app.py

import streamlit as st
from groq import Groq
import os
from PyPDF2 import PdfReader
import requests
import numpy as np
import faiss
from sentence_transformers import SentenceTransformer

# ---------------------------
# PAGE CONFIG
# ---------------------------
st.set_page_config(page_title="Krish GPT Multi-Modal RAG", layout="wide")
st.title("🤖 Krish GPT Multi-Modal RAG")
st.caption("PDF + Image OCR + RAG using Groq LLM 🚀")

# ---------------------------
# API KEYS
# ---------------------------
groq_api_key = os.getenv("GROQ_API_KEY")
ocr_api_key = os.getenv("OCR_API_KEY")

if not groq_api_key:
    groq_api_key = st.text_input("Enter GROQ API Key", type="password")

if not ocr_api_key:
    ocr_api_key = st.text_input("Enter OCR.Space API Key", type="password")

if not groq_api_key or not ocr_api_key:
    st.stop()

client = Groq(api_key=groq_api_key)

# ---------------------------
# EMBEDDING MODEL
# ---------------------------
@st.cache_resource
def load_embedder():
    return SentenceTransformer("all-MiniLM-L6-v2")

embedder = load_embedder()

# ---------------------------
# OCR Function
# ---------------------------
def ocr_space_image(file, api_key):
    url = "https://api.ocr.space/parse/image"
    files = {'file': file}
    data = {'apikey': api_key, 'language': 'eng'}
    r = requests.post(url, files=files, data=data)
    try:
        result = r.json()
        text = result['ParsedResults'][0]['ParsedText']
    except:
        text = ""
    return text

# ---------------------------
# FILE UPLOAD
# ---------------------------
uploaded_file = st.file_uploader(
    "Upload PDF or Image", type=["pdf", "png", "jpg", "jpeg"]
)
file_text = ""

if uploaded_file:
    if uploaded_file.type == "application/pdf":
        reader = PdfReader(uploaded_file)
        for page in reader.pages:
            t = page.extract_text()
            if t:
                file_text += t
    elif "image" in uploaded_file.type:
        file_text = ocr_space_image(uploaded_file, ocr_api_key)

# ---------------------------
# TEXT CHUNKING & FAISS
# ---------------------------
def chunk_text(text, chunk_size=500):
    chunks = []
    for i in range(0, len(text), chunk_size):
        chunks.append(text[i:i+chunk_size])
    return chunks

def build_index(chunks):
    embeddings = embedder.encode(chunks)
    dim = embeddings.shape[1]
    index = faiss.IndexFlatL2(dim)
    index.add(np.array(embeddings))
    return index, embeddings

def search(query, chunks, index):
    q_emb = embedder.encode([query])
    D, I = index.search(np.array(q_emb), k=min(3, len(chunks)))
    results = [chunks[i] for i in I[0]]
    return "\n".join(results)

# ---------------------------
# PROCESS FILE
# ---------------------------
if uploaded_file and file_text:
    chunks = chunk_text(file_text)
    index, embeddings = build_index(chunks)
    st.session_state.rag_data = (chunks, index)

# ---------------------------
# CHAT MEMORY
# ---------------------------
if "messages" not in st.session_state:
    st.session_state.messages = []

for msg in st.session_state.messages:
    with st.chat_message(msg["role"]):
        st.markdown(msg["content"])

# ---------------------------
# USER PROMPT
# ---------------------------
prompt = st.chat_input("Ask anything...")

if prompt:
    st.session_state.messages.append({"role": "user", "content": prompt})
    with st.chat_message("user"):
        st.markdown(prompt)

    context = ""
    if "rag_data" in st.session_state:
        chunks, index = st.session_state.rag_data
        context = search(prompt, chunks, index)

    with st.chat_message("assistant"):
        try:
            response = client.chat.completions.create(
                model="llama-3.3-70b-versatile",
                messages=[
                    {"role": "system", "content": f"Context:\n{context}"},
                    *st.session_state.messages
                ],
                temperature=0.7,
                max_tokens=1024
            )
            reply = response.choices[0].message.content
        except Exception as e:
            reply = f"❌ Error: {str(e)}"

        st.markdown(reply)
        st.session_state.messages.append({"role": "assistant", "content": reply})