src/streamlit_app.py

import os
import streamlit as st
import numpy as np
from sentence_transformers import SentenceTransformer
from ingest import build_index
from rag import retrieve
from groq import Groq
from config import EMBEDDING_MODEL
from pdf_utils import answer_to_pdf

# ---------------- CONFIG ----------------
st.set_page_config(page_title="OTT Bot", layout="wide")

# ---------------- LOAD MODELS ----------------
@st.cache_resource
def load_embedder():
    return SentenceTransformer(EMBEDDING_MODEL)

embedder = load_embedder()

# ---------------- GROQ CLIENT ----------------
GROQ_API_KEY = os.environ.get("GROQ_API_KEY")
if not GROQ_API_KEY:
    st.error("❌ GROQ_API_KEY not set in HF Secrets")
    st.stop()

client = Groq(api_key=GROQ_API_KEY)

# ---------------- UI ----------------
st.title("📚 OTT Bot")
st.markdown("Upload PDFs via HF Dataset and query them using AI")

# ---------------- INGESTION ----------------
if st.button("🔨 Build Index"):
    with st.spinner("Building FAISS index from dataset..."):
        build_index()
    st.success("✅ Index built successfully")

st.divider()

# ---------------- QUESTION INPUT ----------------
question = st.text_input(
    "Ask a question about the document",
    placeholder="e.g. What is the main concept discussed?"
)

# ---------------- ANSWER ----------------
if question:
    with st.spinner("Searching document..."):
        query_embedding = embedder.encode([question]).astype("float32")
        contexts = retrieve(query_embedding)

        context_text = "\n\n".join(c["text"] for c in contexts)

        prompt = f"""
You are an expert tutor.

Using ONLY the information in the context below,
provide a detailed and well-structured answer.

Guidelines:
- Explain step by step
- Use headings or bullet points where useful
- Give examples
- Do NOT keep the answer short
- If information is missing, say so clearly

Context:
{context_text}

Question:
{question}

Answer clearly and concisely.
"""

    with st.spinner("Thinking..."):
        response = client.chat.completions.create(
            model="llama-3.1-8b-instant",
            messages=[
                {"role": "system", "content": "You are a helpful assistant."},
                {"role": "user", "content": prompt},
            ],
        )

    answer_text = response.choices[0].message.content

    st.subheader("Answer")
    st.write(answer_text)

    # ---------------- PDF DOWNLOAD ----------------
    pdf_buffer = answer_to_pdf(question, answer_text)

    st.download_button(
        label="📄 Download Answer as PDF",
        data=pdf_buffer,
        file_name="OTT_Bot_Answer.pdf",
        mime="application/pdf"
    )