User Story
As a contributor, I want to add an example provider profile by contributing a reviewed YAML file, without registering its ID in product code or maintaining provider specific documentation.
Problem Statement
The inference provider table duplicates information in example profiles and requires a new row for each contribution. Provider lifecycle telemetry maintains named profile buckets even though gateways have no built-in profiles. A bucket identifies a profile ID, not the definition an operator imported.
Impact / Why This Matters
Contributors are led to change docs, telemetry code, and tests for a sample YAML. Operators may mistake the table for the profiles available on their gateway. Continuing to extend these lists adds maintenance without making either one authoritative.
Proposed Design
Remove the inference provider table and direct readers to the example profile directory and openshell profile list. Remove all named provider profile telemetry buckets, the ID-to-bucket mapping, and the providerProfile field from newly emitted provider lifecycle events. Keep the event's operation and outcome.
Alternatives Considered
Keep the named buckets: they remain misleading when operators can import modified profiles under the same IDs. Emit raw profile IDs: that could disclose operator-defined names.
Acceptance Criteria
User Story
As a contributor, I want to add an example provider profile by contributing a reviewed YAML file, without registering its ID in product code or maintaining provider specific documentation.
Problem Statement
The inference provider table duplicates information in example profiles and requires a new row for each contribution. Provider lifecycle telemetry maintains named profile buckets even though gateways have no built-in profiles. A bucket identifies a profile ID, not the definition an operator imported.
Impact / Why This Matters
Contributors are led to change docs, telemetry code, and tests for a sample YAML. Operators may mistake the table for the profiles available on their gateway. Continuing to extend these lists adds maintenance without making either one authoritative.
Proposed Design
Remove the inference provider table and direct readers to the example profile directory and
openshell profile list. Remove all named provider profile telemetry buckets, the ID-to-bucket mapping, and theproviderProfilefield from newly emitted provider lifecycle events. Keep the event's operation and outcome.Alternatives Considered
Keep the named buckets: they remain misleading when operators can import modified profiles under the same IDs. Emit raw profile IDs: that could disclose operator-defined names.
Acceptance Criteria
providerProfileand new events without it.