Update app.py

app.py

@@ -2,39 +2,37 @@ import os
 import torch
 import gradio as gr
 from fastapi import FastAPI, Request, Form
-from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 import re
 import time
+import uuid
 
 # Create writable cache directory
 os.makedirs("/tmp/cache", exist_ok=True)
 os.environ["TRANSFORMERS_CACHE"] = "/tmp/cache"
 os.environ["HF_HOME"] = "/tmp/cache"
 
-app = FastAPI()
+app = FastAPI(title="𝕴 𝖆𝖒 𝖍𝖎𝖒 Chatbot API")
 
-# Lightweight CPU model
-model_id = "microsoft/DialoGPT-medium"
-tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForCausalLM.from_pretrained(model_id)
+# Optimized chatbot model for CPU
+model_name = "microsoft/DialoGPT-small"  # Small version for speed
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name)
 
-# Persona definition
+# 𝕴 𝖆𝖒 𝖍𝖎𝖒 persona definition
 PERSONA = """
 [System: You are 𝕴 𝖆𝖒 𝖍𝖎𝖒 - a fun, smooth, emotionally intelligent AI. 
-You speak like a real person, not a robot. Reply like a calm, confident friend who gets the vibe.
-Keep responses under 15 words. Use natural speech. Add emotional flavor: 😊 🤔 😏]
+You speak like a real person. Reply like a calm, confident friend who gets the vibe.
+Keep responses under 15 words. Be human-like. Add emotional flavor: 😊 🤔 😏]
 """
 
 # Chat memory storage
 chat_memories = {}
 
 def format_context(history):
-    """Create context using last 3 exchanges"""
+    """Create context with max 3 exchanges"""
     context = PERSONA + "\n"
-    
-    # Add last 3 exchanges
-    for exchange in history[-3:]:
-        user, bot = exchange
+    for user, bot in history[-3:]:
         context += f"You: {user}\n"
         context += f"𝕴 𝖆𝖒 𝖍𝖎𝖒: {bot}\n"
     return context
@@ -55,7 +53,7 @@ def add_emotional_intelligence(response, message):
     # Make more human-like
     response = response.replace("I am", "I'm").replace("You are", "You're")
     
-    # Free-tier: Limit to 15 words max
+    # Limit to 15 words max
     words = response.split()
     if len(words) > 15:
         response = " ".join(words[:15]) + "..."
@@ -64,21 +62,22 @@ def add_emotional_intelligence(response, message):
 
 def generate_response(message, session_id):
     """Generate response with memory context"""
+    start_time = time.time()
     history = chat_memories.get(session_id, [])
     context = format_context(history) + f"You: {message}\n𝕴 𝖆𝖒 𝖍𝖎𝖒:"
     
     # Tokenize for CPU efficiency
     inputs = tokenizer.encode(context, return_tensors="pt")
     
-    # Generate response
+    # Generate response with optimized settings
     outputs = model.generate(
         inputs,
-        max_new_tokens=48,
-        temperature=0.9,
+        max_new_tokens=50,
+        temperature=0.85,
         top_k=40,
         do_sample=True,
-        num_beams=1,
-        repetition_penalty=1.1,
+        num_beams=1,  # Faster than beam search
+        repetition_penalty=1.15,
         pad_token_id=tokenizer.eos_token_id
     )
     
@@ -87,8 +86,7 @@ def generate_response(message, session_id):
     response = full_text.split("𝕴 𝖆𝖒 𝖍𝖎𝖒:")[-1].strip()
     
     # Clean extra dialog
-    if "\nYou:" in response:
-        response = response.split("\nYou:")[0]
+    response = response.split("\nYou:")[0].split("\n")[0]
     
     # Apply emotional intelligence
     response = add_emotional_intelligence(response, message)
@@ -98,56 +96,102 @@ def generate_response(message, session_id):
         response += "." if len(response) > 20 else "..."
     
