Retriever is a typed runtime for robotics pipelines.
Retriever lets teams author typed flows in Python, debug them locally with
Pipeline.step()
and carry the same graph into multiprocessing or dora-backed execution.
This homepage stays intentionally short: install, verify fit, then continue on
Start
or the integrated docs.
Treat openretriever.org as the open-source front door, openretriever-docs.pages.dev as the core docs, and retriever-space.pages.dev as the Golden examples gallery.
Retriever exposes flow I/O explicitly so interfaces are inspectable before deployment.
The docs keep local stepping and breakpoint-friendly debugging near the front door.
The public repository and license are part of the credibility story, not hidden in footnotes.
Install first
Keep the first command visible
curl -fsSL https://pixi.sh/install.sh | bash
pixi run demo-webcam-detection
# fallback if your environment needs it
python -m pip install -e ".[demo,dora,recording]"Choose the next route
Home stays compact. Depth lives elsewhere.
Use the Pixi-first setup path, with pip or uv fallback documented in the same guide.
Use the dedicated onboarding page for the longer builder path, route selection, and next steps.
Browse the integrated docs front door for tutorials, runtime guides, architecture, handbook, and API.
Keep the domain and docs split legible.
Docs and onboarding stay integrated with the main site flow here. Showcase and media live on the docs and Golden examples surfaces, rather than competing entry points with the canonical entry point.
Install-first project front door, quick start path, docs, GitHub, and community links.
MkDocs site for installation, tutorials, runtime concepts, API, and release notes.
Examples, tutorials, and integration demos live on the companion Golden surface.