Skip to content
Hosted service in public preview. 3.0 installs from GitHub until it is on PyPI.

Storage and disk

The store root is init(dir=...), else AGENTICML_DIR, else ~/.agenticml.

<root>/
objects/<hh>/<sha256> content-addressed bytes, shared by every run
runs/<run_id>/
meta.json project, config, status, summary, env, git
events.jsonl config updates, code, artifacts, finish
notes.md run note (Markdown, up to 256 KiB)
sync.json upload cursor per server and workspace, after a sync
scalars/<id>.bin 8 bytes per point: uint32 step, float32 value
scalars/index.json
media/index.json

Run ids look like 20261027T021500-ab12cd34 (UTC). Each scalar point takes 8 bytes. Series longer than 65,536 points also get a small overview file at finish, so charts load fast.

Code snapshots and media are stored once, keyed by SHA-256, so the same bytes logged twice share one object. Snapshot caps default to 10 MB per file and 100 MB total. Artifacts are stored as a path, size, and mtime unless you pass copy=True.

The full on-disk format is documented in FORMAT.md.

Terminal window
agenticml du
agenticml gc --dry-run
agenticml gc --older-than DAYS --keep-last N
agenticml gc --include-unsynced
agenticml sync --all --prune
  • agenticml du reports how much space the store uses.
  • agenticml gc removes runs. Only synced runs are eligible unless you pass --include-unsynced. --dry-run prints the plan without deleting.
  • agenticml sync --prune drops raw series files that already have an overview. A pruned run cannot be resumed or synced to another host.

agenticml warns when the store grows past AGENTICML_MAX_BYTES (default 5 GiB).