app.py

import os
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
import gradio as gr
import fitz  # PyMuPDF
import easyocr
from PIL import Image
from sentence_transformers import SentenceTransformer
from chromadb import Client, Settings

# Load Zephyr 7B (fine-tuned for chat)
zephyr_tokenizer = AutoTokenizer.from_pretrained("HuggingFaceH4/zephyr-7b-alpha")
zephyr_model = AutoModelForCausalLM.from_pretrained(
    "HuggingFaceH4/zephyr-7b-alpha",
    torch_dtype=torch.float16,  # Use half-precision for faster inference
    device_map="auto"  # Automatically loads the model on GPU if available
)

# Load a sentence transformer model for embeddings
embedding_model = SentenceTransformer("all-MiniLM-L6-v2")

# Initialize Chroma client for RAG
chroma_client = Client(Settings())
collection = chroma_client.create_collection(name="knowledge_base")

# Function to extract text from PDF
def extract_text_from_pdf(pdf_path):
    doc = fitz.open(pdf_path)
    text = ""
    for page in doc:
        text += page.get_text()
    return text

# Function to extract text from image
def extract_text_from_image(image_path):
    reader = easyocr.Reader(['en'])
    results = reader.readtext(image_path)
    extracted_text = " ".join([res[1] for res in results])
    return extracted_text

# Function to generate a response
def generate_response(prompt):
    # Structure the input prompt for chat
    formatted_prompt = f"<|user|>\n{prompt}\n<|assistant|>\n"
    
    # Tokenize the input prompt
    inputs = zephyr_tokenizer(formatted_prompt, return_tensors="pt").to(zephyr_model.device)
    
    # Generate the response
    outputs = zephyr_model.generate(**inputs, max_length=200)
    
    # Decode the response
    response = zephyr_tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    # Extract only the assistant's response
    response = response.split("<|assistant|>")[-1].strip()
    return response

# Function to add documents to the knowledge base
def add_to_knowledge_base(text_chunks):
    embeddings = embedding_model.encode(text_chunks)
    for idx, (chunk, embedding) in enumerate(zip(text_chunks, embeddings)):
        collection.add(
            documents=[chunk],
            embeddings=[embedding.tolist()],
            ids=[str(idx)]
        )

# Function to retrieve relevant chunks
def retrieve_relevant_chunks(query, top_k=3):
    query_embedding = embedding_model.encode(query)
    results = collection.query(
        query_embeddings=[query_embedding.tolist()],
        n_results=top_k
    )
    return results["documents"][0]

# Chatbot function to handle text, PDF, and image inputs
def chatbot(input_type, text_input, pdf_input, image_input):
    if input_type == "Text":
        if not text_input:
            return "Please enter some text."
        query = text_input
    elif input_type == "PDF":
        if pdf_input is None:
            return "Please upload a PDF file."
        pdf_text = extract_text_from_pdf(pdf_input)
        query = f"Extracted text from PDF:\n{pdf_text}\n\nQuestion: {text_input}"
    elif input_type == "Image":
        if image_input is None:
            return "Please upload an image file."
        image_text = extract_text_from_image(image_input)
        query = f"Extracted text from image:\n{image_text}\n\nQuestion: {text_input}"
    else:
        return "Invalid input type."

    # Retrieve relevant chunks from the knowledge base
    relevant_chunks = retrieve_relevant_chunks(query)
    context = "\n\n".join(relevant_chunks)

    # Generate response using the model
    prompt = f"Context:\n{context}\n\nQuestion: {query}\n\nAnswer:"
    response = generate_response(prompt)
    return response

# Gradio interface
input_components = [
    gr.Dropdown(choices=["Text", "PDF", "Image"], label="Input Type"),
    gr.Textbox(lines=2, placeholder="Enter text...", label="Text Input"),
    gr.File(label="Upload PDF", file_types=[".pdf"]),
    gr.Image(label="Upload Image", type="filepath")
]

# Create the Gradio interface
interface = gr.Interface(
    fn=chatbot,
    inputs=input_components,
    outputs="text",
    title="RAG Chatbot with PDF and Image Support",
    description="Select the input type (Text, PDF, or Image) and provide your input."
)

# Launch the app
interface.launch()