app.py

import os
import shutil
from getpass import getpass

import gradio as gr
from llama_index.llms.openai import OpenAI
from llama_index.embeddings.openai import OpenAIEmbedding
from llama_index.core import Settings, SimpleDirectoryReader, VectorStoreIndex, StorageContext
from llama_index.vector_stores.qdrant import QdrantVectorStore
from llama_index.core.memory import ChatMemoryBuffer
import qdrant_client

# Set your OpenAI API key from environment variable
openai_api_key = os.getenv('OPENAI_API_KEY')
if not openai_api_key:
    raise ValueError("Please set your OPENAI_API_KEY environment variable.")

# Define a system prompt as a global constant
SYSTEM_PROMPT = (
    "You are an AI assistant who answers the user questions, "
    "use the schema fields to generate appropriate and valid json queries"
)

# Configure the LLM and embedding models
Settings.llm = OpenAI(model="gpt-3.5-turbo", temperature=0.4)
Settings.embed_model = OpenAIEmbedding(model="text-embedding-ada-002")

# Load initial documents from a directory called "new_file"
documents = SimpleDirectoryReader("new_file").load_data()

# Set up the Qdrant vector store (using an in-memory collection for simplicity)
client = qdrant_client.QdrantClient(location=":memory:")
vector_store = QdrantVectorStore(
    collection_name="paper",
    client=client,
    enable_hybrid=True,
    batch_size=20,
)
storage_context = StorageContext.from_defaults(vector_store=vector_store)

# Build the initial index and query/chat engines
index = VectorStoreIndex.from_documents(documents, storage_context=storage_context)
chat_memory = ChatMemoryBuffer.from_defaults(token_limit=3000)
chat_engine = index.as_chat_engine(
    chat_mode="context",
    memory=chat_memory,
    system_prompt=SYSTEM_PROMPT,
)

def process_uploaded_file(uploaded_file):
    """
    Process the uploaded file:
      1. Save the file to an "uploads" folder.
      2. Copy it to a temporary folder ("temp_upload") to load using SimpleDirectoryReader.
      3. Extend the global documents list and rebuild the index and chat engine.
    """
    if uploaded_file is None:
        return "No file uploaded."

    # 'uploaded_file' is a temporary file path provided by Gradio.
    file_name = os.path.basename(uploaded_file)
    uploads_dir = "uploads"
    os.makedirs(uploads_dir, exist_ok=True)
    dest_path = os.path.join(uploads_dir, file_name)
    shutil.copy(uploaded_file, dest_path)

    # Prepare a temporary directory to read the file
    temp_dir = "temp_upload"
    os.makedirs(temp_dir, exist_ok=True)
    # Clear previous files in temp_dir (optional, to avoid mixing files)
    for f in os.listdir(temp_dir):
        os.remove(os.path.join(temp_dir, f))
    shutil.copy(dest_path, temp_dir)

    # Load the new document(s) from the temporary folder
    new_docs = SimpleDirectoryReader(temp_dir).load_data()

    # Update the global documents list and rebuild the index and chat engine
    global documents, index, chat_engine
    documents.extend(new_docs)
    index = VectorStoreIndex.from_documents(documents, storage_context=storage_context)
    chat_engine = index.as_chat_engine(
        chat_mode="context",
        memory=chat_memory,
        system_prompt=SYSTEM_PROMPT,
    )

    return f"File '{file_name}' processed and added to index."

def chat_with_ai(user_input, chat_history):
    """
    Send the user input to the chat engine and update the conversation history.
    """
    response = chat_engine.chat(user_input)
    # Collect reference filenames from the response (if available)
    references = response.source_nodes
    ref = []
    for node in references:
        if "file_name" in node.metadata and node.metadata["file_name"] not in ref:
            ref.append(node.metadata["file_name"])
    # Create a complete response string with references if present
    complete_response = str(response)
    if ref:
        complete_response += "\n\nReferences: " + ", ".join(ref)
    chat_history.append((user_input, complete_response))
    return chat_history, ""

def clear_history():
    return [], ""

def gradio_chatbot():
    """
    Create a Gradio interface with two tabs:
      - "Chat" for interacting with the chat engine.
      - "Upload" for uploading new files to update the index.
    """
    with gr.Blocks() as demo:
        gr.Markdown("# Chat Interface for LlamaIndex with File Upload")
        
        with gr.Tab("Chat"):
            chatbot = gr.Chatbot(label="LlamaIndex Chatbot")
            user_input = gr.Textbox(
                placeholder="Ask a question...", label="Enter your question"
            )
            submit_button = gr.Button("Send")
            btn_clear = gr.Button("Delete Context")
            chat_history = gr.State([])
            submit_button.click(chat_with_ai, inputs=[user_input, chat_history],
                                  outputs=[chatbot, user_input])
            user_input.submit(chat_with_ai, inputs=[user_input, chat_history],
                              outputs=[chatbot, user_input])
            btn_clear.click(fn=clear_history, outputs=[chatbot, user_input])
        
        with gr.Tab("Upload"):
            gr.Markdown("### Upload a file to add its content to the index")
            file_upload = gr.File(label="Choose a file")
            upload_button = gr.Button("Upload and Process")
            upload_status = gr.Textbox(label="Upload Status")
            upload_button.click(process_uploaded_file, inputs=[file_upload], outputs=[upload_status])
    
    return demo

if __name__ == "__main__":
    gradio_chatbot().launch(debug=True)