Oracle has done a good job exposing REST APIs for AIDP. However, the APIs currently expose operations without exposing enough metadata for customers to build robust automation around them.
As a result, developers frequently have to:
- inspect browser network traffic
- experiment with parameter values
- rely on incomplete documentation
- hard-code values that may change over time
This makes CI/CD, Infrastructure as Code and automated environment provisioning significantly more difficult than it needs to be.
The APIs should expose metadata describing valid values.
For example:
- supported cluster types
- supported compute shapes
- supported Spark versions
- notebook runtime versions
- workspace types
- catalog options
- region-specific capabilities
Configuration metadata
Rather than publishing static tables, expose endpoints that return:
GET /supportedClusterConfigurations
GET /supportedComputeShapes
GET /supportedSparkVersions
GET /supportedNotebookRuntimes
GET /supportedWorkspaceTypes
That way automation tools never become outdated.
The current REST APIs expose the required operations but do not expose sufficient metadata for reliable automation. Enhancing the REST APIs with discoverability, metadata endpoints and complete parameter documentation would significantly improve automation, reduce support requests and make AIDP easier to integrate into enterprise DevOps processes.
