ADL S3 Plugin

Pushes observations out of an ADL instance as CSV objects into an S3-compatible bucket — Amazon S3, a self-hosted MinIO, or any service speaking the S3 API (DigitalOcean Spaces, Backblaze B2, Wasabi). It is a dispatch plugin: it collects nothing, and adds two dispatch channel types, S3 Upload and MinIO Upload, to the Dispatch Channels chooser. Each channel writes either one object per observation or a daily object per station that later runs append to.

Operator guide: docs/guide.md — prerequisites, IAM permissions, every channel field, the CSV layout, write modes, parameter mappings, diagnostics and troubleshooting. The guide is also published on the central ADL documentation site.

Development setup

The plugin runs inside the ADL core image. Build the adl:latest image from the ADL core repository first, then:

git clone https://github.com/wmo-raf/adl-s3-plugin.git
cd adl-s3-plugin
cp .env.sample .env        # set PLUGIN_BUILD_UID=$(id -u), PLUGIN_BUILD_GID=$(id -g), ADL_DB_PASSWORD
docker compose build
docker compose up
docker compose exec adl adl createsuperuser

The admin is served on PORT (default 8080). The plugin source is bind-mounted, so code changes reload the dev server. If the build fails with pull access denied for adl:latest, prefix the build with DOCKER_BUILDKIT=0.

Tests are Django-runner tests under plugins/adl_s3_plugin/src/adl_s3_plugin/tests/; the CSV and append-merge tests are pure functions and need no database. Lint and format from plugins/adl_s3_plugin/ with make lint and make format. See CONTRIBUTING.md — a change to any channel field must update the guide in the same PR.