This page includes a list of topics to be discussed with the team to find good solutions. Feel free to add/remove from it.

    • Users provide their own Google API Keys (need to figure out details)
    • how do we review the pull request? We have a API key for internal testing
    • Currently temporarily replacing key in extractor for student routes
    • Trying to see what we can do without API key
  • Box integration
    • Box Skills (beta)
      • Register Skill (DES Galaxy Recognizer)
      • Sends a JSON body to clusterman
        • fileid (unique identifier), read token (download file), write token (upload metadata to box)
      • bdbox.ndslabs.org runs clusterman/rabbitmq/mongo/extractors
        • clusterman creates a variation of the Clowder extraction message
        • extractor is a branch of pyclowder that reads/writes to box and calls Mattias ML code
  • Propagation of Fence credentials to Clowder
    • If an extractor creates derived file/dataset they cannot access them
    • Depends on Clowder 2.0 and Silhouette implementation
    • Working on propagating User API Keys to extractors
    • Need to figure out how to match fence users to clowder users and keys
  • User authentication for getting API access token
  • Adoption of Datasets/Spaces in the API to leverage extraction by space
  • Full docker stack deployment / writing new extractors/converters
    • Ben has in NDS workbench (Benjamin Galewsky where is this code?, can you paste the link here)
  • Defining parameters for extractors in extractor-info.json
    • so that clients can provide interface/support for which parameters can be set
    • required/available
  • Using binder repoToDocker as alternative to docker files for creating extractors
  • Tool Catalog user experience
    • Discuss internally ways to improve UX
    • Can we do an informal user study? Provide tasks and see where users get lost
  • No labels