- Site config: site.example.json is committed; the local site.json (git-ignored) or
$JFJOCH_BATTERY_SITE / --site names this machine's paths and the private manifest.
Data roots are the arms' own directories, so copying them and editing the site
config moves the battery.
- Run states: manifest records the runner's pid; a run is complete, running, aborted
(Ctrl-C/SIGTERM, or `abort RUN`) or unfinished. compare and report refuse a run
that did not finish unless --allow-incomplete, and the report says so.
- `list` shows runs and their state; `run` defaults its label to the rugnux version
and its baseline to the site's persisted one; a baseline must be complete and of
the same privacy as the run; a private report cannot be written into the repo.
- XDS reference d_min: when the finest shell of XDS's table has no significant
CC1/2, XDS merged past its own signal (no resolution cut, or a cut at the detector
edge) and its limit is the detector's reach. The reference is then where XDS's
CC1/2 falls through 0.30 (rugnux's own target), and that is also the range forced
on rugnux. The manifest keeps XDS's limit (dmin_xds) and the rule (dmin_rule).
Forcing the old 1.08 A on the weak insulin set turned rugnux's I23 into I222
(ISa 12.5); forced at the derived 1.81 A it is I23 again (ISa 18.0). Changed in
the inhouse arm: insu_I_weak 1.08 -> 1.81, lyso_half_image 0.80 -> 2.60,
cytc_eiger 2.04 -> 2.27, lyso_strong 1.18 -> 1.24; every other inhouse limit stands.
- inhouse.json follows the directory rename (ids and inputs, data root
/home/data/inhouse); `remap` applies such a rename and keeps the old ids as
aliases, so runs from before it still compare set by set.
- Optional open-arm model check (model_check.py, when present): R-free of the
deposited model against each merge, recorded and plotted.
- results.json has a fixed, documented row schema (results_schema 1); the HTML
report gains inline-SVG plots (verdicts per arm, d_min and R-free ratios per set).
- README.md: arms, data layout and provenance, prerequisites, running, run
directory, results schema, report, comparing, protocol, adding datasets.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
One runner for the open (vs PDB deposition) and in-house (vs XDS) arms,
replacing the scattered scripts outside the repository. Each run writes a
new, read-only directory named <date>_<rugnux git>_<label> holding a copy
of the binary with its sha256 and build flags, manifest.json (inputs,
references, options, host, GPU sharing) and results.json in one schema
for all arms, plus per-set work directories and a Markdown and a
self-contained HTML report (per-arm and per-population pass rates,
resolution gain, ISa/R_meas distributions, timing, failures, delta vs a
baseline run).
- Dataset lists and references are data files: open.json (151 public
sets, deposited group/cell/resolution, pinned inputs) and inhouse.json
(standard test crystals and no-crystal controls, XDS references). The
private arm's manifest stays outside the repository; a private run gets
its own run directory and report.
- Scoring is setting-invariant: lattice by Niggli-reduced primitive cell
(beta vs 180-beta, C2 vs I2), symmetry via sgequiv on the open arm and
by point group against XDS, then merge usability. Re-scoring the rc169
and rc170 battery outputs reproduces their verdicts.
- Input discovery (`discover`) ranks sweeps by frame count, never by
file size or sort order; a run never searches, it runs what the
manifest names.
- `compare A B` flags per-set changes beyond noise; `--rerun-changed`
reruns them with A's saved binary to separate real changes from
run-to-run noise.
- `--tier smoke` selects a 22-set subset spanning the populations; a
full-arm run takes a lock; `--gpulock` queues on the shared GPU and
marks the timing as not a reference.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
leonarski_fandClaude Opus 5