iloveeyjafjalla opened a new issue, #6275: URL: https://github.com/apache/rocketmq-dashboard/issues/6275
### Before Creating the Bug Report - [x] Searched all-state Issues/PRs and inspected related current PR head differences and public comments. - [x] This is a RocketMQ Studio frontend defect in this repository. - [x] Reproduced on the exact current revision below. ### Studio Version `rocketmq-studio@5e4c39b053c35a5598cc79ff331b8a54e649d7b1`. ### Runtime Environment / Build Toolchain Windows 11, Node 24.15.0, npm 11.12.1. Automated reproduction renders real HomePage → AiPage routing and the real Composer model selector with controlled API/stream boundaries. No live provider or deployment was used. ### Connected RocketMQ Cluster No cluster was contacted. The defect concerns the model selector after the existing Home draft handoff. ### Describe the Bug `selectModel` adds a handed-off model to the runtime's options, but an in-flight `getLlmModels` response replaces the entire array. A Home model absent from that catalog therefore disappears from the AI selector's menu. Its initial selected label and outgoing request are still correct; selecting a different model then prevents switching back through the menu. ### Steps to Reproduce 1. Use a saved model absent from its catalog. The repository already saves `tongyi` / `qwen-plus` in `LlmConfigServiceTest.listModelsShouldUseSavedProvider`; the current builtin Tongyi catalog contains twelve other IDs. Home explicitly appends the saved model to its options. 2. Select `qwen-plus` on Home and submit a prompt normally. 3. Allow the AI page's catalog request to complete after the Home draft has been applied. 4. In the AI model selector, select `qwen3.8-max`, then open the menu to select `qwen-plus` again. ### What Did You Expect to See? The explicitly handed-off model stays in the menu alongside the current catalog, without a duplicate if the catalog already contains it. ### What Did You See Instead? The menu has no `qwen-plus` option after the catalog replaces the array. The controlled old-code page regression reaches `/ai/c/7`, confirms the first request still carries `qwen-plus`, switches to `qwen3.8-max`, and then fails to find the original option. ### Additional Context The corrected regressions produce three expected failures and one passing control against unchanged production code. The same two test files pass 29/29 after retaining explicit handoff options separately from catalog entries. The PR documents build/lint and complete frontend-suite results. This is a menu-option preservation fix, not a claim that a model provider accepts or serves a particular model. No inference failure, provider outage or production incident is asserted. AI assistance: OpenAI Codex investigated, implemented and ran automated checks, including separate automated review. ### Are You Willing to Submit a Pull Request? - [x] Yes. ### Implementation PR https://github.com/apache/rocketmq-dashboard/pull/6274 -- 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]
