app.py

import os
import gradio as gr
from llama_index import VectorStoreIndex, SimpleDirectoryReader, SummaryIndex
from llama_index.readers import SimpleWebPageReader
from llama_index.llms import MistralAI
from llama_index.embeddings import MistralAIEmbedding
from llama_index import ServiceContext
from llama_index.query_engine import RetrieverQueryEngine


description = "Example of an assistant with Gradio and Mistral AI via its API"
placeholder = "Ask a question"
placeholder_url = "Extract text from this url"
placeholder_api_key = "API key"

query_engine = None

with gr.Blocks() as demo:

    gr.Markdown(""" ### Welcome to Gaia Level 2 Demo 
    Add an URL and your API key at the bottom of the interface before interacting with the Chat. This demo allows you to interact with a webpage and then ask questions to Mistral APIs. Mistral will answer with the context extracted from the webpage.
    """)
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.ClearButton([msg, chatbot])

    with gr.Row():
        api_key_text_box = gr.Textbox(placeholder=placeholder_api_key, container=False, scale=7)

    def setup_with_url(url, api_key):
        global query_engine
        # Set-up clients
        llm = MistralAI(api_key=api_key,model="mistral-medium")
        embed_model = MistralAIEmbedding(model_name='mistral-embed', api_key=api_key)
        service_context = ServiceContext.from_defaults(chunk_size=1024, llm=llm, embed_model=embed_model)

        # Set-up db
        documents = SimpleWebPageReader(html_to_text=True).load_data([url])
        index = VectorStoreIndex.from_documents(documents, service_context=service_context)
        query_engine = index.as_query_engine(similarity_top_k=15)
        return "I'm ready, please add a question here."

    with gr.Row():
        url_msg = gr.Textbox(placeholder=placeholder_url, container=False, scale=7)
        url_btn = gr.Button(value="Set-up API and process url ✅", interactive=True)
        url_btn.click(setup_with_url, [url_msg, api_key_text_box], msg, show_progress= "full")

    

    def respond(message, chat_history):
        response = query_engine.query(message)
        chat_history.append((message, str(response)))
        return chat_history

    msg.submit(respond, [msg, chatbot], [chatbot])

demo.launch()