aind-dynamic-foraging

A repository for the Dynamic Foraging task.

---

📋 General instructions

This repository follows the project structure laid out in the Aind.Behavior.Services repository.

---

🔧 Prerequisites

Pre-requisites for running the project can be found here.

---

🚀 Deployment

For convenience, once third-party dependencies are installed, Bonsai and python virtual environments can be bootstrapped by running:

./scripts/deploy.ps1

from the root of the repository.

⚙️ Generating settings files

The Dynamic Foraging tasks is instantiated by a set of three settings files that strictly follow a DSL schema. These files are:

• task_logic.json
• rig.json
• session.json

Examples on how to generate these files can be found in the ./Examples directory of the repository. Once generated, these are the the only required inputs to run the Bonsai workflow in ./src/main.bonsai.

The workflow can thus be executed using the Bonsai CLI:

"./bonsai/bonsai.exe" "./src/main.bonsai" -p SessionPath=<path-to-session.json> -p RigPath=<path-to-rig.json> -p TaskLogicPath=<path-to-task_logic.json>

However, for a better experiment management user experience, it is recommended to use the provided experiment launcher below.

[> ] CLI tools

The platform exposes a few CLI tools to facilitate various tasks. Tools are available via:

uv run dynamic-foraging <subcommand>

for a list of all sub commands available:

uv run dynamic-foraging -h

You may need to install optional dependencies depending on the sub-commands you run.

🎮 Experiment launcher (temporarily CLABE)

To manage experiments and input files, this repository contains a launcher script that can be used to run the Dynamic Foraging task. A default script is located at ./scripts/aind-launcher.py. It can be run from the command line as follows:

uv run clabe run ./scripts/aind-launcher.py
# or uv run ./scripts/main.py

Additional arguments can be passed to the script as needed. For instance to allow the script to run with uncommitted changes in the repository, the --allow-dirty flag can be used:

uv run clabe run ./scripts/aind-launcher.py --allow-dirty

or via a ./local/clabe.yml file. Additional custom launcher scripts can be created and used as needed. See documentation in the clabe repository for more details.

🔍 Primary data quality-control

Once an experiment is collected, the primary data quality-control script can be run to check the data for issues. This script can be launcher using:

uv run dynamic-foraging data-qc <path-to-data-dir>

🔄 Regenerating schemas

DSL schemas can be modified in ./src/aind_behavior_dynamic_foraging/rig.py (or (...)/task_logic.py`).

Once modified, changes to the DSL must be propagated to json-schema and csharp API. This can be done by running:

uv run dynamic-foraging regenerate