[I] Desktop: Test Connection reports a different model than the one selected (OpenCode Free) [maka]

["liuxiaocs7 (via GitHub)" <gi...@apache.org>]

liuxiaocs7 opened a new issue, #4222:
URL: https://github.com/apache/maka/issues/4222

   ### What happened
   
   On **Settings → Models**, for an **OpenCode Free** connection with only **MiMo V2.5 Free** enabled, clicking **测试连接 (Test Connection)** reports success for a *different* model:
   
   > 连接成功 · OpenCode Free
   > nemotron-3-ultra-free · 2154 ms
   
   The toast names a model the user never selected (`nemotron-3-ultra-free`), which reads as if the model selection didn't take effect.
   
   **Expected:** the test should verify the selected model, or — if it must fall back — clearly state that the selected model didn't respond and which model it used instead, rather than silently reporting success for another model.
   
   ### How to reproduce
   
   1. Open **Settings → Models** and select an **OpenCode Free** connection.
   2. In the model multiselect, enable **only** "MiMo V2.5 Free".
   3. Click **测试连接** while MiMo is rate-limited / temporarily unavailable / returns an empty completion.
   4. The toast reports **success for `nemotron-3-ultra-free`** instead of MiMo.
   
   ### Root cause
   
   For `opencode-free`, `testConnectionStrict` (`packages/runtime/src/test-connection.ts:166-195`) intentionally probes each candidate — enabled models first, then the provider's `fallbackModels` — and returns the **first one that answers**. When the selected model (`mimo-v2.5-free`) does not respond (rate limit, temporary outage, or an empty completion rejected by `isOpenAIChatCompletion`), the loop falls through to the next candidate — `nemotron-3-ultra-free`, the provider default (`OPENCODE_FREE_DEFAULT_MODEL`). The success toast in `runTest` (`apps/desktop/src/renderer/settings/use-connection-detail.ts`) then reports `result.modelTested` = the fallback model.
   
   The probe-until-one-answers behavior is **intended** for free models (they rotate and rate-limit; a "connection usable at all" check is more useful than a hard per-model failure). The defect is only that the toast is **misleading** — it presents a fallback as a plain success and hides that the selected model is currently down.
   
   ### Environment
   
   - Maka commit: `593efb9df`
   - OS: macOS 15.7.3 (24G419)
   - Surface: Desktop
   - Node.js: v24.14.0
   
   ### Logs / screenshots
   
   The success toast shows `连接成功 · OpenCode Free / nemotron-3-ultra-free · 2154 ms` while the model selector above it shows "MiMo V2.5 Free" as the only enabled model.


-- 
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: commits-unsubscribe@maka.apache.org.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@maka.apache.org
For additional commands, e-mail: commits-help@maka.apache.org

Re: [I] Desktop: Test Connection reports a different model than the one selected (OpenCode Free) [maka]

["Astro-Han (via GitHub)" <gi...@apache.org>]

Astro-Han closed issue #4222: Desktop: Test Connection reports a different model than the one selected (OpenCode Free)
URL: https://github.com/apache/maka/issues/4222


-- 
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: commits-unsubscribe@maka.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org


---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscribe@maka.apache.org
For additional commands, e-mail: commits-help@maka.apache.org