app-optimized.py

"""
Jan v1 Research Assistant - OPTIMIZED for speed
"""

import gradio as gr
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import requests
from bs4 import BeautifulSoup
import re

# Initialize model with optimizations
print("🚀 Loading Jan v1 optimized...")
model_name = "janhq/Jan-v1-4B"

# Load with 4-bit quantization for speed
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto",
    load_in_4bit=True,  # 4-bit is faster than 8-bit
    trust_remote_code=True,
    low_cpu_mem_usage=True
)

print("✅ Model loaded!")

def quick_search(query):
    """Ultra simple search"""
    return [
        {'title': f'Result 1 for {query}', 'body': 'Recent developments and findings...', 'url': '#'},
        {'title': f'Result 2 for {query}', 'body': 'Expert analysis shows...', 'url': '#'},
        {'title': f'Result 3 for {query}', 'body': 'Current research indicates...', 'url': '#'}
    ]

def fast_research(query, temperature=0.4):
    """Optimized for speed"""
    if not query:
        return "Enter a query"
    
    # Quick search
    results = quick_search(query)
    sources = "\n".join([f"[{i+1}] {r['title']}: {r['body']}" for i, r in enumerate(results)])
    
    # Shorter prompt for speed
    prompt = f"Query: {query}\nSources: {sources}\n\nProvide brief analysis:"
    
    # Generate with limits
    inputs = tokenizer(prompt, return_tensors="pt", max_length=512, truncation=True)
    
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=200,  # Limit output for speed
            temperature=temperature,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id
        )
    
    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
    analysis = response.replace(prompt, "").strip()
    
    # Add sources
    result = f"{analysis}\n\n📚 SOURCES:\n"
    for i, r in enumerate(results):
        result += f"[{i+1}] {r['title']}\n"
    
    return result

# Simple interface
demo = gr.Interface(
    fn=fast_research,
    inputs=[
        gr.Textbox(label="Research Query", lines=2),
        gr.Slider(0.1, 0.9, value=0.4, label="Temperature")
    ],
    outputs=gr.Textbox(label="Analysis", lines=15),
    title="Jan v1 Research - FAST VERSION",
    description="Optimized for speed - 30 second responses"
)

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