Skip to content

Experiment repository catalog#

The repositories built on aind-behavior-services, surveyed on main. All pin the framework to the same 0.13.x line. Use this to pick the right reference implementation. See repo anatomy and behavior-vs-experiment for the patterns.

Single-unit repos (Aind.Behavior.*, Aind.Physiology.*)#

Repo PyPI / package Services pin Task logic? Maturity
Aind.Behavior.VrForaging aind-behavior-vr-foraging via workspace Yes Most mature / reference. uv workspace with a separate curricula package, data mappers, MkDocs, curriculum + RPC, richest Extensions/.
Aind.Behavior.Telekinesis aind-behavior-telekinesis >=0.13.5,<0.14 Yes Mature, clean, copier-templated. Best example of the current canonical single-task repo; uses the newer @experiment() decorator + clabe.ui.
Aind.Behavior.DynamicForaging aind-behavior-dynamic-foraging >=0.13.5 Yes Rich but structurally divergent: task_logic/ split into interventions/, trial_generators/, trial_models.py; uses scikit-learn; Sphinx docs. main checked out in a worktree.
Aind.Behavior.IsoForce aind-behavior-iso-force >=0.13.5 Yes Solid mid-tier; single package, versioned data_contract/v0_1_0.py. Version 0.1.0rc0 (earlier stage).
Aind.Behavior.JustFrames aind-behavior-just_frames >=0.13.6 No (rig-only-ish) Leaner, benchmark-oriented (online video acquisition/encoding). Defines AindJustFramesRig + SatelliteRig; schema under src/DataSchemas/.
Aind.Physiology.Fip aind-physiology-fip >=0.13.0 No (physiology) Mature physiology repo; rig-only, rich C# FIP operators + data mappers; schema under src/DataSchemas/.

Composition repos (Aind.Experiment.*)#

Repo Version Composes (submodules) Maturity
Aind.Experiment.VrForaging-Fip 0.2.0 Aind.Behavior.VrForaging + Aind.Physiology.Fip More developed: dual-Bonsai concurrent launch, FIP mapper, curriculum, QC.
Aind.Experiment.VrForaging-OpenEphys 0.1.0 + Aind.Physiology.OpenEphys Earlier: single-Bonsai launch, RPC client stubbed, third submodule declared but not yet wired; clabe pinned to a feature branch.

How to read the "task logic?" column#

A repo with both rig.py and task_logic.py is a behavior task. A repo with a Rig but no task_logic.py is usually a physiology/acquisition unit — a common indicator, not a requirement, as described in behavior-vs-experiment.

VR Foraging leads; the rest follow#

VR Foraging is the reference implementation: the most mature repo and the one where the framework's patterns are worked out first. The other behavior repos are expected to converge on the same conventions over time (workspace layout, curricula, data mappers, the newer clabe launcher API). Divergences noted in the table above are generally maturity gaps, not deliberate alternatives — when in doubt, follow what VR Foraging does.

  • Learning the canonical pattern → Aind.Behavior.Telekinesis (cleanest current single-task repo).
  • Seeing the full-power version / where the rest are heading → Aind.Behavior.VrForaging.
  • Seeing composition → Aind.Experiment.VrForaging-Fip.