app.py

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

# ── Choose the Mistral‑7B‑Instruct checkpoint ───────────────────────────────────
model_name = "mistralai/Mistral-7B-Instruct-v0.1"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(model_name)
model = torch.compile(model)

prompt_prefix = """
You are an energy-saving expert tasked to help households reduce their monthly electricity bills.
Given the user's appliance usage information (device name, wattage, hours used per day, days used per week):
1. Flag the highest energy consumers.
2. Recommend practical, empathetic, achievable actions.
3. Suggest appliance swaps (e.g. LED, inverter AC) and habit changes.
Format with bullet points.
Here is the input of the user:
"""

def generate_recommendation(appliance_info: str) -> str:
    # Build the full prompt
    prompt = prompt_prefix + appliance_info + "\n\nRecommendations:"
    # Tokenize and move inputs to the model device
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    # Generate with no grad, limited tokens
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=120,
            use_cache=True
        )
    # Decode and return only the recommendations section
    text = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return text.split("Recommendations:")[-1].strip()

# Define the Gradio interface
iface = gr.Interface(
    fn=generate_recommendation,
    inputs=gr.Textbox(lines=10, placeholder="Enter appliance usage details..."),
    outputs="text",
    title="Energy-Saving Recommendation Generator",
    description="Provide appliance usage details to receive energy-saving tips."
)

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