[wylovelyi]

What this PR does

Closes #459 — when testing a model connection, if the configured model is not supported, the provider returns 404 with body "code":"model_not_supported". But ModelConfigOpsService.parseErrorMessage matched the generic 404 branch first and showed the misleading message "接口未找到 (404)，请检查 Base URL 或者路径配置地址", which sends users down the wrong path (they think the URL/path is wrong, not the model).

Root cause

In parseErrorMessage, the order of checks was:

1. 401 → auth error
2. 404 → "interface not found" ← matched first for the unsupported-model 404
3. 429 → quota
4. else → original message

Since the unsupported-model response is a 404, it was caught by step 2 and never surfaced model_not_supported.

Fix

• Added a model_not_supported check before the generic 404 branch. It matches both the error code (model_not_supported) and the message text (unsupported model, case-insensitive), and returns a clear "模型不支持 (model_not_supported)：请确认模型名称正确，且当前厂商/兼容模式支持该模型。"
• A plain 404 (genuinely wrong Base URL / path) still reports "接口未找到" as before — no regression for that case.
• Made parseErrorMessage package-private (was private) so it can be unit-tested directly, and added 4 cases to ModelConfigOpsServiceTest: model_not_supported, generic 404 regression guard, 401, 429.

Files changed

File	Change
.../service/aimodelconfig/ModelConfigOpsService.java	model_not_supported detection before 404; parseErrorMessage → package-private
.../service/aimodelconfig/ModelConfigOpsServiceTest.java	4 new parseErrorMessage test cases

Note

Only error-message mapping logic changed; no DB schema, no API contract, no behavior change beyond the corrected message.

[wylovelyi]

@zxuexingzhijie 你好，这个 PR 在模型不支持时返回 model_not_supported 而非误导性的 404（#459）。准备好 review 了，谢谢！