app.py

import os
from docx import Document
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
import pickle
from langchain_community.llms import HuggingFaceEndpoint
from langchain_community.vectorstores import FAISS
from langchain_community.embeddings import HuggingFaceEmbeddings

# Initialize the embedding model
embedding_model = SentenceTransformer('all-MiniLM-L6-v2')

# Initialize the HuggingFace LLM
llm = HuggingFaceEndpoint(
    endpoint_url="https://api-inference.huggingface.co/models/gpt2",
    model_kwargs={"api_key": os.getenv('HUGGINGFACEHUB_API_TOKEN')}
)

# Initialize the HuggingFace embeddings
embedding = HuggingFaceEmbeddings()

# Function to extract text from a Word document
def extract_text_from_docx(docx_path):
    text = ""
    try:
        doc = Document(docx_path)
        text = "\n".join([para.text for para in doc.paragraphs])
    except Exception as e:
        print(f"Error extracting text from DOCX: {e}")
    return text

# Load or create FAISS index
index_path = "faiss_index.pkl"
document_texts_path = "document_texts.pkl"

document_texts = []

if os.path.exists(index_path) and os.path.exists(document_texts_path):
    try:
        with open(index_path, "rb") as f:
            index = pickle.load(f)
            print("Loaded FAISS index from faiss_index.pkl")
        with open(document_texts_path, "rb") as f:
            document_texts = pickle.load(f)
            print("Loaded document texts from document_texts.pkl")
    except Exception as e:
        print(f"Error loading FAISS index or document texts: {e}")
else:
    # Create a new FAISS index if it doesn't exist
    index = faiss.IndexFlatL2(embedding_model.get_sentence_embedding_dimension())
    with open(index_path, "wb") as f:
        pickle.dump(index, f)
        print("Created new FAISS index and saved to faiss_index.pkl")

def preprocess_text(text):
    # Add more preprocessing steps if necessary
    return text.strip()

def upload_files(files):
    global index, document_texts
    try:
        for file in files:
            file_path = file.name  # Get the file path from the NamedString object
            if file_path.endswith('.docx'):
                text = extract_text_from_docx(file_path)

                # Process the text and update FAISS index
                sentences = text.split("\n")
                sentences = [preprocess_text(sentence) for sentence in sentences if sentence.strip()]
                embeddings = embedding_model.encode(sentences)
                index.add(np.array(embeddings))
                document_texts.extend(sentences)  # Store sentences for retrieval

        # Save the updated index and documents
        with open(index_path, "wb") as f:
            pickle.dump(index, f)
            print("Saved updated FAISS index to faiss_index.pkl")
        with open(document_texts_path, "wb") as f:
            pickle.dump(document_texts, f)
            print("Saved updated document texts to document_texts.pkl")
        
        return "Files processed successfully"
    except Exception as e:
        print(f"Error processing files: {e}")
        return f"Error processing files: {e}"

def query_text(text):
    try:
        # Encode the query text
        query_embedding = embedding_model.encode([text])
        
        # Search the FAISS index
        D, I = index.search(np.array(query_embedding), k=5)
        
        top_documents = []
        for idx in I[0]:
            if idx != -1 and idx < len(document_texts):  # Ensure that a valid index is found
                top_documents.append(document_texts[idx])  # Append the actual sentences for the response
        
        # Prepare the prompt
        context = "\n".join(top_documents)
        prompt = f"Context:\n{context}\n\nQuestion:\n{text}\n\nAnswer:\n"

        # Query the LLM
        response = llm(prompt)
        return response
    except Exception as e:
        print(f"Error querying text: {e}")
        return f"Error querying text: {e}"

# Sample Gradio integration (for illustration)
import gradio as gr

def main():
    gr.Interface(
        [upload_files, query_text],
        ["files", "text"],
        ["text", "text"],
        title="Document Upload and Query System",
        description="Upload DOCX files to build an index, then query for answers based on uploaded documents.",
    ).launch()

if __name__ == "__main__":
    main()