app.py

import gradio as gr
from auto_gptq import AutoGPTQForCausalLM
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from utils import build_faiss_index, retrieve

# Load documents
with open("documents/1mg_rag.txt") as f:
    docs = [line.strip() for line in f if line.strip()]

# Build FAISS index
index, _ = build_faiss_index(docs)

# Load quantized Mistral 7B
model_id = "TheBloke/Mistral-7B-Instruct-v0.2-GPTQ"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoGPTQForCausalLM.from_quantized(model_id, device_map="auto", trust_remote_code=True)

generator = pipeline("text-generation", model=model, tokenizer=tokenizer)

def answer_question(query):
    context = "\n".join(retrieve(query, index, docs))
    prompt = f"[INST] Use the following context to answer the question.\n\nContext:\n{context}\n\nQuestion: {query} [/INST]"
    result = generator(prompt, max_new_tokens=256, do_sample=True, temperature=0.7)
    return result[0]['generated_text']

gr.Interface(fn=answer_question, inputs="text", outputs="text", title="Mistral RAG").launch()