Skip to content

INSTRUCTIONS.md

Latest commit

 

History

History
62 lines (53 loc) · 2.17 KB

INSTRUCTIONS.md

File metadata and controls

62 lines (53 loc) · 2.17 KB

INSTRUCTIONS.md

This file is for protocol users testing tasks on the new framework.

What Changed

  • Task runtime now uses a structured event object internally (BehaviorEvent) with wall-clock timestamps.
  • Log output format is still events.jsonl with timestamp, event_type, and payload.
  • User/project config should live under users/<your_namespace>/.

What You Need To Do

  1. Update your checkout.
  2. Work on your own branch.
  3. Use your namespace files under users/<your_namespace>/.
  4. Run your task through run_task.py or your wrapper.
  5. Check outputs in .task_runs/<run_id>/.
  6. Report bugs with run artifacts attached (run_metadata.json, events.jsonl, result.json, quality_report.json).

Recommended Workflow

  1. cd /Users/lukesjulson/codex/RPi4_refactor/targets/RPi4_behavior_tasks
  2. git checkout main
  3. git pull origin main
  4. git checkout -b <your_name>/task-test-YYYYMMDD
  5. Edit users/<your_namespace>/mouse_info/<mouse>.json.
  6. Edit users/<your_namespace>/session_templates/<template>.json.
  7. Run task.
  8. Validate output and behavior.

Example (Julia/Duy Go/No-Go)

cd /Users/lukesjulson/codex/RPi4_refactor/targets/RPi4_behavior_tasks
git checkout main
git pull origin main
git checkout -b julia_duy/test-gonogo-20260222

python users/julia_duy/wrappers/run_gonogo_julia_duy.py \
  --yes \
  --output-dir .task_runs \
  --set trial_count=20 \
  --set seed=123

After the run:

  1. Find the run id printed in terminal output.
  2. Inspect .task_runs/<run_id>/events.jsonl.
  3. Confirm event records have wall-clock timestamps and expected event sequence.
  4. If needed, run: python tools/verify_run_artifacts.py .task_runs/<run_id>.

Example (Matt Context)

cd /Users/lukesjulson/codex/RPi4_refactor/targets/RPi4_behavior_tasks
python users/matt_context/wrappers/run_context_matt.py \
  --yes \
  --output-dir .task_runs \
  --set trial_count=30 \
  --set seed=456

Notes

  • Experimental protocols are blocked unless you pass --allow-experimental.
  • Production mode is stricter: use --run-mode production.
  • Optional stricter lock for production: --require-release-tag.
  • Do not commit per-session outputs from .task_runs/.