v1.0.0

VrForaging 1.0.0 release

This tag marks the first stable release of the experimental workflow. Major versions are used to track dataset compatibility. Within a given major version, we may continue to evolve the experimental workflow, experiment control, UI, and curricula, but try our best to keep the data contract backwards-compatible.

Major features

• Updated against the latest versions of aind-behavior-services and clabe.
• Introduced script-driven workflows, including task launch, calibration, and session recovery.
• Refactored infrastructure to use external schema-generator references instead of package-duplicated schemas.
• Added saturation visualizer support across camera visualizers.
• Added support for odor mixtures across multiple olfactometers.
• Added Olfactometer extension support to the data contract and QC pipeline.
• Added support for terminating patches after N reward sites via OnSiteCountPatchTerminator.
• Migrated docs from Sphinx to MkDocs.
• Added delayed experiment start and stop support.
• Moved curricula into the mono-repo and streamlined project structure and launcher behavior.

Improvements

• Surfaced annotations in QC reports.
• Enforced rig_name and computer_name from environment variables.
• Improved metadata mapping, including olfactometer stimulus mapping.
• Improved photodiode event detection and aligned QC tests.
• Refactored ValidateOlfactometerDefinition to return Unit in Process.
• Updated schema folder layout and refactored Bonsai directory placement.
• Refactored QA startup ordering to run before task engine startup.
• Reduced QC stack-trace verbosity.
• Refactored and stabilized visualizer architecture, including saturation slider and velocity limits.
• Added commit hash encoding to metadata.
• Added waterlog integration to the experiment script.
• Added flexible tracking for external curricula repositories.
• Added support for automatic dataset version inference.
• Added curriculum README rendering.

Breaking changes

• Removed max_open_time from task logic.
• Rig schema is not backwards compatible in this release due to upstream simplification in aind-behavior-services. (See #499 for migration instructions.)
• Curriculum state used for sessions is now serialized to trainer_state.json as per aind-file-standards.
• odor_specification is now represented as List[float].
• task_logic schema backward deserialization support was explicitly extended for versions >= 0.6.

Bug fixes

• Fixed lens models in instrument mapping.
• Fixed Pandas 3.x related QC issues by avoiding in-place mutation of read-only dataframes.
• Fixed deployment script issues.
• Fixed native assertion and startup-related native errors in visualizers.
• Added graceful pre-exit error handling.
• Generalized environment sampling and implemented sequential sampling.
• Fixed task logic and bounds for CTCM algorithm.
• Applied rig-test driven bug fixes.

What's Changed

• Move curricula to mono repo by @bruno-f-cruz in #519
• Add curriculum metadata to aind-data-schema data mapper by @bruno-f-cruz in #518
• Release v1 by @bruno-f-cruz in #476
• Update schema folder by @bruno-f-cruz in #520
• Refactor QA to run prior to task engine startup by @bruno-f-cruz in #522
• Ensure curriculum does not override root logger during import by @bruno-f-cruz in #524
• Small bug fixes from tests at the rig by @bruno-f-cruz in #525
• Refactor visualizers by @bruno-f-cruz in #527
• Encode commit hash in metadata by @bruno-f-cruz in #526
• Reduce verbosity level for stack traces in quality-control by @bruno-f-cruz in #530
• Refactor ImGui visualizers to prevent native errors at startup by @bruno-f-cruz in #531
• Refactor visualizer to include saturation slider and limits for velocity by @bruno-f-cruz in #534
• Fix native assertion from visualizers by @bruno-f-cruz in #535
• Refactor bonsai directory to ./bonsai by @bruno-f-cruz in #539
• Handle errors gracefully before exiting by @bruno-f-cruz in #540
• Add waterlog to aind experiment script by @bruno-f-cruz in #541
• Allow flexible tracking of external curricula repositories by @bruno-f-cruz in #542
• Ensure patch reward available depletes on reward delivery by @bruno-f-cruz in #544
• Generalize environment sampling and implement sequential sampling by @bruno-f-cruz in #547
• Refactor available reward to not deplete implicitly by @bruno-f-cruz in #548
• Add support for automatically inferring dataset version by @bruno-f-cruz in #550
• Ensure previous task logic schema (>=0.6) is backwards deserializable by @bruno-f-cruz in #551
• Add deterministic reversals curriculum by @tiffanyona in #545
• Render curriculum readme by @bruno-f-cruz in #556
• Fix task logic and bounds for Ctcm algorithm by @bruno-f-cruz in #553
• Replace single_site_matching with single_site by @galenlynch in #552
• Add "Learning Sets" example by @bruno-f-cruz in #557

New Contributors

• @galenlynch made their first contribution in #552

Full Changelog: v0.6.8...v1.0.0