vLLM
vLLM Multimodal Processor Cache Bug — StopIteration (400) due to an audio item cache conflict in Qwen3-Omni's `use_audio_in_video`. Identified, analyzed, and reported.
[Tool Parser] Fixed exception handling for backslash escapes
Fixed exception handling for backslash escapes in `make_valid_python`
[CI/Build] Fixed a test for invalid CLI argument validation in test_cli_args.py
https://github.com/vllm-project/vllm/pull/46779
[CI/Build] Support for `bad_words` at the `/v1/completions` endpoint
https://github.com/vllm-project/vllm/pull/46793