Update app.py

app.py

@@ -79,7 +79,7 @@ pages = loader.load_and_split(text_splitter=text_splitter)
 pages_chunks = [page.page_content for page in pages]
 print(f"Generated {len(pages_chunks)} chunks of {pm_kisan_doc}")
 
-pages_chunks[8]
+# pages_chunks[8]
 
 embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
 
@@ -107,9 +107,7 @@ def get_gemini_output(prompt, temperature=0.6):
 
     genai.configure(api_key= gemini_api_key)
     model =  genai.GenerativeModel(model_name='gemini-pro')
-    answer = model.generate_content(prompt,
-                                    generation_config=genai.types.GenerationConfig(
-        temperature=0.1))
+    answer = model.generate_content(prompt)
 
     return answer.text
 
@@ -235,7 +233,7 @@ def run_tts(text, tts_lang):
   tts_lang = lang_map[tts_lang]
   print(f"Lang code: {tts_lang}")
 
-
+  print("TTS WORKING DIRECTORY:",os.getcwd())
   tts_command = f'python3 -m TTS.bin.synthesize --text "{text}" \
     --model_path /models/v1/{tts_lang}/fastpitch/best_model.pth \
     --config_path /models/v1/{tts_lang}/fastpitch/config.json \
@@ -254,7 +252,7 @@ def run_tts(text, tts_lang):
 
   os.system(tts_command)
 
-
+os.makedirs('asr_models', exist_ok=True)
 def download_ai4b_asr_model(lang: str):
 
   available_langs = {
@@ -285,11 +283,10 @@ def download_ai4b_asr_model(lang: str):
       url = f"https://objectstore.e2enetworks.net/indic-asr-public/indicConformer/ai4b_indicConformer_{available_langs[lang]}.nemo"
       try:
           result= subprocess.run(['wget', url, '-O', download_path], check=True, capture_output=True, text=True)
-          print(result.stdout)
+          print("ASR MODEL DOWNLOADED SUCCESSFULLY",result.stdout)
       except subprocess.CallProcessError as e:
           print(f"Error occured: {e.stderr}")
           raise
-          
 
   return download_path
 
@@ -309,14 +306,14 @@ ensure_numpy_version()
 subprocess.run(['pip', 'install', 'numba==0.60.0'], check=True)
 
 
-import librosa
+# import librosa
 
-def preprocess_audio(audio_path):
-    audio,sr = librosa.load(audio_path,sr=None,mono=True)
-    return audio, sr
+# def preprocess_audio(audio_path):
+#     audio,sr = librosa.load(audio_path,sr=None,mono=True)
+#     return audio, sr
 
 def transcribe(audio: str, lang: str):
-    audio, sr = preprocess_audio(audio)
+    # audio, sr = preprocess_audio(audio)
 
     lang_map = {
       "odia": "or",
@@ -361,26 +358,12 @@ def process_user_query(user_query, retrieved_doc):
 
   prompt = prompt_template.format(user_query=user_query, retrieved_doc=retrieved_doc)
   print("Input prompt:", prompt)
+  
   processed_doc = get_gemini_output(prompt)
   print("Output prompt:",processed_doc)
 
   return processed_doc
 
-#Context awareness
-# from collections import deque
-
-# class ContextManger:
-#   def __init__(self,max_history=7):
-#     self.history = deque(maxlen=max_history)
-
-#   def add_interaction(self,query,response):
-#     self.history.append((query,response))
-
-#   def get_context(self):
-#     return list(self.history)
-
-# context_manager = ContextManger()
-
 
 import traceback
 
@@ -393,7 +376,6 @@ def process_gradio_input(audio, user_lang):
     # Convert the Indic text from transcription to English, so that GPT-3.5 can process it
     print(f"Translating indic to en..")
     indic_to_en = indic_translate(src_lang=user_lang, tgt_lang="english", sents_to_translate=[query_transcription])[0]
-    print("IIIIIIIIIIIIIIINNNNNNNNNNNNNNDDDDIIIIIIIICCCCCCCCC_TRANSLATE:",indic_to_en)
 
 
     # context = context_manager.get_context()
@@ -425,6 +407,7 @@ def process_gradio_input(audio, user_lang):
     run_tts(text=en_to_indic_doc, tts_lang=user_lang)
     print("Finished running TTS")
 
+    print("PRESENT WORKING DIRECTORY OF AUDIO SAVED:", os.getcwd())
     audio_outfile_path = "tts_output.wav"