v2.23.1

What's Changed

Features

• MultiModalEmbeddingParam Parameter Enhancements (#273)
  Aligned with Python SDK by adding multiple explicit builder fields: dimension (output vector dimension), outputType, fps (video frame sampling rate), instruct (custom task instruction), enableFusion (fuse multiple contents into single vector), resLevel (resolution level), maxVideoFrames (max video frames). Resolves user-reported issue where dimension could not be set when calling Multimodal-Embedding for image vectorization (expected 1152 dimensions, not 1024). Generic .parameter(k, v) channel remains available and takes precedence, ensuring full backward compatibility.

• Omni Model Video Parameter Rename (#274)
  Renamed the representation_compact parameter for input video in the Omni model.

Contributors

• @lzsweb
• @songguocola

Full Changelog: v2.23.0...v2.23.1