Reverse-geocode a fleet

Telemetry volume is set by your ping interval and your device count, so a fleet produces reverse lookups at a rate with no relationship to revenue. Priced per lookup, the bill starts shaping the telemetry design instead of following it: sample the pings, batchbatch sizeHow many examples the model processes before each parameter update. Larger batches give smoother gradients but cost more memory; gradient accumulation simulates a big batch on a small GPU. them overnight, or drop the lookup and store a raw coordinate that no report can read.

What changes

Reverse geocodinggeocodingThe process of converting an address into geographic coordinates (latitude and longitude). Mailwoman geocodes in a multi-tier cascade: exact address-point match → street interpolation → locality centroid. Each tier is progressively coarser but more widely available. becomes a local read. WOFReverseGeocoder pulls every place whose bounding box contains the point from an R*Tree index, smallest area first, then ray-casts each candidate's real boundary against a polygon sidecar; the first polygon containing the point wins. No socket opens, so the rate is bounded by disk and CPU rather than by a quota, and reverseGeocodeSync does the same work on the calling threadthreadA parallel workstream within a release. Threads compose; they are not sequential milestones like phases. for an ingest path that is already synchronous.

Every result carries a containment field reading polygon or approximate, which is how you tell a boundary-confirmed answer from a nearest-centroid one before anything downstream acts on it.

What you still carry

Three limits decide whether this fits, and all three arrive before the first lookup.

The database is a local build. Reverse geocodinggeocodingThe process of converting an address into geographic coordinates (latitude and longitude). Mailwoman geocodes in a multi-tier cascade: exact address-point match → street interpolation → locality centroid. Each tier is progressively coarser but more widely available. needs a full WOFWOF (Who's On First). An open-source gazetteer of places maintained by Mapzen/whosonfirst. Mailwoman builds a custom SQLite database from WOF GeoJSON repos, extended with postcode data, importance scores, and coincident-role relations. distribution carrying the place_bbox index — a different database from the candidate gazetteergazetteerA geographical index that maps place names and postcodes to real-world coordinates. Mailwoman uses a custom-built Who's On First (WOF) SQLite database as its gazetteer — the 'atlas' half of the grammar/atlas architecture. forward geocodinggeocodingThe process of converting an address into geographic coordinates (latitude and longitude). Mailwoman geocodes in a multi-tier cascade: exact address-point match → street interpolation → locality centroid. Each tier is progressively coarser but more widely available. reads, and not one of the four published bundlesevidence bundleThe pair of retrieval-augmented input channels (street-type + locality-surface) that feed lexicon membership as soft per-token evidence alongside the text. Shipped in 6.7.0; trained natively from step 0 in the from-scratch base line.. mailwoman data pull ships candidate, us, fr and poi, and none of them is it. You build it with mailwoman gazetteer build admin, which costs hours of wall clock and tens of gigabytes of source data.

The polygon sidecar is a second build, and its coveragecoverageThe fraction of a population or region for which a data source has real, non-placeholder entries — e.g. 47% rooftop coverage on Texas addresses. Distinct from accuracy on the rows that are present. is per-countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head.. mailwoman gazetteer polygons produces it. Without it every result reads approximate by construction, because no point-in-polygon test is attempted; with it, only the countries in that sidecar get boundary confirmation. So a fleet running outside those countries reads approximate on every row, and the containment discriminator above stops discriminating. Decide the sidecar's countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head. list from where your vehicles are, before you build it.

There is no address-level reverse. The result is an administrative chain — neighborhood, localitylocalityThe city / town / settlement component of an address: a populated place sitting between region and neighborhood in the hierarchy., county, regionregionThe first-level administrative subdivision of a country — a US state, a French region, a province. The component between country and locality., countrycountryThe top-level address component (an ISO country). Closed-vocabulary, so it is best handled by a deterministic matcher feeding a proposal rather than a retrained model head. — and never a house numberhouse numberThe numeric or alphanumeric identifier of a building on a street. Mailwoman's house_number component; its position relative to the street name flips between locales.. Having the US rooftoprooftopGeocoding precision at the building or parcel level — coordinates within a few meters — the highest tier of the geocode cascade. Sourced from address-point and situs data. shardsshardA partial output file of the corpus build, written in Parquet format. The training pipeline streams shards row by row. on disk does not change that, because the reverse path never consults an address-point database.

Reverse geocode a coordinate carries these with their transcripts, including the failure worth designing for: a coordinate outside every boundary you hold polygons for still gets an answer, so threshold distanceKm whenever containment reads approximate.

Next

• Your first ten minutes — the install and the gazetteergazetteerA geographical index that maps place names and postcodes to real-world coordinates. Mailwoman uses a custom-built Who's On First (WOF) SQLite database as its gazetteer — the 'atlas' half of the grammar/atlas architecture. pull, which come before the admin build.
• Pricing — the same monthly figure at any ping interval.