Patterns by sensor_name (a property, or a window redrawn from a DIFF image as the device's pattern editor does), re-measured with optical's emulator, camera correction as the cloud's process_shifts (one measurement at a time, or the series), written in the files of data/ (displacement/, camera/, targets.csv; images/, metrics/, weather/ linked to data_diff/), with run.yaml and diagnostics/ beside the stations. configs/cloud.yaml, empty, reproduces data/JUNGFRAU 2026-10-12..16: corr_x/y within 0.03 px, the same rejections, the camera pose within 5e-4 deg; 16 s. configs/jungfrau_example.yaml: PM_G back to 0.7, BL_C's window redrawn. .gitignore: data_diff/, runs/, remeasure/. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2.9 KiB
run_config Implementation Plan
For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (
- [ ]) syntax for tracking.
Goal: processing/run_config.sh <config> [SERIAL...] writes runs/<name>/<SERIAL>/ with the files of data/.
Architecture: one script; optical imported from $OPTICAL; re-measurement through reprocess_images.main;
correction through remove_camera_movement, per epoch by default.
Tech Stack: uv (numpy, pandas, pillow, pyyaml, scipy, scikit-image, opencv-python-headless).
Spec: docs/superpowers/specs/2026-11-19-run-config-design.md
Global Constraints
- Writes only under
runs/<name>/(guarded); readsdata_diff/, neverdata/except for the acceptance comparison. - Columns of
data/'s advanced CSV, in that order:time_utc, sensor_name, x, y, error_value, error_threshold, relative_x, relative_y, corr_x, corr_y, reproj_err, inlier, sigma0, pattern_type, peak, flux, sharpness, fwhm, dominance, snr, corr_x_mm, corr_y_mm. - Camera CSV columns:
time_utc, thenmedian_/min_/max_ofpan, tilt, scale, rotation, sigma0.
Review Focus
- A configuration with a typo (
error_treshold, unknown sensor): error before any work. - An epoch where only one reference survives: translation only, no camera row (as the cloud).
- A day without images in the range: no file for it.
- Re-running the same configuration: the station folder replaced, other stations untouched.
- optical's working tree dirty: recorded in
run.yaml.
Task 1: configuration and correction, with --check
Files: Create processing/run_config.py.
Interfaces: load_config(path) -> dict; camera_settings(config, station, unit_props) -> dict;
cloud_properties(sensors) -> {sid: props}; apply_patterns(props, overrides, images_dir) -> {sid: props};
correct(meas: DataFrame, props, cam) -> (advanced: DataFrame, camera: DataFrame).
check(): unknown top / station / camera / pattern key raises; unknown sensor_name raises; camera order station > common > unit > defaults; synthetic correction: three static references, a static target (corr ~0), a target above its threshold (corr empty, x kept), relative = x - pattern x, sigma0 on every row, one camera row per minute.- Implement;
--checkprints ok.
Task 2: the run, the cloud configuration, the acceptance
Files: Modify processing/run_config.py (main); create processing/run_config.sh, configs/cloud.yaml;
modify .gitignore (runs/), README.md.
main: stations, days, patterns,reprocess_images.main,correct, per-day files, links,run.yaml.run_config.sh cloud JUNGFRAU --from 2026-10-12 --to 2026-10-17;data.pyreadsruns/cloud/JUNGFRAU.- Compare with
data/JUNGFRAU: corr_x / corr_y (all, and the rejected-reference night of 10-15), camera pose. - Docs.