scaffold-service

---

name: scaffold-service description: This skill should be used when the user asks to "scaffold a GIS ingestion service", "create a GIS upload API", "build a FastAPI service for shapefile/DWG/.gdb ingestion", "generate a geospatial ingestion microservice", or invokes /gis-to-db:scaffold-service. Produces a complete, production-grade FastAPI service that accepts GIS/CAD file uploads, parses them via fiona/ezdxf/LibreDWG, reprojects to a target SRID, and writes geometries to PostGIS, MongoDB, or MySQL spatial. Includes API key auth, Pytest tests, Arq async job queue, Prometheus metrics, structured logging, and Docker compose wiring. argument-hint: [--name service-name] [--db-target postgis|mongo|mysql] [--out-dir ./services] [--target-srid 4326] allowed-tools: Bash, Read, Write, Edit, Glob

Scaffold a full FastAPI GIS ingestion service

v0.1 status — design doc only. Pre-built templates land in v0.2 (Slice B). In v0.1, Claude generates files inline from the specs in this SKILL.md — expect 25+ generated files, ~3000-5000 lines. Generation is slower and less consistent than template-based v0.2 will be. For production use today, prefer make-cli (lighter) or convert (one-shot).

This skill generates a complete, runnable FastAPI service for ingesting GIS and CAD files into a spatial database. The output is production-grade — auth, tests, async jobs, observability, and Docker — not a hello-world.

When to Use

Invoke when the user wants a standalone service they can deploy. Typical phrasings:

• /gis-to-db:scaffold-service --name parcels-ingest --db-target postgis
• "Scaffold a FastAPI service that ingests shapefiles into PostGIS."
• "Create a microservice for uploading DWG files and storing parcels."
• "Build a geospatial ingestion API with async jobs."

What It Generates

Directory structure under <out-dir>/<service-name>/:

<service-name>/
├── app/
│   ├── main.py                  # FastAPI app, /upload, /jobs/{id}, /metrics, /health
│   ├── auth.py                  # X-API-Key dependency
│   ├── settings.py              # pydantic-settings, reads from .env
│   ├── db/
│   │   ├── postgis.py           # asyncpg + SQLAlchemy GeoAlchemy2 (if --db-target=postgis)
│   │   ├── mongo.py             # motor (if --db-target=mongo)
│   │   └── mysql.py             # aiomysql (if --db-target=mysql)
│   ├── parsers/
│   │   ├── shapefile.py         # fiona-based reader
│   │   ├── geodatabase.py       # fiona FileGDB driver
│   │   ├── dxf.py               # ezdxf entity-to-geometry mapper
│   │   └── dwg.py               # subprocess LibreDWG dwg2dxf → dxf.py
│   ├── jobs/
│   │   ├── worker.py            # Arq worker
│   │   └── tasks.py             # parse_and_ingest task
│   ├── observability.py         # structlog + Prometheus instrumentation
│   └── schemas.py               # Pydantic models
├── tests/
│   ├── conftest.py
│   ├── data/                    # tiny sample .shp, .geojson fixtures
│   ├── test_upload.py
│   ├── test_parsers.py
│   └── test_integration.py
├── Dockerfile                   # python:3.12-slim + GDAL + LibreDWG
├── docker-compose.yml           # api, worker, redis, db
├── pyproject.toml               # uv/poetry compatible
├── .env.example
├── .dockerignore
├── README.md
└── Makefile                     # make dev / make test / make build

How It Works

1. Validator agent runs first. Use the Task tool to invoke gis-preflight-validator with mode=scaffold-service. The validator confirms the chosen DB target is reachable (if URI is set in settings) and that Docker is available for the build step.
2. Confirm parameters with the user. Service name, DB target, target SRID, out-dir. Read defaults from .claude/gis-to-db.local.md. Ask only for missing values.
3. Copy templates from assets/templates/service/. Templates use Jinja-style {{ var }} placeholders that the skill substitutes for service name, DB target, target SRID, etc.
4. Filter generated files by DB target. Only emit db/<target>.py, parsers/ adapters for the chosen target, and the relevant docker-compose.yml service stanza.
5. Write all files via the Write tool. Show a tree of what was created.
6. Offer to run. Print the commands to start the service (cd <out-dir>/<service-name> && make dev), but do not run automatically — the user reviews first.

Template Variables

Templates expect:

• SERVICE_NAME — kebab-case, used for directory and docker-compose service names.
• DB_TARGET — postgis | mongo | mysql.
• TARGET_SRID — int, default 4326.
• PYTHON_VERSION — default 3.12.
• INPUT_FORMATS — list, default ["shapefile", "geodatabase", "dxf", "dwg"].

Implementation Reference

• assets/templates/service/ — the Jinja-style template tree mirrored above. Treat this as the source of truth for what gets generated.
• references/db-target-matrix.md — exactly which files differ between postgis / mongo / mysql modes and the spatial schema each emits.
• references/arq-pattern.md — how the Arq worker is wired (entry point, redis URL, task signatures).
• references/observability.md — structlog config, Prometheus metric names, /metrics endpoint contract.
• examples/postgis-output/ — full snapshot of a generated service for PostGIS to verify against.

Important Constraints

• Never overwrite files in the user's filesystem without confirmation. If <out-dir>/<service-name>/ exists, ask before proceeding.
• Always include a working .env.example and a one-line make dev quickstart in the generated README.
• Generated code must pass ruff check and mypy --strict out of the box. Templates are kept clean.
• All file I/O uses async paths — no blocking calls in request handlers.