What I do
- •Define logging conventions that enable fast feedback loops.
- •Specify log locations, formats, and minimum content.
- •Connect logs to Memory Bank updates and reports.
Feedback loop principles
- •Logs are the primary evidence of system behavior.
- •Every meaningful action should leave a trace that can be reviewed later.
- •A log entry should answer: what happened, when, where, and with what result.
Common log locations
- •RAPID:
logs/(core runtime) + optional subfolders. - •DS:
logs/for training/evaluation metrics, run summaries, and errors. - •Session activity reports:
logs/activity/<Role>/(see apm-gov).
Log format (default)
code
[YYYY-MM-DD HH:MM:SS] [LEVEL] - Message
Templates
- •RAPID example:
references/LOGGING_RAPID_TMP.md - •DS example:
references/LOGGING_DS_TMP.md
RAPID logging requirements
- •Core runtime: write to
logs/. - •Log key events: start/stop, user actions, important decisions, errors.
- •Include identifiers when available (request id, task id, file names).
DS logging requirements
- •Training logs: metrics per epoch/iteration (loss, primary/secondary metrics).
- •Config snapshot: model params and data versions used.
- •Artifacts: where the model and reports were saved.
- •Resource use: basic timing and (if applicable) GPU usage notes.
Minimum logging checklist
- • Start/end of operation
- • Inputs and configuration (high level)
- • Result summary or metric
- • Error details with context (if any)
How logs feed Memory Bank
- •Summarize log outcomes in
memory-bank/STATE.md. - •If logs reveal a deviation from architecture, document it.
- •Use logs as evidence when writing reports.