### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

When configuring a model via the database (DB model), parameters set in litellm_params (such as use_chat_completions_api) are not applied to the actual LLM request.

For example, setting use_chat_completions_api: true in a DB model's litellm_params JSON field has no effect - the request still tries to use the native Responses API endpoint instead of bridging to chat completions.

Expected behavior: Parameters in DB model's litellm_params should be merged into the request kwargs, just like they are when using YAML configuration.

### Steps to Reproduce

1. Create a DB model with `use_chat_completions_api: true` in `litellm_params`: ```json { "litellm_params": { "model": "glm-5.1", "api_base": "https://open.bigmodel.cn/api/coding/paas/v4", "custom_llm_provider": "openai", "use_chat_completions_api": true } } ```

2. Send a request to `/v1/responses` using this model

3. Observe the error: ```json { "error": { "message": "litellm.NotFoundError: NotFoundError: OpenAIException - {\"error\":\"Not Found\",\"path\":\"/v4/resp…