app.py

import asyncio
from aiohttp import web, WSMsgType
from google import genai
from google.genai import types
from PyPDF2 import PdfReader
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
import os

def load_pdf_chunks(pdf_path, max_chunk_length=500):
    reader = PdfReader(pdf_path)
    full_text = ""

    for page in reader.pages:
        text = page.extract_text()
        if text:
            full_text += text + "\n"

    # Split into smaller chunks
    chunks = [full_text[i:i + max_chunk_length] for i in range(0, len(full_text), max_chunk_length)]
    return chunks


PDF_PATH = "resume.pdf"
chunks = load_pdf_chunks(PDF_PATH)


embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
chunk_embeddings = embedding_model.encode(chunks, convert_to_numpy=True)

dimension = chunk_embeddings.shape[1]
index = faiss.IndexFlatL2(dimension)
index.add(chunk_embeddings)

def retrieve_context(query, top_k=7):
    query_embedding = embedding_model.encode([query], convert_to_numpy=True)
    distances, indices = index.search(query_embedding, top_k)
    return [chunks[i] for i in indices[0]]

GEMINI_API_KEY = os.getenv("GEMINI_API_KEY")
if not GEMINI_API_KEY:
    raise ValueError("GEMINI_API_KEY environment variable not set!")
# Initialize Gemini client
client = genai.Client(api_key=GEMINI_API_KEY)
model = "gemini-2.0-flash-live-001"
config = types.LiveConnectConfig(
    response_modalities=["TEXT"],
    system_instruction=types.Content(
        parts=[
            types.Part.from_text(
                text="You are Soumyajit's AI Assistant which is chatbot on Soumyajit's profile website, you have to act like a professional AI Assistant, and answer precisely in maximum 2-3 lines."
            )
        ],
        role="user"
    )
)

# WebSocket handler
async def ws_handler(request):
    ws = web.WebSocketResponse()
    await ws.prepare(request)

    async with client.aio.live.connect(model=model, config=config) as session:
        async for msg in ws:
            if msg.type == WSMsgType.TEXT:
                user_query = msg.data

                # 🔍 RAG Retrieval
                retrieved_docs = retrieve_context(user_query)
                context_text = "\n".join(retrieved_docs)

                # Combine into final prompt
                final_prompt = f"Context:\n{context_text}\n\nQuestion:\n{user_query}"

                await session.send_client_content(
                    turns={"role": "user", "parts": [{"text": final_prompt}]},
                    turn_complete=True
                )

                full_response = ""
                async for response in session.receive():
                    if response.text:
                        full_response += response.text

                await ws.send_str(full_response)

            elif msg.type == WSMsgType.ERROR:
                print('WebSocket connection closed with exception:', ws.exception())

    return ws

# Health check
async def health_check(request):
    return web.Response(text="OK")

# Main app setup
app = web.Application()
app.router.add_get("/", health_check)
app.router.add_get("/ws", ws_handler)  # WebSocket endpoint

if __name__ == "__main__":
    port = int(os.environ.get("PORT", 7860))  # Hugging Face uses 7860
    web.run_app(app, port=port, host="0.0.0.0")