stac-fastapi-geoparquet

A stac-fastapi with a stac-geoparquet backend.

stac-fastapi-geoparquet can serve a full-featured STAC API from one or more stac-geoparquet files located (e.g.) in blob storage — no database required.

Warning

👷 This project is under active development and may change and break at any time.

Usage

To start a STAC API server pointing to a single stac-geoparquet file:

$ python -m pip install 'stac-fastapi-geoparquet[serve]'
$ STAC_FASTAPI_GEOPARQUET_HREF=data/naip.parquet uvicorn stac_fastapi.geoparquet.main:app
INFO:     Started server process [47920]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)

This will start the server on http://127.0.0.1:8000. The collection will be auto-generated from the items in the stac-geoparquet file.

Using collections

Instead of providing the href to a single file, you can provide the href to a file containing a JSON list of collections. Any collection assets with a application/vnd.apache.parquet type field will be added to the server as sources of items. For an example, see data/collections.json.

To start a server with one or more collections:

$ STAC_FASTAPI_COLLECTIONS_HREF=data/collections.json uvicorn stac_fastapi.geoparquet.main:app
INFO:     Started server process [47920]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://127.0.0.1:8000 (Press CTRL+C to quit)

To auto-generate the collections file, we provide a script:

scripts/generate-collections s3://my-bucket/a.parquet s3://my-bucket/b.parquet

This will update ./data/collections.json.

Limitations

• Currently, only supports one collection per file (tracking issue: #27)

Development

Get uv, then:

git clone [email protected]:stac-utils/stac-fastapi-geoparquet.git
cd stac-fastapi-geoparquet
uv sync

To run the tests:

uv run pytest