Guillaume Calmettes

Hello, Thanks for this project! `make rofi` was failing due to the need of `xcb-aux` as dependency. This MR adds it.

Add Scikit-Learn

8

Once Scipy is included, would Scikit-Learn be trivial to add? (dependencies Numpy/Scipy)

- Add uploads minio backend (see https://github.com/redpwn/rctf/issues/609) - Update fastify stack to remove [deprecated expect-ct header](https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Expect-CT) (see https://github.com/helmetjs/helmet/issues/310) which are enforced in latest Google chrome

**Is your feature request related to a problem? Please describe.** Currently, the files uploads providers are limited to either local path or GCS for object storage, which might be too...

Make default hitsAddend minimum value configurable

2

fix: https://github.com/envoyproxy/ratelimit/issues/730 In the current implementation of the `ratelimit` service, the `hitsAddend` value that is associated with each request is set to a minimum of `1`. This is done to...

Make the default hitsAddend configurable

1

The current minimum `hitsAddend` value is set to `1` (enforced via a [Max function](https://github.com/envoyproxy/ratelimit/blob/28b1629a21e885bdd2b527d6a1c1de8483dc47d4/src/redis/fixed_cache_impl.go#L52)). This prevents uses cases where the request itself would not carry any "weight" (e.g.: request to...

[BugFix]: properly catch templating error when preprocess input

6

This is a rework of https://github.com/vllm-project/vllm/pull/13039 following the inputs from @robertgshaw2-redhat (see in particular [this comment](https://github.com/vllm-project/vllm/pull/13039#issuecomment-2648030594)) Currently only `ValueError` exceptions are catched for the preprocessing input step on the frontend...

[Bugfix]: do not shutdown server is `skip_special_use=False` for MistralTokenizer

1

The `MistralTokenizer` [does not currently support](https://github.com/vllm-project/vllm/blob/bc6ccb987877000ec271e0076317b03a66cde4bc/vllm/transformers_utils/tokenizers/mistral.py#L455) the `skip_special_tokens=False` argument. Because an `assert` is used to check for the `skip_special_tokens` value, if one makes a requests passing the `"skip_special_tokens": false` parameter...

[Bugfix]: correctly propagate errors message caught at the chat_templating step to the client

2

This PR ensures the error message from the exception caught during the chat templating step is correctly propagated to the client. The previous implementation converted the error so it is...

[Bugfix] ensure tool_choice is popped when `tool_choice:null` is passed in json payload

9

When request is made with `"tool_choice":null,"tools":[]` in a json payload, vllm returns a `500 Internal Server Error`. The minimalist example below triggers the error: ``` curl -X POST -H "Content-Type:...