Short, task-shaped guides to logging, finding, and exporting your runs. Pick a guide from the list, or search to jump straight to one.
Install the CLI, register a project, and log your first run in about two minutes. Everything stays on your machine, no account required.
You can record older runs retroactively, with whatever metadata you can still identify, and you do not need a config file to do it. Inline parameters work for new runs too.
For array jobs and logging after the fact: have your script drop a small params.json beside its outputs, then ingest the whole batch in one single-writer fieldnotes log --manifest call.
A SLURM or PBS array should never call fieldnotes log from the compute nodes. Have each task emit a manifest instead, then ingest the whole array in one single-writer call once the results land on your workstation.
Record which script produced which data, and which script turned that data into a figure, so you can trace any result back to the exact code that made it, even after the code changes.
Two free drop-in files teach any AI coding assistant to inject a manifest emitter into the scripts it writes, then log finished runs with fieldnotes log, no hand-editing required.
If you submit cluster jobs with ClusterPilot, it can log each finished job to your local Fieldnotes automatically, offline, with one setting. This is the fully automatic end of the logging spectrum.