Hugging Face Tool Use
Add web search tools to a GPT-OSS 120B agent on Hugging Face.
"""
Huggingface Tool Use
====================
Cookbook example for `huggingface/tool_use.py`.
"""
from agno.agent import Agent
from agno.models.huggingface import HuggingFace
from agno.tools.websearch import WebSearchTools
# ---------------------------------------------------------------------------
# Create Agent
# ---------------------------------------------------------------------------
agent = Agent(
model=HuggingFace(id="openai/gpt-oss-120b"),
tools=[WebSearchTools()],
markdown=True,
)
# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------
if __name__ == "__main__":
# --- Sync ---
agent.print_response("What is the latest news on AI?")
# --- Sync + Streaming ---
agent.print_response("What is the latest news on AI?", stream=True)The current Hugging Face streaming parser can omit additional tool calls when a single streamed delta contains more than one call. Use the non-streaming invocation for this example when all requested tools must execute.
Run the Example
Set up your virtual environment
uv venv --python 3.12
source .venv/bin/activateInstall dependencies
uv pip install -U agno ddgs huggingface-hubExport environment variables
export HF_TOKEN="your_hf_token_here"Use the non-streaming tool path
Save the source as tool_use.py and remove stream=True from the second print_response call. Both calls will then use the non-streaming path.
Run the example
Save the code above as tool_use.py, then run:
python tool_use.pyFull source: cookbook/90_models/huggingface/tool_use.py