Skip to content

v0.6.3

Choose a tag to compare

View all tags
@slrelease slrelease released this 10 Jun 20:41
· 157 commits to main since this release
v0.6.3
b204605

Highlights

Chat requests can now combine text and images, while the model catalog gains more classification, extraction, and object-detection options.

Features

  • The gateway's OpenAI-compatible chat endpoint now accepts inline image data for vision-capable models while preserving text/image ordering.
  • Added ModernBERT-base-zeroshot-v2.0 and BART-large-MNLI classification configurations, GLiNER2-large-v1 extraction, and OWLv2-large-patch14-ensemble object detection.
  • Added Azure Blob support for model caching and large request payloads.

Bug fixes

  • Corrected image preprocessing and removed padding from document embeddings for nvidia/llama-nemoretriever-colembed-3b-v1.
  • Workers are removed promptly from gateway discovery when they shut down, reducing routing to stale workers.
Assets 2
Loading