Basic Agent

The async examples share one event loop because the model caches its inference client. See the Hugging Face setup for token permissions and model-provider availability.

Code

import asyncio

from agno.agent import Agent
from agno.models.huggingface import HuggingFace

model = HuggingFace(
    id="mistralai/Mistral-7B-Instruct-v0.2", max_tokens=4096, temperature=0
)
agent = Agent(model=model)


async def async_examples():
    # Reuse and close the cached async client in one event loop.
    try:
        await agent.aprint_response(
            "What is meaning of life and then recommend 5 best books to read about it"
        )
        await agent.aprint_response(
            "What is meaning of life and then recommend 5 best books to read about it",
            stream=True,
        )
    finally:
        await model.get_async_client().close()


if __name__ == "__main__":
    # --- Sync ---
    agent.print_response(
        "What is meaning of life and then recommend 5 best books to read about it"
    )

    # --- Sync + Streaming ---
    agent.print_response(
        "What is meaning of life and then recommend 5 best books to read about it",
        stream=True,
    )

    # --- Async and Async + Streaming ---
    asyncio.run(async_examples())
    model.get_client().close()

Usage

Set up your virtual environment

uv venv --python 3.12
source .venv/bin/activate

Set your API key

export HF_TOKEN=xxx

Install dependencies

uv pip install -U huggingface_hub agno

Run Agent

Save the code above as huggingface_basic.py, then run:

python huggingface_basic.py