npm search discoverability, as measured

Declaring keywords is not discoverability. This page records where the packages this repository publishes actually rank in npm registry search for the discovery terms the package metadata claims — including the queries where they do not rank at all, which is the half a keyword list can never tell you.

Every row below is projected from a dated record in docs/data/npm-search-observations.v1.json, validated against schemas/npm-search-observations.v1.json. Each observation must report every tracked package, found or not, so a query where Honua does not appear cannot be dropped from the page. Records are immutable per observation: re-observing a query adds a record and the page projects the newest one, so a rank that got worse stays in the file as provenance.

How the numbers were taken

• Endpoint: https://registry.npmjs.org/-/v1/search — the registry itself, not a mirror or a third-party rank tracker.
• Page size: 20 results, matching npmjs.com's own paging. "Page 1" below means the first 20 ranked results.
• Scan depth: 250 ranked results per query. A package reported as "not in the top 250" is exactly that claim and no stronger.
• Reproduce:

npm run docs:npm-search:observe   # re-query the live registry and rewrite the records
npm run docs:npm-search           # regenerate this page from the committed records

Search ranking is a moving, unowned target: it depends on npm's own scoring, on download counts, and on what else was published that week. Treat every row as an observation with a date on it, not as a property of the package.

One consequence worth stating plainly: these ranks reflect the metadata as published, not the metadata in the working tree. Improving a package's keywords cannot move its rank until the next release carries them to the registry, so a row here can lag a merged fix by a whole release cycle.

Tracked packages

• @honua/sdk-js
• @honua/mcp-server
• create-honua-app

Observed rankings

What each query is for

• maplibre arcgis migration (tracked target) — the exact success-metric query declared on #499 ("npm search for 'maplibre arcgis migration' surfaces @honua/sdk-js on page one"). Re-observe by 2026-11-25.
• arcgis migration — the arcgis-migration keyword's primary discovery term (REQ-004). Re-observe by 2026-11-25.
• maplibre gis sdk — the maplibre discovery term as a MapLibre developer shopping for a data client would type it (REQ-004). Re-observe by 2026-11-25.
• ogc api features client — the ogc-api discovery term (REQ-004). Re-observe by 2026-11-25.
• stac client typescript — the stac discovery term (REQ-004). Re-observe by 2026-11-25.
• geocoding maplibre — the geocoding discovery term (REQ-004). Re-observe by 2026-11-25.
• mcp server geospatial — the discovery path to @honua/mcp-server for an agent developer. Re-observe by 2026-11-25.

Tracked targets

• maplibre arcgis migration → @honua/sdk-js on page 1: not met — page 5 (rank 86) as observed 2026-08-25.

A target that is not met is reported here rather than removed. These are tracked, non-gating metrics on #499: CI fails when this page drifts from its records, never because a rank is disappointing — the SDK does not own npm's ranking function.

What this page does not claim

• Not a download count. Rank is not adoption; the download baseline is tracked separately on #499.
• Not a competitor comparison. Which packages outrank Honua for a term is not a product claim; that comparison lives in docs/comparison.md under its own evidence contract.
• Not stable. Nothing here is a guarantee about the next query you run; rerun the observe command.

The declared keywords and descriptions these queries test are gated separately by npm run verify:discoverability, which covers every package this repository publishes.

Ecosystem directory entries and the external submission ledger live in maplibre-plugin-directory.md.