Update app.py

app.py

@@ -1,105 +1,101 @@
 import gradio as gr
 import os
-import json
 import uuid
-import threading
-import time
-import re
-from dotenv import load_dotenv
 from openai import OpenAI
 from realtime_transcriber import WebSocketClient, connections, WEBSOCKET_URI, WEBSOCKET_HEADERS
 
-# ------------------ Load Secrets ------------------
-load_dotenv()
-OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
-ASSISTANT_ID = os.getenv("ASSISTANT_ID")
-
-if not OPENAI_API_KEY or not ASSISTANT_ID:
-    raise ValueError("Missing OPENAI_API_KEY or ASSISTANT_ID")
-
+# Load OpenAI API key
+OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY")
+if not OPENAI_API_KEY:
+    raise ValueError("OPENAI_API_KEY environment variable must be set")
 client = OpenAI(api_key=OPENAI_API_KEY)
-session_threads = {}
-
-# ------------------ Chat Logic ------------------
-def reset_session():
-    session_id = str(uuid.uuid4())
-    session_threads[session_id] = client.beta.threads.create().id
-    return session_id
-
-def process_chat(message, history, session_id):
-    thread_id = session_threads.get(session_id)
-    if not thread_id:
-        thread_id = client.beta.threads.create().id
-        session_threads[session_id] = thread_id
-
-    client.beta.threads.messages.create(thread_id=thread_id, role="user", content=message)
-    run = client.beta.threads.runs.create(thread_id=thread_id, assistant_id=ASSISTANT_ID)
-
-    while client.beta.threads.runs.retrieve(thread_id=thread_id, run_id=run.id).status != "completed":
-        time.sleep(1)
-
-    messages = client.beta.threads.messages.list(thread_id=thread_id)
-    for msg in reversed(messages.data):
-        if msg.role == "assistant":
-            return msg.content[0].text.value
-    return "⚠️ Assistant did not respond."
 
-def extract_image_url(text):
-    match = re.search(r'https://raw\.githubusercontent\.com/[^\s"]+\.png', text)
-    return match.group(0) if match else None
-
-def handle_chat(message, history, session_id):
-    response = process_chat(message, history, session_id)
-    history.append((message, response))
-    image = extract_image_url(response)
-    return history, image
+# Session state
+session_id = str(uuid.uuid4())
+if session_id not in connections:
+    connections[session_id] = WebSocketClient(WEBSOCKET_URI, WEBSOCKET_HEADERS, session_id)
+    connections[session_id].start()
+
+# Functions for Document Assistant
+def process_user_input(message, history):
+    if not message:
+        return "Please enter a message.", history
+
+    try:
+        thread = client.beta.threads.create()
+        client.beta.threads.messages.create(
+            thread_id=thread.id,
+            role="user",
+            content=message
+        )
+        run = client.beta.threads.runs.create(
+            thread_id=thread.id,
+            assistant_id=os.environ.get("ASSISTANT_ID")
+        )
+        while True:
+            run_status = client.beta.threads.runs.retrieve(
+                thread_id=thread.id,
+                run_id=run.id
+            )
+            if run_status.status == "completed":
+                break
+        messages = client.beta.threads.messages.list(thread_id=thread.id)
+        assistant_reply = next((m.content[0].text.value for m in reversed(messages.data) if m.role == "assistant"), "No response.")
+        history.append((message, assistant_reply))
+        return "", history
+    except Exception as e:
+        return f"❌ Error: {str(e)}", history
+
+# Functions for Realtime Voice Transcription
+def send_audio_chunk_realtime(mic_chunk):
+    if session_id not in connections:
+        return "Initializing voice session..."
+    if mic_chunk is not None:
+        sr, y = mic_chunk
+        connections[session_id].enqueue_audio_chunk(sr, y)
+    return connections[session_id].transcript
+
+def clear_transcript():
+    if session_id in connections:
+        connections[session_id].transcript = ""
+    return ""
 
