SDK
mostlyrightmd is a local-first SDK for working with authoritative weather and climate observations — daily and hourly temperature, dewpoint, wind, precipitation, and more. It calls public APIs (AWC, IEM, GHCNh, NWS CLI) directly and writes a parquet cache to ~/.mostlyright/cache/. The SDK is currently in private beta.
The Python and TypeScript SDKs share the same surface: dataset(station, from_date, to_date) returns one row per local-standard-time day in the requested window, joining observations with the climate normal for that station and date. (research() is a fully working alias of dataset().)
Access
The SDK is in private beta. Request early access and we’ll send an invite with install instructions for Python and TypeScript.
Where to start
Section titled “Where to start”- Python quickstart — request access, then a first call in under 60 seconds.
- TypeScript quickstart — request access, then a first call in under 60 seconds.
Concepts
Section titled “Concepts”- Temporal safety —
KnowledgeViewandLeakageDetector. Train-time data is filtered to what was knowable at the as-of cutoff, and a separate audit path raises on leakage. - Source identity — the fused (source-blind) default vs a single-source
source=pin, plusSourceMismatchErrorfor the case where training data and inference data come from different feeds.
Migration
Section titled “Migration”- → v1.13 —
dataset()as the primary composer,obs()per-report grain, standaloneclimate(), covariate columns, and the numbered-mode deprecation train. - Cache directory —
~/.tradewinds/to~/.mostlyright/and theTRADEWINDS_CACHE_DIRtoMOSTLYRIGHT_CACHE_DIRrename.
API reference
Section titled “API reference”The auto-generated reference lands under python/ and typescript/ once a non-rc tag fires the SDK’s docs-publish workflow. Until then these subtrees are empty placeholders.
python/— Sphinx + autodoc, regenerated on everyv*Python tag.typescript/— TypeDoc, regenerated on everyvts-*TS tag.