Lmstudio Image Agent

Send image bytes to a llama3.2-vision model in LM Studio and stream the description.

image_agent.py
"""
Lmstudio Image Agent
====================

Cookbook example for `lmstudio/image_agent.py`.
"""

import httpx
from agno.agent import Agent
from agno.media import Image
from agno.models.lmstudio import LMStudio

# ---------------------------------------------------------------------------
# Create Agent
# ---------------------------------------------------------------------------

agent = Agent(
    model=LMStudio(id="llama3.2-vision"),
    markdown=True,
)

response = httpx.get(
    "https://upload.wikimedia.org/wikipedia/commons/0/0c/GoldenGateBridge-001.jpg"
)

agent.print_response(
    "Tell me about this image",
    images=[Image(content=response.content)],
    stream=True,
)

# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------

if __name__ == "__main__":
    pass

Run the Example

Set up your virtual environment

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

Install dependencies

uv pip install -U agno openai

Prepare LM Studio

Load llama3.2-vision in LM Studio and start its local server at http://127.0.0.1:1234/v1.

Run the example

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

python image_agent.py

Full source: cookbook/90_models/lmstudio/image_agent.py