app.py

import gradio as gr
from huggingface_hub import InferenceClient

# Initialize a list to store the conversation history
conversation_history = []

# Function to interact with the model using the Inference API
def chat_with_model(user_input, hf_api_key):
    global conversation_history
    
    if not hf_api_key:
        return "Error: Please provide your Hugging Face API key."

    try:
        # Initialize the InferenceClient with the provided API key
        client = InferenceClient(api_key=hf_api_key)

        # Add the user's message to the conversation history
        conversation_history.append({"role": "user", "content": user_input})

        # Define the system message (defining the assistant role)
        system_message = {
            "role": "system", 
            "content": "You are a code assistant that helps with code generation, debugging, and explanations."
        }

        # Add system message to the conversation history
        if len(conversation_history) == 1:  # Add system message only once
            conversation_history.insert(0, system_message)

        # Ensure the conversation history doesn't exceed token limits
        if len(conversation_history) > 10:  # Keep the last 10 messages
            conversation_history = [system_message] + conversation_history[-10:]

        # Create a stream for chat completions using the API
        stream = client.chat.completions.create(
            model="Qwen/Qwen2.5-Coder-32B-Instruct", 
            messages=conversation_history, 
            max_tokens=500,
            stream=True
        )
        
        # Collect the generated response from the model
        response = ""
        for chunk in stream:
            response += chunk.choices[0].delta.content
        
        # Add the assistant's response to the conversation history
        conversation_history.append({"role": "assistant", "content": response})
        
        return response

    except Exception as e:
        return f"Error: {e}"

# Create the Gradio interface with a light theme and black text
with gr.Blocks(
    css="""
        body {
            background-color: #f9f9f9;
            color: #000000;
        }
        .gradio-container {
            background-color: #ffffff;
            color: #000000;
        }
        .gradio-container input, .gradio-container textarea, .gradio-container button {
            color: #000000;
            background-color: #ffffff;
            border: 1px solid #cccccc;
        }
        .gradio-container textarea::placeholder, .gradio-container input::placeholder {
            color: #777777;
        }
        .gradio-container button {
            background-color: #f0f0f0;
            border: 1px solid #cccccc;
        }
        .gradio-container button:hover {
            background-color: #e6e6e6;
        }
        #title, #description {
            color: #000000 !important;
        }
    """
) as demo:
    gr.Markdown("<h1 id='title'>Code Assistant with Qwen2.5-Coder</h1>")
    gr.Markdown("<p id='description'>Ask me anything about coding! Enter your Hugging Face API key to start.</p>")
    
    # Create the input and output interface
    with gr.Row():
        user_input = gr.Textbox(lines=5, placeholder="Ask me anything about coding...")
        api_key = gr.Textbox(lines=1, placeholder="Enter your Hugging Face API key", type="password")
    
    # Create the output display
    output = gr.Textbox(label="Response")

    # Button for submitting queries
    submit_button = gr.Button("Submit")
    submit_button.click(chat_with_model, inputs=[user_input, api_key], outputs=output)

# Launch the app
demo.launch()