Example of eoAPI customization prepared for FOSS4GNA 2024.

---

Documentation: https://eoapi.dev/customization/

Source Code: https://github.com/developmentseed/eoapi-foss4gna

---

This repository shows an example of how users can build a business application based on eoAPI services, starting from eoapi-devseed.

Custom

Runtimes

business.logic

A FastAPI application backed by a Postgres database with functions for finding suitable parcels for natural capital projects, like forested areas that have experienced disturbance in recent years (e.g. wildfire, timber harvesting, etc).

• /parcels: POST GeoJSON features to the parcel table in the database
• /parcels/{id}/landcover_summary: Return the area covered by each landcover class in a given year for a parcel
• /map: Load a leaflet map with vector tile features from the parcel table. Optionally filter down to features that have experienced disturbance in forested areas over a specified time period.

eoapi.stac

Built on stac-fastapi.pgstac application, adding a TiTilerExtension and a simple Search Viewer.

When the EOAPI_STAC_TITILER_ENDPOINT environment variable is set (pointing to the raster application) and titiler extension is enabled, additional endpoints will be added to the stac-fastapi application (see: stac/extension.py):

• /collections/{collectionId}/items/{itemId}/tilejson.json: Return the raster tilejson for an item
• /collections/{collectionId}/items/{itemId}/viewer: Redirect to the raster viewer

eoapi.raster

The dynamic tiler deployed within eoapi-devseed is built on top of titiler-pgstac and pgstac. It enables large-scale mosaic based on the results of STAC search queries.

The service includes all the default endpoints from titiler-pgstac application and:

• /: a custom landing page with links to the different endpoints
• /mosaic/builder: a virtual mosaic builder UI that helps create and register STAC Search queries
• /collections: a secret (not in OpenAPI documentation) endpoint used in the mosaic-builder page
• /collections/{collection_id}/items/{item_id}/viewer: a simple STAC Item viewer

eoapi.vector

OGC Features and Tiles API built on top of tipg.

The API will look for tables in the database's public schema by default. We've also added three functions that connect to the pgSTAC schema:

• pg_temp.pgstac_collections_view: Simple function which returns PgSTAC Collections
• pg_temp.pgstac_hash: Return features for a specific searchId (hash)
• pg_temp.pgstac_hash_count: Return the number of items per geometry for a specific searchId (hash)

Infrastructure

The CDK code is almost similar to the one found in eoapi-template. We just added some configurations for our custom runtimes.

Local testing

Before deploying the application on the cloud, you can start by exploring it with a local Docker deployment

docker compose up

Once the applications are up, you'll need to add STAC Collections and Items to the PgSTAC database.

To load the Impact Observatory Landcover STAC collection and items to the local database:

# set env vars to point at the database in the docker network
export BUSINESS_API_ENDPOINT=http://localhost:8084
export PGUSER=username
export PGPASSWORD=password
export PGDATABASE=postgis
export PGHOST=localhost
export PGPORT=5439

After the STAC metadata are loaded up it is time to add some parcels to the database:

# bootstrap the database with some landcover class definitions
curl -X POST ${BUSINESS_API_ENDPOINT}/bootstrap-data | jq

# load a sample of parcel data from two counties in the US
scripts/load-parcel-data siskiyou 0-200
scripts/load-parcel-data st_louis 0-200

Deployment

Requirements

• python >=3.9
• docker
• node >=14
• AWS credentials environment variables configured to point to an account.
• Optional a config.yaml file to override the default deployment settings defined in config.py.

Installation

Install python dependencies with

python -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt

And node dependencies with

npm install

Verify that the cdk CLI is available. Since aws-cdk is installed as a local dependency, you can use the npx node package runner tool, that comes with npm.

npx cdk --version

First, synthesize the app

npx cdk synth --all

Then, deploy

npx cdk deploy --all --require-approval never

After deployment, follow the same types of steps to seed the database with STAC metadata and/or parcel records.

Development

source .venv/bin/activate

python -m pip install -e \
  'runtimes/business/logic' \
  'runtimes/eoapi/raster' \
  'runtimes/eoapi/stac' \
  'runtimes/eoapi/vector'