-# ------------------ Voice Logic ------------------
-def create_websocket_client():
-    client_id = str(uuid.uuid4())
-    connections[client_id] = WebSocketClient(WEBSOCKET_URI, WEBSOCKET_HEADERS, client_id)
-    threading.Thread(target=connections[client_id].run, daemon=True).start()
-    return client_id
+# Gradio UI Components
+doc_image = gr.Image(label="📘 Extracted Document Image", show_label=True, elem_id="docimg", height=500, width=360)
+chatbot = gr.Chatbot(label="🧠 Document Assistant", elem_id="chatbox", bubble_full_width=False)
+prompt = gr.Textbox(placeholder="Ask about the document...", label="Ask about the document")
+send_btn = gr.Button("Send")
 
-def clear_transcript(client_id):
-    if client_id in connections:
-        connections[client_id].transcript = ""
-    return ""
+# Voice Section
+audio_in = gr.Audio(label="🎵 Audio", type="numpy", streaming=True)
+live_transcript = gr.Textbox(label="Live Transcript", lines=6)
+clear_btn = gr.Button("Clear Transcript")
 
-def send_audio_chunk(audio, client_id):
-    if client_id not in connections:
-        return "Initializing connection..."
-    sr, y = audio
-    connections[client_id].enqueue_audio_chunk(sr, y)
-    return connections[client_id].transcript
+with gr.Blocks(theme=gr.themes.Base(), css="""
+    #docimg img { object-fit: contain !important; }
+    #chatbox { height: 500px; }
+    .gr-box { border-radius: 12px; }
+""") as demo:
 
-# ------------------ UI ------------------
-with gr.Blocks(theme=gr.themes.Soft()) as demo:
     gr.Markdown("# 🧠 Document AI + 🎙️ Voice Assistant")
-
-    session_id = gr.State(value=reset_session())
-    client_id = gr.State()
+    with gr.Row():
+        with gr.Column(scale=1):
+            doc_image.render()
+        with gr.Column(scale=2):
+            chatbot.render()
 
     with gr.Row():
-        image_display = gr.Image(label="📑 Extracted Document Image", show_label=True, height=480, width=400)
-        with gr.Column():
-            chatbot = gr.Chatbot(label="💬 Document Assistant", height=480)
-            text_input = gr.Textbox(label="Ask about the document", placeholder="e.g. What is clause 3.2?")
-            send_btn = gr.Button("Send")
+        prompt.render()
+        send_btn.render()
 
-    send_btn.click(handle_chat, inputs=[text_input, chatbot, session_id], outputs=[chatbot, image_display])
-    text_input.submit(handle_chat, inputs=[text_input, chatbot, session_id], outputs=[chatbot, image_display])
+    send_btn.click(fn=process_user_input, inputs=[prompt, chatbot], outputs=[prompt, chatbot])
 
-    # Toggle Section
-    with gr.Accordion("🎤 Or Use Voice Instead", open=False):
-        with gr.Row():
-            transcript_box = gr.Textbox(label="Live Transcript", lines=7, interactive=False, autoscroll=True)
+    with gr.Accordion("🎙️ Or Use Voice Instead", open=False):
+        live_transcript.render()
         with gr.Row():
-            mic_input = gr.Audio(streaming=True)
-            clear_button = gr.Button("Clear Transcript")
-
-        mic_input.stream(fn=send_audio_chunk, inputs=[mic_input, client_id], outputs=transcript_box)
-        clear_button.click(fn=clear_transcript, inputs=[client_id], outputs=transcript_box)
-        demo.load(fn=create_websocket_client, outputs=client_id)
+            audio_in.render()
+            clear_btn.render()
+        audio_in.stream(fn=send_audio_chunk_realtime, inputs=audio_in, outputs=live_transcript)
+        clear_btn.click(fn=clear_transcript, outputs=live_transcript)
 
 demo.launch()