app.py

import gradio as gr
import PyPDF2
import os
from openai import OpenAI
import sys # Import sys to write to stderr

# --- Configuration & Client Initialization ---
# It's good practice to check for the API key at startup.
# This provides a clear error if the secret isn't set in Hugging Face Spaces.
NEBIUS_API_KEY = os.getenv("NEBIUS_API_KEY")
if not NEBIUS_API_KEY:
    # Use a more descriptive error message.
    raise ValueError("API Key not found. Please set the NEBIUS_API_KEY secret in your Hugging Face Space settings.")

# Initialize the OpenAI client with your custom endpoint.
# Ensure this base_url is correct and publicly accessible.
client = OpenAI(
    base_url="https://api.studio.nebius.com/v1/",
    api_key=NEBIUS_API_KEY
)

# --- Core Functions ---

def extract_text_from_pdf(pdf_file):
    """Extracts text from an uploaded PDF file object."""
    # pdf_file is a temporary file object from Gradio.
    if not pdf_file:
        return ""
    try:
        reader = PyPDF2.PdfReader(pdf_file.name)
        text = ""
        # Extract text from all pages except the first one.
        for i, page in enumerate(reader.pages):
            if i == 0:
                continue
            page_text = page.extract_text()
            if page_text:
                text += page_text + "\n"
        return text
    except Exception as e:
        print(f"Error reading PDF: {e}", file=sys.stderr)
        return ""


def get_llm_answer(pdf_text, question, history):
    """
    Sends the context, history, and question to the LLM and returns the answer.
    """
    # Truncate the PDF text to avoid exceeding the model's context limit.
    context = pdf_text[:16000]

    # The system prompt guides the model's behavior.
    system_prompt = '''You are a helpful assistant who specializes in body composition, diet, and exercise.
    Answer questions based on the provided document. Encourage the user to seek a professional
    if they have serious concerns whenever appropriate.'''

    # Construct the message payload for the API.
    messages = [{"role": "system", "content": system_prompt}, 
                {"role": "user", "content": f"Use the following document to answer my question:\n\n{context}"}, 
                {"role":"user", "content": f"Question: {question}"}
    ]

    # Add the conversation history.
    if history:
        for msg in history:
            if msg["role"] in ["user", "assistant"]:
                messages.append(msg)
    # Add the new user question
    messages.append({"role": "user", "content": question})

    try:
        response = client.chat.completions.create(
            model="meta-llama/Meta-Llama-3.1-70B-Instruct",
            temperature=0.6,
            top_p=0.95,
            messages=messages
        )
        return response.choices[0].message.content
    except Exception as e:
        print(f"Error calling OpenAI API: {e}", file=sys.stderr)
        # Return a user-friendly error message.
        return "Sorry, I encountered an error while trying to generate a response. Please check the logs."


# --- Gradio Interface Logic ---

# Use a class to manage state (the extracted PDF text).
class PDFChatbot:
    def __init__(self):
        self.pdf_text = None
        self.pdf_filename = None

    def upload_pdf(self, pdf_file):
        if pdf_file is None:
            return "Status: No PDF uploaded."
        
        self.pdf_text = extract_text_from_pdf(pdf_file)
        self.pdf_filename = os.path.basename(pdf_file.name)

        if not self.pdf_text:
            return f"Status: Could not extract text from {self.pdf_filename}. It might be empty, scanned, or protected."
            
        return f"Status: Successfully processed {self.pdf_filename}. You can now ask questions."

    def chat(self, user_message, history):
        if self.pdf_text is None:
            # Add an instruction to the chatbot window if no PDF is uploaded.
            #history.append([user_message, "Please upload a PDF document first."])
            return "Please upload a PDF document first.", history
        
        if history is None:
            history = []
        context_history = [msg for msg in history if msg["role"] in ["user", "assistant"]]
        # Get the answer from the LLM.
        answer = get_llm_answer(self.pdf_text, user_message, context_history)
        
        # Append the user message and the assistant's answer to the history.
        history = history + [
            {"role": "user", "content": user_message }, 
            {"role": "assistant", "content": answer }
        ]
        
        # Return an empty string to clear the input textbox and the updated history.
        return "", history

# Instantiate the bot.
pdf_bot = PDFChatbot()

# Build the Gradio UI.
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("# Body Composition Agent\nUpload a document about your body composition and ask questions about its content.")
    
    with gr.Row():
        with gr.Column(scale=1):
            pdf_file = gr.File(label="Upload PDF", file_types=[".pdf"])
            upload_btn = gr.Button("Process PDF", variant="primary")
            upload_status = gr.Textbox(label="Status", interactive=False, value="Status: Waiting for PDF...")
        
        with gr.Column(scale=2):
            chatbot = gr.Chatbot(type="messages", label="Chat History", height=500)
            msg_textbox = gr.Textbox(label="Your Question:", interactive=True, placeholder="Type your question here...")
            # Clear button is useful for starting a new conversation.
            clear_btn = gr.ClearButton([msg_textbox, chatbot], value="Clear Chat")

    # Wire up the event listeners.
    upload_btn.click(
        fn=pdf_bot.upload_pdf,
        inputs=[pdf_file],
        outputs=[upload_status]
    )
    
    # Allow submitting questions with Enter key or the button.
    msg_textbox.submit(
        fn=pdf_bot.chat,
        inputs=[msg_textbox, chatbot],
        outputs=[msg_textbox, chatbot]
    )

# Launch the app.
demo.launch(debug=True) # Use debug=True to see errors in the console.