Skip to content

:>> ABIO → ABIO Ref → ABIO Data

ABIO Data

What the data/ folder actually holds, and why nothing is filed there on purpose.

Rewritten 2026-10-07 (T067). The page used to describe an intent-based tree — chem/kegg1/ for a reusable chemistry, world/simple1/chem/ for a world's own, test/T1/world/chem/ for a test's — with the rule that a top-level category names the thing's primary purpose. None of it was ever built, and the design it belonged to (the M1 scenario runtime, its spec language and the do-system that resolved those names) was deleted in M47.7 and T056.

What is there now

data/ is gitignored scratch: the default dat folder dvc_dat writes into when something creates a real Dat rather than a MockDat (ABIO DAT). On this machine it is ~10 MB of data/anonymous/Dat_<n>/ folders — 2,662 of them, each a _spec_.yaml left by a test that did not name its dat. Nothing reads them back, nothing is durable, and deleting the folder costs nothing.

The outputs that are meant to be read each have their own place, all gitignored:

  • runs/<name>/ — one experiment run: records.jsonl, manifest.json, report.txt, key.png / key.json. Written by bio suite run, read back by bio suite resume|aggregate|report.
  • reports/ — bio report's page: report.md, report.html, junit.xml.
  • art/ — dvc-dat 3.0's artifact store, beside the dat folder.

Where the durable things live instead

Nothing a later run depends on is in data/. The inputs are checked in: catalog/experiments/*.yaml (the thirteen zeros), catalog/examples/*/ (the nine coverage worlds), catalog/registrations.yaml (the licences), and the goldens pinned in tests/suite/test_golden_experiments.py. That is the whole contract — a clone with an empty data/ reproduces every zero.

See Also

  • ABIO DAT — the dvc-dat surface alienbio actually uses
  • ~~ABIO Files~~ — the full repository layout