Program closed-loop robot behavior in Python.

Build perception, planning, memory, and control as reusable Python modules, or load ready-made pieces from Retriever Hub. Compose them like PyTorch modules; Retriever schedules each part independently at the behavior level.

InstallPython 3.11+

$ pip install retriever-core

For the camera demo: pip install "retriever-core[demo]".

Run camera demo Read the docs

AI agent?Read /llms.txt →

hub_webcam.py

from retriever import Latest, Pipeline, Rate, Trigger, hub

Camera = hub.use("openretriever/webcam-demo:CameraSource")
Detect = hub.use("openretriever/webcam-demo:ColorDetector")
View = hub.use("openretriever/webcam-demo:RerunDisplay")

with Pipeline("webcam") as pipe:
    camera = Camera() @ Rate(hz=20)
    detector = Detect() @ Trigger("image")
    viewer = View() @ Rate(hz=20)
    camera.then(detector, sync=Latest())
    detector.then(viewer, sync=Latest())

pipe.run(duration=60)

Import Hub Flows

Connect with .then()

Run locally with Rerun