app.py

# Chat agent using LlamaIndex SimpleChatEngine + Gradio
from llama_index.core.chat_engine import SimpleChatEngine
from llama_index.core.memory import ChatMemoryBuffer
from llama_index.llms.huggingface_api import HuggingFaceInferenceAPI
import gradio as gr
import os
from dotenv import load_dotenv

load_dotenv()

hf_token = os.getenv("HF_TOKEN")

llm = HuggingFaceInferenceAPI(
    model_name="Qwen/Qwen2.5-Coder-32B-Instruct",
    temperature=0.7,
    max_tokens=1024,
    token=hf_token,
    provider="auto",
)

memory = ChatMemoryBuffer.from_defaults(token_limit=3000)
chat_engine = SimpleChatEngine.from_defaults(llm=llm, memory=memory)


def respond(message, history):
    response = chat_engine.chat(message)
    return str(response)


demo = gr.ChatInterface(respond)
demo.launch()