Every endpoint is public and CORS-enabled. No API key. Try any endpoint directly below.
https://sector-slow.fourieranalys.workers.dev. All responses are JSON. All endpoints are read-only except ?snapshot=1.All surfaced pairs with full statistical fields. This is the primary endpoint.
curl -s "https://sector-slow.fourieranalys.workers.dev/_correlation" | jq '.pairs_count, .pairs_evaluated'Fields: a, b, r, n, p_value, p_bh, bh_significant, perm_p, ci_low, ci_high, ci_width, provenance, stale_data, verdict, tier, contradiction. Full definitions in methodology.
Stability tracking over the last N snapshots. Shows which pairs persist across days.
Additional fields: days_present, stability_pct, r_mean, r_min, r_max, pareto. Also returns drifts[] (Frobenius distance between consecutive snapshots) and top-level transitivity.
Manually write today's snapshot to _corr_<date>. Idempotent — skips if already written. The cron writes it automatically at 03:57 UTC, so this is only for manual triggers.
Current value and delta for one sector.
Returns delta (current, previous, pct_change, actual_window_days), context (history_points), sources (healthy/failed), and a pre-formatted summary string.
Full time-series for one sector. This is what the correlation engine reads.
Returns metric, points, first, last, and history[] array of {t, v} — Unix-millisecond timestamp and value.
Basic health check. Returns {ok: true} if the worker is up.
Deep health check: cron status, calibration status, shard coverage, audit findings.
Key fields: checks.cron.shards_today.{seen,expected}, checks.cron.last_run, checks.calibration, checks.sectors.shards_merged.
Pipeline health: how many sectors are tracked, healthy, failing.
| Field | Meaning |
|---|---|
r | Pearson correlation on detrended residuals |
n | Common daily observations (>= 10 required) |
p_bh | Benjamini-Hochberg adjusted p-value (q=0.05, m≈5000) |
bh_significant | Survives multiple-testing correction |
perm_p | Permutation p-value (2000 shuffles, assumption-free) |
ci_width | Fisher 95% CI width on r (higher = less confident) |
provenance | Two sources of same underlying metric |
stale_data | Artifact of batch-update counters |
verdict | confirm:provenance, surface, suppress:*, dispute:* |
tier | P (provenance) / A / B / C / D |
contradiction | Prior violated (only fires for known priors) |
stability_pct | Fraction of available days this pair appeared |
r_range | r_max - r_min across snapshots |
drifts[] | Frobenius distance between consecutive snapshots |
pareto | On frontier of (stability, |r|) — not dominated by any other pair |
No API key, no explicit rate limit. Please be considerate: this runs on Cloudflare Workers Free tier, which allows ~100k requests/day. A single page load of the dashboard makes ~2 requests. Reasonable for exploration and light automation.
All endpoints send Access-Control-Allow-Origin: *. You can call them from any origin, including a browser on another domain or a static page like this one.
The cron runs 5 shards between 03:17 and 03:57 UTC daily. Data from /history and /signal refreshes nightly. The /_correlation endpoint computes on-demand from current KV state, so it reflects the most recent history points (typically up to yesterday).