ascii-city-engine

ASCII City Engine

Build a portable city experience in three layers: an engine, a city-provider contract, and a city pack. Keep city-specific facts out of engine code.

Workflow

1. Define the pack boundary and local meter-based CRS.
2. Acquire elevation, building, and road/path data from public sources.
3. Verify downloads, licenses, units, and provenance before conversion.
4. Reproject all geometry to local meters and emit the manifest plus world tiles.
5. Validate the pack offline:

   python3 scripts/validate-city-pack.py path/to/city-pack
   

6. Load the pack's first tile in assets/ascii-city-engine.html and test movement, grade following, and solid footprints.
7. Record data limitations and human acceptance evidence.

Required invariants

• Compute feet_z from terrain and eye_z from feet plus eye height.
• Reject null terrain, over-steep steps, and movement touching a solid footprint.
• Use deterministic building colors and clear missed rays every frame.
• Treat v1 as one ground height per (x, y) column; reserve, but do not implement, a surface graph.
• Keep the engine independent of any city, vendor, agent harness, or private infrastructure.
• Do not commit large or full-resolution source data. Commit only a small redistributable sample; document acquisition for the rest.

Reference routing

• Read references/engine-architecture.md for world math, rendering, collisions, and the scaffold contract.
• Read references/city-provider-contract.md when creating or validating a provider.
• Read references/gis-ingestion.md before downloading or converting GIS sources.
• Read references/raleigh-poc.md for the reproducible Raleigh proof of concept.
• Use templates/city-pack-manifest.schema.json and templates/world.schema.json as authoritative data contracts.

Available Scripts

This skill bundles one script; there are no others to discover.

Script	Purpose	Invocation
scripts/validate-city-pack.py	Offline validator for a city pack: checks the manifest against the schema contract and verifies world tiles. Run it at workflow step 5, after emitting the manifest and tiles and before loading anything in the engine — do not proceed to in-engine testing until it passes.	python3 scripts/validate-city-pack.py path/to/city-pack

Prerequisites

• Python 3 with standard library only; the validator requires no third-party packages.
• A city pack produced through the ingestion workflow: manifest plus world tiles reprojected to local meters (see references/gis-ingestion.md).
• The engine itself is a standalone HTML file (assets/ascii-city-engine.html) that runs in any modern browser; no build step or server is required.

Limitations

• v1 supports one ground height per (x, y) column: no interiors, bridges with traversable space below, tunnels, or multi-level geometry (a surface graph is reserved but not implemented).
• The engine is software-rendered CPU canvas — no WebGL/GPU rendering or mobile controls.
• The skill does not include full-resolution GIS archives; packs are built reproducibly from public sources, committing only small redistributable samples.
• The validator is offline and structural: passing it does not substitute for human movement testing in the engine (workflow step 6).

Boundaries

Use a different skill or implementation approach for WebGL/GPU rendering, mobile controls, combat, interiors, bridges with traversable space below, tunnels, or general-purpose geospatial analysis. If asked to specialize this skill for one agent runtime, preserve the portable core and put runtime integration outside this skill. If asked to commit large GIS binaries, refuse and provide reproducible download/conversion commands instead.