Python API

The stable API (from ntpstats import api as nt) lists the names covered by the deprecation policy, in one namespace. The module pages below document everything, including the provisional parts.

from ntpstats import load, load_one
from ntpstats.stability import series_stability
from ntpstats.analysis import summary, compare

s = load_one("/var/log/chrony/measurements.log", peer="192.0.2.10")
print(summary(s)["range_90"])
for r in series_stability(s, kinds=("oadev", "tdev"), ci=0.95):
    print(r.kind, r.taus, r.dev, r.lo, r.hi, r.alpha)

A TimeSeries has t (POSIX seconds), offset (seconds, reference − local), extra (per-sample columns such as delay) and meta.

Module	Contents
ntpstats.series	TimeSeries
ntpstats.parsers	load, load_one, per-format parsers
ntpstats.stability	estimators, confidence intervals, noise ID, dynamic
ntpstats.edf	equivalent degrees of freedom
ntpstats.analysis	summary, detrending, outliers, compare
ntpstats.network	delay floor, wedge, FPP, min-delay filter
ntpstats.filters	Kalman filter and RTS smoother
ntpstats.estimators	estimator API and reference algorithms
ntpstats.simulate	clocks, paths, scenarios with ground truth
ntpstats.bench	benchmark runner
ntpstats.trace, ntpstats.ptpsim	delay traces and replay; PTP servos and boundary-clock chains
ntpstats.sntp, ntpstats.nts	measurement clients (NTPv4, RFC 9769 interleaved, NTPv5, NTS, NTS pools)
ntpstats.roughtime	Roughtime client, chained measurements, malfeasance reports
ntpstats.spectrum, ntpstats.noisefit	PSDs, L(f); power-law noise fit h_α
ntpstats.hat, ntpstats.holdover	N-cornered hat / Groslambert covariance; holdover prediction
ntpstats.research	CGGTTS, RINEX clock, Circular T, RIPE Atlas, NTP Pool, the interop dataset
ntpstats.otlp	OpenTelemetry (OTLP/HTTP JSON) export
ntpstats.plugins	plugin entry points and discovery
ntpstats.adapters	pandas, xarray, Parquet/Arrow
ntpstats.api, ntpstats.stream, ntpstats.deprecation	stable surface, large files, deprecation helpers
ntpstats.masks, ntpstats.report	masks, HTML reports