app.py

import streamlit as st
import chromadb
import requests
import os

# HF model to use (small + free)
MODEL_ID = "mistralai/Mistral-7B-Instruct-v0.1"
API_URL = f"https://api-inference.huggingface.co/models/{MODEL_ID}"

API_TOKEN = st.secrets["HUGGINGFACEHUB_API_TOKEN"]

# Setup headers
headers = {
    "Authorization": f"Bearer {API_TOKEN}"
}

# Load Chroma DB
chroma_client = chromadb.PersistentClient(path="chroma_store")
collection = chroma_client.get_or_create_collection(name="tech_docs")

# HF API call
def query_huggingface(prompt):
    payload = {
        "inputs": prompt,
        "options": {"wait_for_model": True}
    }
    response = requests.post(API_URL, headers=headers, json=payload)

    # Check for HTTP or decoding errors
    if response.status_code != 200:
        return f"❌ HF API Error: {response.status_code} - {response.text}"

    try:
        result = response.json()
        if isinstance(result, list) and "generated_text" in result[0]:
            return result[0]["generated_text"]
        elif isinstance(result, dict) and "error" in result:
            return f"❌ Model Error: {result['error']}"
        else:
            return "⚠️ Unexpected model response format."
    except Exception as e:
        return f"⚠️ Failed to parse response: {e}"

# UI
st.title("💬 Ask Me Anything - Tech RAG Chatbot")

user_query = st.text_input("🔎 Ask your question:")

if user_query:
    # Retrieve top 3 matching docs from vector DB
    results = collection.query(query_texts=[user_query], n_results=3)
    context = "\n".join(results["documents"][0]) if results["documents"] else ""

    # Build prompt
    prompt = f"""Answer the question using the context below:

    Context:
    {context}

    Question:
    {user_query}

    Answer:"""

    # Send to HF API
    with st.spinner("Thinking..."):
        answer = query_huggingface(prompt)

    st.markdown("### 📢 Answer:")
    st.write(answer)