youngkermit8-coder opened a new pull request, #1736: URL: https://github.com/apache/rocketmq-dashboard/pull/1736
## Summary - route LLM configuration tests according to the selected engine - keep HTTP engines probing the provider model endpoint - test Claude Code and Qoder through their registered CLI provider availability - return `llm.provider.cli_missing` with an actionable hint when the selected CLI is unavailable ## Root cause `LlmConfigService.testConfig(...)` considered only whether the provider was OpenAI-compatible. It called `OpenAiCompatibleLlmClient.listModels(...)` even when the configured engine was `claude-code` or `qoder`, although production requests execute those engines through `AgentProviderRegistry`. This allowed an unrelated HTTP endpoint failure to reject a healthy CLI configuration, and an HTTP success could hide a missing CLI executable. ## Impact The settings "test connection" action now exercises the same provider type that runtime requests use. HTTP behavior is unchanged, while CLI users receive an immediate availability result without an unrelated network request. Closes #1731. ## Validation - baseline regression on exact `rocketmq-studio` `c2488cc`: failed deterministically (`expected: 0`, `but was: 1`) when the selected CLI engine was coupled to a failing HTTP model endpoint - `LlmConfigServiceTest`: 23/23 passed - AI controller/service/config/client/gateway/provider regression set: 62/62 passed - Maven Checkstyle: 0 violations - `mvn -DskipTests package`: passed - `git diff --check`: clean An expanded 95-test AI run passed 93 tests; its only two failures are the existing CRLF-sensitive `ToolGatewayServiceTest` fixture mutations independently covered by #1705. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
