app1

app.py

@@ -119,10 +119,8 @@ qa_chain = ConversationalRetrievalChain.from_llm(
     output_key = 'output',
 )
 
-import soundfile as sf
-import gradio as gr
-from gradio_client import Client
 import numpy as np
+import soundfile as sf
 
 # Load ASR pipeline
 transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-large")
@@ -130,43 +128,31 @@ transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-lar
 def chat_interface(question, audio_input=None, history=None):
     if audio_input is not None:
         # Function to transcribe the audio input
-
         def transcribe(audio):
-            sr, y = audio
-            y = y.astype(np.float32)
-            y /= np.max(np.abs(y))
+            # If the audio input has multiple channels, take the first channel
+            audio = audio[:, 0]  # Take the first channel
 
-            return transcriber({"sampling_rate": sr, "raw": y})["text"]
-
-        # Transcribe the audio input
-        question = transcribe(audio_input)
+            # Normalize audio
+            audio /= np.max(np.abs(audio))
 
+            # Write the audio to a temporary file
+            temp_audio_file = "temp_audio.wav"
+            sf.write(temp_audio_file, audio, 16000)  # Assuming 16kHz sample rate
 
-        return question
+            # Transcribe the audio from the temporary file
+            return transcriber(temp_audio_file)[0]['transcription']
 
-    # Original chatbot logic
-    result = qa_chain.invoke({'question': question})
-    output_string = result['output']
-
-    # Find the index of the last occurrence of "answer": in the string
-    answer_index = output_string.rfind('"answer":')
-
-    # Extract the substring starting from the "answer": index
-    answer_part = output_string[answer_index + len('"answer":'):].strip()
-
-    # Find the next occurrence of a double quote to get the start of the answer value
-    quote_index = answer_part.find('"')
-
-    # Extract the answer value between double quotes
-    answer_value = answer_part[quote_index + 1:answer_part.find('"', quote_index + 1)]
+        # Transcribe the audio input
+        question = transcribe(audio_input)
 
-    return answer_value
+    # Original chatbot logic goes here
+    return question
 
 chatbot_gradio_app = gr.Interface(
     fn=chat_interface,
     inputs=[
         gr.Textbox(lines=3, label="Type your message here"),
-        gr.Audio(label="Record your voice", type='numpy')  # Change type to "microphone"
+        gr.Audio(label="Record your voice", type='numpy')
     ],
     outputs=gr.Textbox(label="Bot's Response"),
 )