ -61,7 +54,7 @@ def save_to_supabase(input_text, output_text, direction):
         print("SAVE EXCEPTION:", e)
         return "❌ Save request error."
 
-# Fetch translation history
+
 def get_translation_history(direction="en_to_ks"):
     table_name = "translations" if direction == "en_to_ks" else "ks_to_en_translations"
 
@@ -86,7 +79,7 @@ def get_translation_history(direction="en_to_ks"):
         print("HISTORY FETCH ERROR:", e)
         return "Error loading history."
 
-# Translation function
+
 def translate(text, direction):
     if not text.strip():
         return "Please enter some text.", gr.update(), gr.update()
@@ -122,7 +115,7 @@ def translate(text, direction):
         print("Translation Error:", e)
         return "⚠️ Translation failed.", gr.update(), gr.update()
 
-# Transcribe English audio
+
 def transcribe_audio(audio_path):
     try:
         result = asr(audio_path)
@@ -131,7 +124,7 @@ def transcribe_audio(audio_path):
         print("STT Error:", e)
         return "⚠️ Transcription failed."
 
-# Synthesize English TTS using gTTS for ks_to_en direction
+
 def synthesize_tts(text, direction):
     if direction == "ks_to_en" and text.strip():
         try:
@@ -144,7 +137,7 @@ def synthesize_tts(text, direction):
             print("TTS Error:", e)
     return None
 
-# Direction switch
+
 def switch_direction(direction, input_text_val, output_text_val):
     new_direction = "ks_to_en" if direction == "en_to_ks" else "en_to_ks"
     input_label = "Kashmiri Text" if new_direction == "ks_to_en" else "English Text"
@@ -155,7 +148,23 @@ def switch_direction(direction, input_text_val, output_text_val):
         gr.update(value=input_text_val, label=output_label)
     )
 
-# Gradio interface
+
+def handle_audio_translation(audio_path, direction):
+    if direction == "en_to_ks":
+        transcription = transcribe_audio(audio_path)
+        if transcription.startswith("⚠️"):
+            return transcription, "", "", None
+        translation, _, _ = translate(transcription, direction)
+        return transcription, translation, transcription, None
+    else:
+        # Assume audio_path is not used; rely on text in the input box
+        transcription = transcribe_audio(audio_path)
+        translation, _, _ = translate(transcription, direction)
+        tts_audio = synthesize_tts(translation, direction)
+        return transcription, translation, transcription, tts_audio
+
+
+# === Gradio Interface ===
 with gr.Blocks() as interface:
     gr.HTML("""
     <div style="display: flex; justify-content: space-between; align-items: center; padding: 10px;">
@@ -181,10 +190,11 @@ with gr.Blocks() as interface:
 
     with gr.Row():
         audio_input = gr.Audio(type="filepath", label="🎙️ Upload or record English audio")
-        audio_output = gr.Audio(label="🔊 English Output Audio", type="filepath", interactive=False)
+        audio_output = gr.Audio(label="🔊 English Output Audio")
+
     stt_translate_button = gr.Button("🎤 Transcribe & Translate")
 
-    # Click events
+    # Events
     translate_button.click(
         fn=translate,
         inputs=[input_text, translation_direction],
@@ -208,18 +218,10 @@ with gr.Blocks() as interface:
     )
 
     stt_translate_button.click(
-        fn=transcribe_audio,
-        inputs=audio_input,
-        outputs=input_text
-    ).then(
-        fn=translate,
-        inputs=[input_text, translation_direction],
-        outputs=[output_text, input_text, output_text]
-    ).then(
-        fn=synthesize_tts,
-        inputs=[output_text, translation_direction],
-        outputs=audio_output
+        fn=handle_audio_translation,
+        inputs=[audio_input, translation_direction],
+        outputs=[input_text, output_text, input_text, audio_output]
     )
 
 if __name__ == "__main__":
-    interface.queue().launch()
+    interface.queue().launch(share=True)

requirements.txt

@@ -5,3 +5,7 @@ gradio==5.32.0
 requests
 git+https://github.com/VarunGumma/IndicTransToolkit.git
 gTTS
+pydub
+ffmpeg-python
+soundfile
+accelerate