- add native openaiCompatible provider type with flat baseUrl/apiKey/dialect
config fields and declared per-model capabilities
- add admin-only probeManagedCopilotProfile GraphQL mutation that dispatches
real probe requests against server-managed copilot profiles, reusing the
BYOK probe engine with an AllowPrivate egress policy for self-hosted
endpoints (Vertex profiles rejected; use workspace BYOK instead)
- expose providers.profiles app config descriptor to the admin console
- add 'AI Providers' settings group with JSON editor and per-profile Test
action that probes declared capabilities and reports verified/failed
status per model and operation
luulam
177150086f