## Check for existing issues

- [x] I searched the existing issues and checked that this is not a duplicate

## What happened?

Some LiteLLM Proxy vector store routes do not carry the request model into provider routing. When no managed registry model is available, `?model=` and `x-litellm-model` can be ignored for create, list, update, and search requests, causing a wrong provider fallback or a `model=None` failure

Model-embedded File IDs can also reach the upstream provider without decoding. This affects vector store routing and provider request preparation in the Proxy

## Steps to Reproduce

1. Configure multiple provider deployments in the Proxy `model_list` 2. Call vector store create, list, update, or search with `?model=<model>` or `x-litellm-model: <model>` 3. Observe that the selected provider does not match the request model, or that routing fails with `model=None` 4. Use a File ID produced by model-based file routing and observe that the model-embedded ID is forwarded unchanged

## Relevant log output

```text Expected: the request model selects the configured deployment and the upstream receives the original File ID Actual: the request falls back to the wrong provid…