GitHub Action: manage S3 artifacts in workflows
action: upload
Upload will tarball and upload artifact(s) to S3 and optionally manage authentification if AWS credentials are specified.
See usage here.
Documentation is found here.
action: download
Download will download artifacts from s3 path and optionally extract files from
tar if s3 path ends with .tgz
or .tar.gz
.
See usage here.
Documentation is found here.
Each Action has a detailed README for how to use it as referenced above. Please review Issues, post new Issues against this repository as needed.
Please see here for guidelines on how to contribute to this project.
- actions-s3-artifact/upload
- Upload a build artifact to S3 for future use externally
- Uploading an ML model to use in a container
- Uploading a Javascript bundle for serving
- Upload an artifact for future use in another job
- Uploading a preprocessed data set for use in a training job
- Uploading a javascript build for use in parallel jobs
- Upload a build artifact to S3 for future use externally
- actions-s3-artifact/download
- Needs to be able to fail without erroring the pipeline for a cache-like behavior
- Needs to be able to output its success status for future pipeline steps
- Download an external artifact from S3 for use in a job
- Downloading data set to work with in CI/CD
- Downloading ML model to build into deployment CD
- Download an artifact from a previous job for use in a job
- Downloading a preprocessed data set for use in a training job
- Downloading a javascript build for use in parallel jobs
- actions-s3-artifact
- TBD: Might not be worth it
- Cache and retrieve an artifact based on a possibly changing key or hash of a file
- Check cache
- If it's present, skip computation
- If it's not present, compute
- Upload to cache, if it was not retrieved