app.py

import os
import tempfile
import numpy as np
import faiss
import gradio as gr
from pypdf import PdfReader
from sentence_transformers import SentenceTransformer
from groq import Groq

# ===============================
# Groq API
# ===============================
GROQ_API_KEY = os.getenv("Rag_key")
if not GROQ_API_KEY:
    raise ValueError("Groq API key not found. Add Rag_key in HuggingFace Secrets.")

client = Groq(api_key=GROQ_API_KEY)

# ===============================
# Embedding Model (Open Source)
# ===============================
embedder = SentenceTransformer("all-MiniLM-L6-v2")

# ===============================
# PDF Reader
# ===============================
def read_pdf(pdf_path):
    reader = PdfReader(pdf_path)
    text = ""
    for page in reader.pages:
        text += page.extract_text() or ""
    return text

# ===============================
# Text Chunking
# ===============================
def chunk_text(text, chunk_size=400, overlap=50):
    chunks = []
    start = 0
    while start < len(text):
        end = start + chunk_size
        chunks.append(text[start:end])
        start = end - overlap
    return chunks

# ===============================
# FAISS Vector Store
# ===============================
def create_faiss(chunks):
    embeddings = embedder.encode(chunks)
    dim = embeddings.shape[1]
    index = faiss.IndexFlatL2(dim)
    index.add(np.array(embeddings))
    return index, embeddings

def retrieve_chunks(chunks, index, question, k=3):
    q_embedding = embedder.encode([question])
    distances, indices = index.search(np.array(q_embedding), k)
    return [chunks[i] for i in indices[0]]

# ===============================
# Groq LLM Call
# ===============================
def ask_llm(context, question):
    response = client.chat.completions.create(
        model="llama-3.3-70b-versatile",
        messages=[
            {
                "role": "system",
                "content": "Answer ONLY from the provided context. Reply in Urdu."
            },
            {
                "role": "user",
                "content": f"Context:\n{context}\n\nQuestion:\n{question}"
            }
        ]
    )
    return response.choices[0].message.content

# ===============================
# Main RAG Pipeline (FIXED)
# ===============================
def rag_pipeline(file, question):
    try:
        if file is None or not question.strip():
            return "براہ کرم PDF اپلوڈ کریں اور سوال لکھیں۔"

        # ✅ FIX: Handle HuggingFace NamedString
        if isinstance(file, str):
            pdf_path = file
        else:
            with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp:
                tmp.write(file.read())
                pdf_path = tmp.name

        text = read_pdf(pdf_path)
        if not text.strip():
            return "PDF سے متن حاصل نہیں ہو سکا۔"

        chunks = chunk_text(text)
        index, _ = create_faiss(chunks)
        relevant = retrieve_chunks(chunks, index, question)
        context = "\n".join(relevant)

        return ask_llm(context, question)

    except Exception as e:
        return f"⚠️ خرابی: {str(e)}"

# ===============================
# Gradio UI
# ===============================
ui = gr.Interface(
    fn=rag_pipeline,
    inputs=[
        gr.File(label="📄 PDF اپلوڈ کریں"),
        gr.Textbox(label="❓ سوال", placeholder="PDF سے سوال پوچھیں")
    ],
    outputs=gr.Textbox(label="📌 جواب"),
    title="Jehan Zada RAG App"
)

ui.launch()