Tutorial Tracks¶
Retriever tutorials are organized as a learning path, not a dump of examples. Run the first lane end-to-end, then branch into the track that matches what you are building.
Recommended first hour
pixi run demo-webcam-detection— see a real timed graph.pixi run demo-basic-flow— learn the smallest API.pixi run demo-rt-execution— run through the runtime path.pixi run demo-stepper— debug the same model in-process.pixi run demo-webcam-record— record and replay an observed run.
Tracks¶
A. Flow FundamentalsTyped flows, clocks, adapters, and pipeline ergonomics.
B. IR and ExecutionValidation, IR, execution graphs, and backend runs.
C. Debug and ReplayStepper-first debugging, traces, MCAP, and replay drills.
D. Closed Loop StateStateful flows, feedback, authority, and replanning.
E. Resource and SyncMulti-rate streams, synchronization, fan-in/fan-out, and joins.
F. Policy BackendsSwapping policy backends behind one graph contract.
G. Operations InterfacesRegistries, composition, typed boundaries, and integration surfaces.
H. Evidence and ManifestsRun manifests, datasets, and acceptance evidence.
How To Choose¶
| If you want to... | Start with |
|---|---|
| Learn the programming model from scratch | Track A: Flow Fundamentals |
| Run the same graph on a backend | Track B: IR and Execution |
| Debug a failing robot run | Track C: Debug and Replay |
| Build feedback/stateful behavior | Track D: Closed Loop State |
| Handle multi-rate data handoff | Track E: Resource and Sync |
| Abstract over a policy/model backend | Track F: Policy Backends |
| Publish reusable components | Track G: Operations Interfaces |
| Capture run evidence and dataset lineage | Track H: Evidence and Manifests |