Status and reliability

Everything below is checked when this page loads, or is a standing commitment you can hold against a delivered report. There is no uptime percentage here, and the reason is the first section.

Live checks

Check	State	Detail
Application database	OK	Responding
Published metro dataset	OK	45 metros, built 23 August 2026
Drying model in service	OK	v2.3.0

Checked 5 September 2026 05:07 UTC, at the moment this page rendered. Machine-readable: /status?format=json

Why there is no uptime figure

Availability here is not monitored. No external monitor watches this service, there is no incident log, and there is no historical record to average. So a “99.9%” on this page would be a number chosen because it reads well, not one that was measured.

That matters more than it might sound. This service sells documents into contract disputes, where the whole value is that every figure traces to something checkable. A page whose job is trust is the worst possible place to publish the one number nobody can verify.

What matters for this product is different anyway. A claim report is produced by a person, reviewed, and delivered the same business day. Whether the web server had a bad twenty minutes in March has no bearing on whether the analysis holds up in front of a resident engineer. Freshness of the data and stability of the method do.

Where the numbers come from

Input	Source	Named on every report?
Weather	NOAA Global Historical Climatology Network – Daily (GHCND), US Government public domain	Yes — station ID, distance from site, and record completeness
Soil	USDA NRCS Soil Data Access, queried at the jobsite coordinates	Yes — map unit name and hydrologic group
Geocoding	US Census Bureau ZCTA Gazetteer, with operator coordinate override	Yes

Method version history

Version	Released	What changed
v2.3.0	2026-07-30	v2.3.0 - Sprinkle exclusion no longer reports a workable day while the drying clock is running (breaks parity with the VibeSprings TypeScript port, which remains on v2.2.1). Abnormality moved from mean-exceedance to mean + 1 standard deviation, with the count of more-adverse baseline periods reported as the headline. Baseline coverage gate raised 50% to 80%. Year-crossing baseline fetch extended into the following year; Feb 29 clamped in non-leap baseline years. Exhibit precipitation gap-corrected to match delay days. Chart bands derived from the local distribution rather than fixed inches. Effective drying constant and NRCS hydrologic group now disclosed on both deliverables.
v2.2.1	2025-12-29	Linear Measurement Evolution: Replaced version 2.2.0's tiered seasonal steps and rain triggers with smooth linear interpolation for thermodynamic drying coefficients. Integrated specific USDA-aligned base soil values (A=0, B=1.5, C=2.5, D=3.5) and refined Freeze-Thaw hysteresis logic for more accurate thaw-weakening prediction. | 2026-07: Platform unification — engine extracted to shared DryingModel; mud map, regional grid, and public API v1 now run this exact engine (previously four divergent variants). Estimate output unchanged (verified by 43,200-day differential parity test). Also corrected this row's soil_factors, which misrecorded clay/silt as 3.5/2.5 (those are NRCS groups D/C); the engine has always used 3.0/2.0.
v2.2.0	2025-12-10	Peer Review Implementation: (1) CRITICAL: Thaw Reset Trigger - frozen soil resets drying clock on thaw due to thaw weakening. (2) Seasonal Drying Coefficients - Winter 0.5x, Spring/Fall 0.75x, Summer 1.0x. (3) Tiered Rain Intensity - Light rain (<0.5") resets 50%, Heavy rain (>1.5") resets 150%. (4) Updated soil classifications - Clay 3 days, Silt 2 days per pumping risk assessment.
v1.0.0	2025-11-01	Original release. Basic soil drying formula without sponge factor, sprinkle check, or thermodynamic seasonality.
v2.1.0	2024-12-01	Initial tracked version. Includes: Unified Soil Scientist Formula with sponge factor, sprinkle check, thermodynamic seasonality, granular vs cohesive logic.

A version number appears on every report, so a document issued last year can be matched to the method that produced it. The coefficients themselves ship in the methodology appendix included with a purchased report, alongside the specific numbers they produced.

What this service refuses to do

These are constraints in the code, not aspirations. They are the reason a report can be handed to an opposing party.

• It will not publish a baseline assembled from too few usable years. A thin record is refused rather than reported with a caveat.
• It will not treat an unobserved day as a dry one. Days on which the station reported nothing are excluded from both the count and the denominator, and the drying model holds its state until real observations resume.
• It will not always find a claim. When a disputed period falls inside the normal range, the report says so and prints the distribution supporting that conclusion. A negative finding is delivered as readily as a favourable one.
• It will not treat exceeding an average as evidence. A mean is exceeded in roughly half of all periods by construction. The abnormality threshold is set above it, at the baseline mean plus one standard deviation.
• If station coverage at a site is too thin to produce a defensible document, the report is refunded.

What is not claimed

• No government certification. No agency certifies these reports. Ilystics is not affiliated with NOAA, the USDA, or the AIA.
• Not legal advice. This service does not determine contractual entitlement, causation, or damages. The analysis supports a claim; the claimant and their advisors make it.
• Not validated against contractor daily logs. The model is calibrated and benchmarked against published state contract weather tables, and where it disagrees with them that is published rather than hidden. It has not been checked against superintendent daily logs from live jobsites, and saying otherwise would overstate it.