app.py

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

# Load model and tokenizer
model_id = "Equall/SaulLM-54B-Instruct"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto"
)

@spaces.GPU()
def generate_response(message, history, system_prompt, max_tokens, temperature):
    """Generate legal analysis using Saul-54B"""
    
    # Build conversation history
    messages = []
    
    if system_prompt:
        messages.append({"role": "system", "content": system_prompt})
    
    for human, assistant in history:
        messages.append({"role": "user", "content": human})
        messages.append({"role": "assistant", "content": assistant})
    
    messages.append({"role": "user", "content": message})
    
    # Format for model
    input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    inputs = tokenizer(input_text, return_tensors="pt").to(model.device)
    
    # Generate
    outputs = model.generate(
        **inputs,
        max_new_tokens=max_tokens,
        temperature=temperature,
        do_sample=temperature > 0,
        pad_token_id=tokenizer.eos_token_id
    )
    
    response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
    return response

# Default system prompt for legal queries
DEFAULT_SYSTEM = """You are SaulLM-54B, a specialized legal language model. You provide accurate legal analysis based on U.S. and European legal systems. 

IMPORTANT DISCLAIMERS:
- This is for informational purposes only, not legal advice
- Information may not reflect recent legal developments
- Users should consult qualified legal professionals for actual legal advice
- Do not use this for decisions that could affect legal rights"""

# Build interface
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("# SaulLM-54B Legal Assistant")
    gr.Markdown("*Specialized AI for legal reasoning and analysis. Private queries, powered by Zero GPU (25 min/day free).*")
    
    with gr.Row():
        with gr.Column(scale=3):
            chatbot = gr.Chatbot(label="Legal Analysis", height=500)
            msg = gr.Textbox(
                label="Your Legal Question",
                placeholder="Ask about statutes, case law, legal concepts, or compliance...",
                lines=3
            )
            with gr.Row():
                submit = gr.Button("Submit", variant="primary")
                clear = gr.Button("Clear Chat")
        
        with gr.Column(scale=1):
            system_prompt = gr.Textbox(
                label="System Prompt",
                value=DEFAULT_SYSTEM,
                lines=12,
                max_lines=12
            )
            max_tokens = gr.Slider(
                label="Max Response Tokens",
                minimum=100,
                maximum=2000,
                value=1000,
                step=100
            )
            temperature = gr.Slider(
                label="Temperature",
                minimum=0.0,
                maximum=1.0,
                value=0.7,
                step=0.1
            )
            gr.Markdown("### Usage Tips")
            gr.Markdown("""
            - Be specific about jurisdiction
            - Cite relevant statutes/cases if known
            - Zero GPU resets after 60s idle
            - 25 min/day free compute limit
            """)
    
    def user_submit(message, history):
        return "", history + [[message, None]]
    
    def bot_respond(history, system_prompt, max_tokens, temperature):
        message = history[-1][0]
        history_context = history[:-1]
        
        response = generate_response(message, history_context, system_prompt, max_tokens, temperature)
        history[-1][1] = response
        return history
    
    msg.submit(user_submit, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot_respond, [chatbot, system_prompt, max_tokens, temperature], chatbot
    )
    submit.click(user_submit, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot_respond, [chatbot, system_prompt, max_tokens, temperature], chatbot
    )
    clear.click(lambda: None, None, chatbot, queue=False)

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