rag_server.py

import gradio as gr
import os
import requests
from io import BytesIO
from PyPDF2 import PdfReader
from tempfile import NamedTemporaryFile
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from huggingface_hub import InferenceClient
from gradio.exceptions import Error
from transformers import AutoModel
import streamlit as st

# --- Konfiguration ---
os.environ["HF_HOME"] = "/app/hf_cache"  # Verwenden Sie einen absoluten Pfad innerhalb des Containers und erzwingen den Cache!
HF_API_TOKEN = os.environ.get("HF_API_TOKEN")  # Lesen Sie den Token aus der Umgebungsvariable
MODEL_NAME = "dannyk97/mistral-screenplay-model"

# --- Hilfsfunktionen ---

def query_huggingface_inference_endpoints(prompt):
    """
    Stellt eine Anfrage an die Hugging Face Inference API.
    """
    try:
        client = InferenceClient(token=HF_API_TOKEN)
        result = client.text_generation(prompt, model=MODEL_NAME)
        return result
    except Exception as e:
        return f"Fehler bei der Anfrage an Hugging Face API: {e}"

# Function to download PDF from Google Drive
def download_pdf_from_drive(drive_link):
    file_id = drive_link.split('/d/')[1].split('/')[0]
    download_url = f"https://drive.google.com/uc?id={file_id}&export=download"
    response = requests.get(download_url)
    if response.status_code == 200:
        return BytesIO(response.content)
    else:
        raise Exception("Failed to download the PDF file from Google Drive.")

# Function to extract text from a PDF
def extract_text_from_pdf(pdf_stream):
    pdf_reader = PdfReader(pdf_stream)
    text = ""
    for page in pdf_reader.pages:
        text += page.extract_text()
    return text

# Function to split text into chunks
def chunk_text(text, chunk_size=500, chunk_overlap=50):
    text_splitter = RecursiveCharacterTextSplitter(
        chunk_size=chunk_size, chunk_overlap=chunk_overlap
    )
    return text_splitter.split_text(text)

def create_embeddings_and_store(chunks):
    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
    vector_db = FAISS.from_texts(chunks, embedding=embeddings)
    return vector_db

# Function to query the vector database and interact with Hugging Face Inference API
def query_vector_db(query, vector_db):
    # Retrieve relevant documents
    docs = vector_db.similarity_search(query, k=3)
    context = "\n".join([doc.page_content for doc in docs])

    # Interact with the Text Generation API
    prompt = f"Nutze diesen Kontext um die Frage zu beantworten: {context}\nFrage: {query}"
    try:
        output = query_huggingface_inference_endpoints(prompt)
        return output
    except Exception as e:
        return f"FEHLER: {str(e)}"

# Streamlit app
st.title("RAG-Based Application with Google Drive Support")

# Predefined list of Google Drive links - HIER DEFINIERT!
drive_links = [
    "https://drive.google.com/file/d/1PW8PJQC1EqYpsk8AhqrE4OS5cy57sqJ4/view?usp=drive_link"
    # Add more links here as needed
]

st.write("Processing the predefined Google Drive links...")

all_chunks = []

# Process each predefined Google Drive link
for link in drive_links:
    try:
        st.write(f"Processing link: {link}")
        # Download PDF
        pdf_stream = download_pdf_from_drive(link)
        st.write("PDF Downloaded Successfully!")
        
        # Extract text
        text = extract_text_from_pdf(pdf_stream)
        st.write("PDF Text Extracted Successfully!")
        
        # Chunk text
        chunks = chunk_text(text)
        st.write(f"Created {len(chunks)} text chunks.")
        all_chunks.extend(chunks)
    except Exception as e:
        st.write(f"Error processing link {link}: {e}")

if all_chunks:
    # Generate embeddings and store in FAISS
    vector_db = create_embeddings_and_store(all_chunks)
    st.write("Embeddings Generated and Stored Successfully!")
    
    # User query input
    user_query = st.text_input("Enter your query:")
    if user_query:
        response = query_vector_db(user_query, vector_db)
        st.write("Response from LLM:")
        st.write(response)