Update app.py

app.py

@@ -1,10 +1,6 @@
 import streamlit as st
-from streamlit_webrtc import webrtc_streamer, WebRtcMode, RTCConfiguration
 import whisper
 import numpy as np
-import av
-from typing import List
-import queue
 from langchain_community.llms import HuggingFaceEndpoint
 from langchain_community.embeddings import HuggingFaceBgeEmbeddings
 from langchain.memory import ConversationBufferMemory
@@ -16,6 +12,9 @@ from dotenv import load_dotenv
 import requests
 from requests.adapters import HTTPAdapter
 from requests.packages.urllib3.util.retry import Retry
+import time  # Imported time module
+from streamlit_chat import message
+from streamlit_audiorecorder import audiorecorder  # For audio recording
 
 # Load environment variables
 load_dotenv()
@@ -23,8 +22,10 @@ load_dotenv()
 # Initialize session state
 if "messages" not in st.session_state:
     st.session_state.messages = []
-if "audio_buffer" not in st.session_state:
-    st.session_state.audio_buffer = queue.Queue()
+if "audio_data" not in st.session_state:
+    st.session_state.audio_data = None
+if "recording" not in st.session_state:
+    st.session_state.recording = False
 
 # Prompt template
 PROMPT_TEMPLATE = """
@@ -81,7 +82,7 @@ embeddings = HuggingFaceBgeEmbeddings(
 )
 
 vectorstore = FAISS.from_texts(
-    ["Initial therapeutic context"], 
+    ["Initial therapeutic context"],
     embeddings
 )
 
@@ -95,18 +96,10 @@ conversation_chain = ConversationalRetrievalChain.from_llm(
     retriever=vectorstore.as_retriever(),
     memory=memory,
     combine_docs_chain_kwargs={"prompt": qa_prompt},
-    return_source_documents=True
+    return_source_documents=True,
+    output_key='answer'  # Specify output_key to fix the error
 )
 
-def video_frame_callback(frame: av.VideoFrame) -> av.VideoFrame:
-    return frame
-
-def audio_frame_callback(frame: av.AudioFrame) -> av.AudioFrame:
-    if st.session_state.recording:
-        sound = frame.to_ndarray()
-        st.session_state.audio_buffer.put(sound)
-    return frame
-
 def get_ai_response(user_input: str) -> str:
     max_retries = 3
     for attempt in range(max_retries):
@@ -156,42 +149,51 @@ def main():
     
     st.title("Darija AI Therapist 🧠")
     st.subheader("تكلم معايا بالدارجة على اللي كيجول فبالك")
-
-    # WebRTC setup
-    webrtc_ctx = webrtc_streamer(
-        key="speech-to-text",
-        mode=WebRtcMode.SENDONLY,
-        audio_receiver_size=1024,
-        rtc_configuration=RTCConfiguration(
-            {"iceServers": [{"urls": ["stun:stun.l.google.com:19302"]}]}
-        ),
-        video_frame_callback=video_frame_callback,
-        audio_frame_callback=audio_frame_callback,
-        media_stream_constraints={"video": False, "audio": True},
-    )
-
+    
     # Chat interface
-    user_input = st.text_input("اكتب رسالتك هنا:", key="text_input")
+    # Create columns for text input and mic button
+    col1, col2 = st.columns([9, 1])
+    with col1:
+        user_input = st.text_input("اكتب رسالتك هنا:", key="text_input")
+    with col2:
+        # Mic button
+        if st.session_state.recording:
+            mic_label = "🛑"
+        else:
+            mic_label = "🎤"
+        
+        if st.button(mic_label):
+            st.session_state.recording = not st.session_state.recording
+            if st.session_state.recording:
+                st.session_state.audio_data = audiorecorder("Click to stop recording")
+            else:
+                audio_data = st.session_state.audio_data
+                if audio_data is not None:
+                    # Convert byte data to numpy array
+                    audio_array = np.frombuffer(audio_data.tobytes(), dtype=np.int16)
+                    # Normalize audio data
+                    audio_array = audio_array.astype(np.float32) / np.iinfo(np.int16).max
+                    # Transcribe audio using Whisper
+                    result = whisper_model.transcribe(audio_array, language="ar")
+                    if result["text"]:
+                        # Put transcribed text into input field
+                        st.session_state.text_input = result["text"]
+                else:
+                    st.error("No audio data recorded.")
+
+    # Handle text submission
     if user_input:
         process_message(user_input)
-
-    # Process audio when recording stops
-    if webrtc_ctx.state.playing and len(st.session_state.audio_buffer) > 0:
-        audio_frames = []
-        while not st.session_state.audio_buffer.empty():
-            audio_frames.append(st.session_state.audio_buffer.get())
-        
-        if audio_frames:
-            audio_data = np.concatenate(audio_frames, axis=0)
-            text = whisper_model.transcribe(audio_data)["text"]
-            if text:
-                process_message(text)
-                st.session_state.audio_buffer = queue.Queue()  # Clear buffer
+        st.session_state.text_input = ""  # Clear input field after sending
 
     # Display chat history
-    for message in st.session_state.messages:
-        with st.chat_message(message["role"]):
-            st.write(message["content"])
+    for message_data in st.session_state.messages:
+        role = message_data["role"]
+        content = message_data["content"]
+        if role == "user":
+            message(content, is_user=True)
+        else:
+            message(content)
 
 if __name__ == "__main__":
     main()