app.py

import gradio as gr
from transformers import pipeline, AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "thrishala/mental_health_chatbot"

try:
    # Load model with int8 quantization for CPU
    model = AutoModelForCausalLM.from_pretrained(
        model_id,
        device_map="cpu",
        torch_dtype=torch.float16,  # Use float16 for reduced memory
        low_cpu_mem_usage=True,     # Enable memory optimization
    )
    
    # Load tokenizer
    tokenizer = AutoTokenizer.from_pretrained(model_id)
    
    # Create pipeline with optimizations
    pipe = pipeline(
        "text-generation",
        model=model,
        tokenizer=tokenizer,
        torch_dtype=torch.float16,
    )

except Exception as e:
    print(f"Error loading model: {e}")
    exit()

def respond(
    message,
    history: list[tuple[str, str]],
    system_message,  # You can use this for initial instructions
    max_tokens,
    temperature,
    top_p,
):
    # 2. Construct the Prompt (Crucial!)
    prompt = f"{system_message}\n" 
    for user_msg, bot_msg in history:
        prompt += f"User: {user_msg}\nAssistant: {bot_msg}\n"
    prompt += f"User: {message}\nAssistant:"
    
    # 3. Generate with the Pipeline
    try:
        response = pipe(
            prompt,
            max_new_tokens=max_tokens,
            temperature=temperature,
            top_p=top_p,
        )[0]["generated_text"]
        #Extract the bot's reply (adjust if your model format is different)
        bot_response = response.split("Assistant:")[-1].strip()
        yield bot_response
    
    except Exception as e:
        print(f"Error during generation: {e}")
        yield "An error occurred during generation." #Handle generation errors.


# 4. Gradio Interface (No changes needed here)
demo = gr.ChatInterface(
    respond,
    additional_inputs=[
        gr.Textbox(
            value="You are a friendly and helpful mental health chatbot.",
            label="System message",
        ),
        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
        gr.Slider(
            minimum=0.1,
            maximum=1.0,
            value=0.95,
            step=0.05,
            label="Top-p (nucleus sampling)",
        ),
    ],
)

if __name__ == "__main__":
    demo.launch()