The PointSav GIS Engine is a high-performance, customer-owned location intelligence platform built in Rust for offline-first, flat-file operation — a structural departure from traditional geographic information systems that rely on centralised database instances and continuous network connectivity. The engine 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. The map data lives in the customer's archive; nothing leaves the deployment unless the operator explicitly chooses to publish.

Architectural Principles

The engine is engineered to operate as a stateless application surface, adhering to the PointSav principle of complete data sovereignty.

Flat-File Substrate

Unlike traditional GIS stacks (e.g., PostGIS, Esri) which 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.

Sovereign Rendering Stack

The platform avoids commercial SaaS mapping dependencies by using a high-performance, 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 with high client-side performance. [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]

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:

1. Ingestion: Reads retail and civic infrastructure records from the Totebox Archive via service-business-clustering and service-places-filtering.
2. Analysis: Executes spatial joins and proximity queries to identify co-location clusters across 1.0 km, 3.0 km, and 5.0 km radii.
3. Ranking: Applies the 12-rank named-anchor matrix to generate site quality tiers.
4. 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.

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