PointSav GIS engine — PointSav Documentation

A location-intelligence platform that depends on a central database and a live network is a platform a customer rents, not owns — outages, per-seat cloud cost, and air-gap ineligibility follow. The PointSav GIS Engine runs offline-first from flat files: it reads from a static PMTiles archive on the customer's own filesystem, renders interactively through MapLibre GL JS in the browser, and serves every query without an external dependency. For a regulated buyer, the spatial record is auditable, portable, and never leaves the building.

[edit]Architectural Principles

The engine operates as a stateless application surface, decoupling the data layer from the runtime so either can be updated or replaced independently.

[edit]Flat-File Substrate

Unlike centralised GIS stacks that require persistent database management, the PointSav engine uses a flat-file substrate. It consumes geographic data directly from JSONL, GeoParquet, and YAML formats versioned within a Totebox Archive. This architecture ensures the data layer remains entirely decoupled from the application logic, eliminating database maintenance overhead and preventing vendor lock-in.

[edit]Sovereign Rendering Stack

The platform avoids commercial SaaS mapping dependencies by using an open-source rendering stack:

PMTiles: A single-file archive format for tiled data that enables maps to be served directly from standard web servers (Nginx) or blob storage without a dedicated tile server. ^{[pmtiles-spec]}
MapLibre GL JS: A WebGL-based library for rendering interactive vector maps in the browser. ^{[maplibre-gl-js]}
Tippecanoe: A tool used to compile massive flat-file datasets into optimized vector tiles, ensuring rapid delivery of complex co-location clusters. [tippecanoe-tool]

[edit]Spatial Processing and Orchestration

The engine's core logic resides in the app-orchestration-gis service. This component executes the Woodfine co-location methodology deterministically:

Ingestion: Reads retail and civic infrastructure records from the Totebox Archive via service-business-clustering and service-places-filtering.
Analysis: Executes spatial joins and proximity queries to identify co-location clusters across 1.0 km, 3.0 km, and 5.0 km radii.
Ranking: Applies the 12-rank named-anchor matrix to generate site quality tiers.
Serialization: Outputs the processed results as tiled data for the visual interface at gis.woodfinegroup.com.

This stateless approach ensures that the entire GIS environment can be re-provisioned instantly from the immutable data layer, providing maximum service resilience and auditability.

[edit]See also

co-location-methodology — the ranking methodology that drives tier assignment in the GIS engine
app-orchestration-gis — the orchestration layer that runs the spatial analysis pipeline
service-business-clustering — retail clustering service feeding the GIS tier computation
service-places-filtering — civic infrastructure filtering service feeding the GIS tier computation
service-fs-data-lake — the flat-file data lake that backs all GIS source data
GIS as BIM Substrate — how the GIS engine functions as a BIM data layer
build-a-colocation-map — step-by-step guide: run a co-location cluster analysis and render the results on a map

[edit]References

Geographic information system — Wikipedia, accessed 2026-06-14