app.py

import gradio as gr
import json
from pathlib import Path
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

# Default system prompt for the chat interface
DEFAULT_SYSTEM_PROMPT = """You are DeepThink, a helpful and knowledgeable AI assistant. You aim to provide accurate, 
informative, and engaging responses while maintaining a professional and friendly demeanor."""

class ChatInterface:
    """Main chat interface handler with memory and parameter management"""
    
    def __init__(self):
        """Initialize the chat interface with default settings"""
        self.model_name = "deepseek-ai/DeepSeek-R1-Distill-Qwen-7B"
        self.tokenizer = AutoTokenizer.from_pretrained(self.model_name)
        self.model = AutoModelForCausalLM.from_pretrained(self.model_name)
        self.chat_history = []
        self.system_prompt = DEFAULT_SYSTEM_PROMPT
        
    def load_context_from_json(self, file_obj):
        """Load additional context from a JSON file"""
        if file_obj is None:
            return "No file uploaded", self.system_prompt
        
        try:
            content = json.load(file_obj)
            if "system_prompt" in content:
                self.system_prompt = content["system_prompt"]
            return "Context loaded successfully!", self.system_prompt
        except Exception as e:
            return f"Error loading context: {str(e)}", self.system_prompt

    def generate_response(self, message, temperature, max_length, top_p, presence_penalty, frequency_penalty):
        """Generate AI response with given parameters"""
        # Format the input with system prompt and chat history
        conversation = f"System: {self.system_prompt}\n\n"
        for msg in self.chat_history:
            conversation += f"Human: {msg[0]}\nAssistant: {msg[1]}\n\n"
        conversation += f"Human: {message}\nAssistant:"

        # Generate response with specified parameters
        inputs = self.tokenizer(conversation, return_tensors="pt")
        outputs = self.model.generate(
            inputs["input_ids"],
            max_length=max_length,
            temperature=temperature,
            top_p=top_p,
            presence_penalty=presence_penalty,
            frequency_penalty=frequency_penalty,
        )
        response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
        
        # Extract assistant's response and update chat history
        response = response.split("Assistant:")[-1].strip()
        self.chat_history.append((message, response))
        
        return response, self.format_chat_history()
    
    def format_chat_history(self):
        """Format chat history for display"""
        return [(f"User: {msg[0]}", f"Assistant: {msg[1]}") for msg in self.chat_history]
    
    def clear_history(self):
        """Clear the chat history"""
        self.chat_history = []
        return self.format_chat_history()

# Initialize the chat interface
chat_interface = ChatInterface()

# Create the Gradio interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    with gr.Row():
        with gr.Column(scale=2):
            # Main chat interface
            chatbot = gr.Chatbot(
                label="Chat History",
                height=600,
                show_label=True,
            )
            
            with gr.Row():
                message = gr.Textbox(
                    label="Your message",
                    placeholder="Type your message here...",
                    lines=2
                )
                submit_btn = gr.Button("Send", variant="primary")
        
        with gr.Column(scale=1):
            # System settings and parameters
            with gr.Group(label="System Configuration"):
                system_prompt = gr.Textbox(
                    label="System Prompt",
                    value=DEFAULT_SYSTEM_PROMPT,
                    lines=4
                )
                context_file = gr.File(
                    label="Upload Context JSON",
                    file_types=[".json"]
                )
                upload_button = gr.Button("Load Context")
                context_status = gr.Textbox(label="Context Status", interactive=False)
            
            with gr.Group(label="Generation Parameters"):
                temperature = gr.Slider(
                    minimum=0.1,
                    maximum=2.0,
                    value=0.7,
                    step=0.1,
                    label="Temperature"
                )
                max_length = gr.Slider(
                    minimum=50,
                    maximum=2000,
                    value=500,
                    step=50,
                    label="Max Length"
                )
                top_p = gr.Slider(
                    minimum=0.1,
                    maximum=1.0,
                    value=0.9,
                    step=0.1,
                    label="Top P"
                )
                presence_penalty = gr.Slider(
                    minimum=0.0,
                    maximum=2.0,
                    value=0.0,
                    step=0.1,
                    label="Presence Penalty"
                )
                frequency_penalty = gr.Slider(
                    minimum=0.0,
                    maximum=2.0,
                    value=0.0,
                    step=0.1,
                    label="Frequency Penalty"
                )
            
            clear_btn = gr.Button("Clear Chat History")

    # Event handlers
    def submit_message(message, temperature, max_length, top_p, presence_penalty, frequency_penalty):
        response, history = chat_interface.generate_response(
            message, temperature, max_length, top_p, presence_penalty, frequency_penalty
        )
        return "", history

    submit_btn.click(
        submit_message,
        inputs=[message, temperature, max_length, top_p, presence_penalty, frequency_penalty],
        outputs=[message, chatbot]
    )
    
    message.submit(
        submit_message,
        inputs=[message, temperature, max_length, top_p, presence_penalty, frequency_penalty],
        outputs=[message, chatbot]
    )
    
    clear_btn.click(
        lambda: (chat_interface.clear_history(), ""),
        outputs=[chatbot, message]
    )
    
    upload_button.click(
        chat_interface.load_context_from_json,
        inputs=[context_file],
        outputs=[context_status, system_prompt]
    )

# Launch the interface
demo.launch()