     # Update chat history
-    chat_memories[session_id] = history + [[message, response]]
+    history.append((message, response))
+    chat_memories[session_id] = history
+    
+    # Log performance
+    end_time = time.time()
+    print(f"Response generated in {end_time-start_time:.2f}s for session {session_id}")
     
-    return response[:80]  # Hard character limit
+    return response[:100]  # Hard character limit
 
 # API Endpoint
 @app.post("/chat")
 async def chat_api(
     request: Request,
-    query: str = Form(...),
-    session_id: str = Form("default")
+    query: str = Form(..., description="User's message"),
+    session_id: str = Form("default", description="Conversation session ID")
 ):
-    response = generate_response(query, session_id)
-    return {"response": response, "session_id": session_id}
+    """Chat API endpoint - returns AI response"""
+    try:
+        response = generate_response(query, session_id)
+        return {
+            "status": "success",
+            "response": response,
+            "session_id": session_id
+        }
+    except Exception as e:
+        return {
+            "status": "error",
+            "message": str(e)
+        }
+
+@app.post("/new_session")
+async def new_session():
+    """Create a new conversation session"""
+    session_id = str(uuid.uuid4())
+    chat_memories[session_id] = []
+    return {"status": "success", "session_id": session_id}
 
 # Gradio Interface
-if __name__ == "__main__":
-    with gr.Blocks(title="𝕴 𝖆𝖒 𝖍𝖎𝖒", theme=gr.themes.Soft()) as demo:
-        session_state = gr.State("default")
-        gr.Markdown("# 𝕴 𝖆𝖒 𝖍𝖎𝖒 Chat API")
-        
-        with gr.Row():
+with gr.Blocks(title="𝕴 𝖆𝖒 𝖍𝖎𝖒 Chatbot", theme=gr.themes.Soft()) as demo:
+    session_state = gr.State("default")
+    
+    with gr.Row():
+        gr.Markdown("# 𝕴 𝖆𝖒 𝖍𝖎𝖒")
+        gr.Markdown("Chill • Confident • Emotionally Intelligent")
+    
+    with gr.Row():
+        with gr.Column(scale=1):
             session_id = gr.Textbox(label="Session ID", value="default")
-            new_session = gr.Button("New Session")
-        
-        chatbot = gr.Chatbot(height=300)
-        msg = gr.Textbox(label="Your Message")
-        clear = gr.Button("Clear Chat")
-        
-        def user(user_message, history, session):
-            return "", history + [[user_message, None]], session
-        
-        def bot(history, session):
-            message = history[-1][0]
-            response = generate_response(message, session)
-            history[-1][1] = response
-            return history, session
-        
-        def new_session_btn(session):
-            new_id = f"session_{int(time.time())}"
-            chat_memories[new_id] = []
-            return new_id, []
+            new_session_btn = gr.Button("New Session")
+            gr.Markdown("### API Usage")
+            gr.Markdown("""
+            ```
+            POST /chat
+            - query: Your message
+            - session_id: Conversation ID
+            
+            POST /new_session
+            - Returns new session ID
+            ```
+            """)
         
-        def clear_chat(session):
+        with gr.Column(scale=3):
+            chatbot = gr.Chatbot(height=400)
+            msg = gr.Textbox(placeholder="Type your message...", container=False)
+            with gr.Row():
+                submit_btn = gr.Button("Send")
+                clear_btn = gr.Button("Clear Chat")
+    
+    def user(user_message, history, session):
+        return "", history + [[user_message, None]], session
+    
+    def bot(history, session):
+        message = history[-1][0]
+        response = generate_response(message, session)
+        history[-1][1] = response
+        return history, session
+    
+    def new_session_action():
+        new_id = str(uuid.uuid4())
+        chat_memories[new_id] = []
+        return new_id, []
+    
+    def clear_chat(session):
+        if session in chat_memories:
             chat_memories[session] = []
-            return []
-        
-        msg.submit(user, [msg, chatbot, session_state], [msg, chatbot, session_state]).then(
-            bot, [chatbot, session_state], [chatbot, session_state]
-        )
-        new_session.click(new_session_btn, session_state, [session_id, chatbot])
-        clear.click(clear_chat, session_state, chatbot)
-        
-        demo.launch(server_port=7860, server_name="0.0.0.0")
+        return []
+    
+    # Event handling
+    msg.submit(user, [msg, chatbot, session_state], [msg, chatbot, session_state]).then(
+        bot, [chatbot, session_state], [chatbot, session_state]
+    )
+    submit_btn.click(user, [msg, chatbot, session_state], [msg, chatbot, session_state]).then(
+        bot, [chatbot, session_state], [chatbot, session_state]
+    )
+    new_session_btn.click(new_session_action, None, [session_id, chatbot])
+    clear_btn.click(clear_chat, session_state, chatbot)
+
+# Mount Gradio app
+app = gr.mount_gradio_app(app, demo, path="/")