High-performance
live weather, in Python.

Pull the current METAR reading, the multi-model forecast trajectory to settlement, and the model that's calling it, fetched on demand from AWC, IEM and Open-Meteo, normalized in your own runtime. Cached locally. TypeScript at parity.

Read the docs

✓ ~5-min refresh ✓ 36 forecast models ✓ CORS-open in the browser

live.py

1from mostlyright import weather
2
3# live obs + forecast trajectory, cached local
4df = weather.pairs(station, today, today)

1import { dataset } from "mostlyright";
2
3// live obs + forecast trajectory, cached local
4const rows = await dataset(station, today, today);

// what's live

Everything you need to track a market in real time

Live METAR

Current observations from the AWC feed, reconciled into a single NOW reading per station: the same number the contract settles against.

36 forecast models

NCEP, ECMWF, DWD, Météo-France and more via Open-Meteo, each with a conservative issued_at so live and backtest never see a forecast before it existed.

Trajectory to settlement

Past observations, the live NOW point, and the forecast curve to the contract close, one continuous, settlement-aligned series.

// the trajectory

Watch the day bend toward the strike

Every reading is timestamped and unit-correct, US extremes in °F, international in °C, so the live curve lines up with the contract's buckets, not a converted approximation.

Split past vs forecast: solid history, dashed forecast, anchored at NOW.

Per-model switching: re-draw the forecast under any of the 36 models.

Stale-aware: readings dim past their freshness budget.

KNYC · today 81° forecast high

// one read API, every runtime

From a Python notebook to a single <script> tag

The same research() call ships in two runtimes. Python drops into notebooks, pandas pipelines, and agent tools; the TypeScript build runs as an ES module in Node, inside browser and Web Workers, on Cloudflare/Bun edge, or as one IIFE tag. CSP-clean, no eval.

Browser integration →

index.html

1<!-- one tag (bundle from your access distribution) -->
2<script src="/vendor/mostlyright.global.js"></script>
3<script>
4  dataset("KNYC", d0, d1)
5    .then(console.log);
6</script>

// go live

From install to live feed in three steps

Request access

Access is currently by request. Tell us about your use case and we will get your team set up with the SDK.

Call pairs()

Pass a station and a date window. First call warms the local cache; the rest are offline.

Render

Schema-versioned rows, JSON-serializable, drop into a chart, an agent tool, or a backtest.

High-performancelive weather, in Python.

Everything you need to track a market in real time

Watch the day bend toward the strike

From a Python notebook to a single <script> tag

From install to live feed in three steps

Go live with public weather in four lines.

High-performance
live weather, in Python.