PointSav Documentation
PointSav builds operating systems and services for regulated businesses that need to own their data, their AI, and their record-keeping outright. The platform runs on customer hardware, produces [[worm-ledger-design|continuous-disclosure-grade records]] by structure, and operates fully without AI for buyers that require an air-gap.
This wiki is the engineering library for that platform, maintained against an internal editorial standard. It documents the architecture, the services, the operating systems, the governance commitments, and the design rationale that binds future development. It is written for institutional readers — auditors, technical due-diligence reviewers, and procurement evaluators — and for the engineers who build on or extend the platform. Where the engineering monorepo holds the code, this wiki holds the reasoning.
Forward-looking statements in this wiki carry planned, intended, or target language per the internal editorial standard.
Browse by area
How-to guides
All 26 guides →Step-by-step operational runbooks — how to install, configure, and run the platform.
- Add a node to a running fleet
Adds a second node to an already-running PPN fleet using service-vm-host's real env-var configuration — the same mechanism as the first node, since nothing about enrollment changes once a fleet exists.
- Build a co-location map
Renders tier-coloured co-location cluster markers in MapLibre GL by loading a PMTiles archive directly — the real flat-file architecture, since no bearer-token REST cluster API exists.
- Close out a construction project
Runs the two job-completion report binaries — a weighted six-category operational scorecard and a job-completion checklist and review pack — either of which can also be run as a live status check on a project still under construction.
- Configure a tenant namespace
Configures a tenant namespace on service-vm-tenant via environment variables and a restart — the real config-driven mechanism, since no runtime tenant-registration API exists.
- Connect to the OSM data pipeline
Ingests a new retail or service chain from OpenStreetMap using the real ingest-osm.py script and taxonomy.py's CATEGORIES/BRAND_FILL dicts, then rebuilds the servable cluster tiles.
- Explore the console for the first time
Orients a first-time operator to os-console — the status bar, the F9 inference-gateway dashboard, and the mandatory F12 input checkpoint that writes to the WORM ledger.
- Export structured data from the platform
Exports platform data through three real paths — DataGraph entity records via MCP tools, wiki Markdown read directly from git, and paginated ledger entries over service-fs's HTTP API.
- Federate archives via content mounts
Federates a second knowledge instance's articles into a running instance through a knowledge.toml [[mount]] entry — a flat, merged namespace with no isolation, not a URL-prefixed federation scheme.