app.py

import gradio as gr
from transformers import RagTokenizer, RagRetriever, RagSequenceForGeneration
import fitz  # PyMuPDF
from datasets import load_dataset
from llama_index.core import Document, VectorStoreIndex, StorageContext, load_index_from_storage, Settings
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.llms.ollama import Ollama

# Load Llama 3 model components
tokenizer = RagTokenizer.from_pretrained("facebook/rag-sequence-nq")
retriever = RagRetriever.from_pretrained("facebook/rag-sequence-nq", index_name="custom", passages_path="my_knowledge_base.faiss")
model = RagSequenceForGeneration.from_pretrained("facebook/rag-sequence-nq", retriever=retriever)

# Load the embedding model
embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")

# Create an LLM object using the deployed Llama3 Ollama instance
llm = Ollama(model="llama3:instruct", request_timeout=60.0)

# Set global settings for the LLM, chunk size, and embedding model
Settings.llm = llm
Settings.chunk_size = 512
Settings.embed_model = embed_model

# Function to extract text from PDFs
def extract_text_from_pdf(pdf_files):
    texts = []
    for pdf in pdf_files:
        doc = fitz.open(pdf.name)
        text = ""
        for page in doc:
            text += page.get_text()
        texts.append(text)
    return texts

# Function to provide answers based on questions and PDFs
def rag_answer(question, pdf_files):
    texts = extract_text_from_pdf(pdf_files)
    context = " ".join(texts)
    inputs = tokenizer(question, return_tensors="pt")
    outputs = model.generate(**inputs, context_input=context)
    return tokenizer.batch_decode(outputs, skip_special_tokens=True)[0]

# Function to create the Vector Store Index from documents
def create_vector_store_index(documents):
    index = VectorStoreIndex.from_documents(documents)
    index.storage_context.persist(persist_dir="pdf_docs")
    return index

# Load dataset and convert to Document format
pdf_docs = load_dataset('your-dataset-name', split='train')  # Replace with your actual dataset name
documents = [Document(text=row['text'], metadata={'title': row['title']}) for index, row in pdf_docs.iterrows()]

# Create or load the vector store index
try:
    storage_context = StorageContext.from_defaults(persist_dir="pdf_docs")
    vector_index = load_index_from_storage(storage_context)
except:
    vector_index = create_vector_store_index(documents)

# Define the query engine powered by the Vector Store
query_engine = vector_index.as_query_engine(similarity_top_k=10)

# Functions for Gradio UI
def query(text):
    z = query_engine.query(text)
    return z

def interface(text):
    z = query(text)
    response = z.response
    return response

# Gradio interface
with gr.Blocks(theme=gr.themes.Glass().set(block_title_text_color="black", body_background_fill="black", input_background_fill="black", body_text_color="white")) as demo:
    gr.Markdown("h1 {text-align: center;display: block;}Information Custodian Chat Agent")
    with gr.Row():
        output_text = gr.Textbox(lines=20)
    with gr.Row():
        input_text = gr.Textbox(label='Enter your query here')
    input_text.submit(fn=interface, inputs=input_text, outputs=output_text)

demo.launch(share=True)