Register Streams AI features with the Inference Feature Registry

[thomheymann]

Summary

Registers Streams AI suggestion features (partitioning suggestions and processing suggestions) with the Inference Feature Registry so that connector selection is managed centrally rather than via a custom internal API.

• Adds registerSuggestionsInferenceFeatures() that registers a parent streams feature and two child features (streams_partitioning_suggestions, streams_processing_suggestions) with the searchInferenceEndpoints plugin at setup time.
• Exports new inference feature ID constants from @kbn/streams-schema.
• Replaces the custom GET /internal/streams/_connectors route with the useLoadConnectors hook from @kbn/inference-connectors, passing featureId so connector selection is scoped to the registered feature.
• Simplifies useGenAIConnectors by removing manual connector loading/filtering logic and exposing isConnectorSelectionRestricted from the registry.
• Updates server-side suggestion routes to resolve the connector/model via the inference feature registry instead of accepting a raw connectorId.
• Updates related tests and Scout specs.

Screenshot

Made with Cursor

[elasticmachine]

Pinging @elastic/obs-onboarding-team (Team:obs-onboarding)

[macroscopeapp]

Approvability

Verdict: Needs human review

This PR introduces new AI feature registrations with the Inference Feature Registry, removes an API endpoint, and refactors connector fetching to use a different library with changed behavior. The author does not own any of the modified files, which are all owned by obs-onboarding-team and obs-sig-events-team. The designated code owners should review these changes.

^{You can customize Macroscope's approvability policy. Learn more.}

[flash1293]

This seems to work well, but I'm not sure about the changes on the server side. We are still passing the connector ID from the client to the server in all cases. So it seems like the branch where it picks the default on the server side is never actually called. Should we still do it then? Is this a preparation for some future simplification?

[thomheymann]

This seems to work well, but I'm not sure about the changes on the server side. We are still passing the connector ID from the client to the server in all cases. So it seems like the branch where it picks the default on the server side is never actually called. Should we still do it then? Is this a preparation for some future simplification?

@flash1293 Yeh good point. It doesn't add much value for us currently. I think where that approach shines is for background tasks or jobs that aren't triggered via UI interaction. But we can add it when it comes to that. I'll revert for now

[elasticmachine]

💛 Build succeeded, but was flaky

• Buildkite Build
• Commit: 1efb371

Failed CI Steps

• FTR Configs #219

Test Failures

• [job] [logs] FTR Configs #219 / integrations For each artifact list under management When on the Event Filters entries list should be able to add a new Event Filters entry

Metrics [docs]

Module Count

Fewer modules leads to a faster build time

id	before	after	diff
streamsApp	1758	1801	+43

Async chunks

Total size of all lazy-loaded chunks that will be downloaded as the user navigates the app

id	before	after	diff
streamsApp	1.9MB	1.9MB	+5.7KB

Unknown metric groups

API count

id	before	after	diff
@kbn/streams-schema	425	428	+3

History

• 💛 Build #424277 was flaky 943c669
• 💛 Build #424134 was flaky 694c74d