v0.7.0

Vexa-ai/vexav0.7.0Oct 16, 2026by Pringled

Upgrading?

AI Summary

This release focuses on performance improvements, including parallel chunking, batch embeddings, and a significant upgrade to the MCP SDK.

Key Highlights

  • Upgrade to MCP SDK 2.x
  • Parallelize chunking for better performance
  • Batch embeddings and upgrade to model2vec
  • Add file-types command
  • Speed up warm queries

New Features

  • Store BM25 postings as numpy arrays
  • Speed up warm queries
  • Use bare MCP tool names in agent instructions
  • Upgrade to MCP SDK 2.x
  • Parallelize chunking
  • Batch embeddings and upgrade model2vec
  • Add file-types command

Full Release Notes

## What's Changed
* feat: Store BM25 postings as numpy arrays by @Pringled in https://github.com/MinishLab/semble/pull/284
* chore: Speed up warm queries  by @Pringled in https://github.com/MinishLab/semble/pull/285
* chore: Walk files with os.scandir by @Pringled in https://github.com/MinishLab/semble/pull/286
* fix: Use bare MCP tool names in agent instructions by @Pringled in https://github.com/MinishLab/semble/pull/289
* chore: Upgrade to MCP SDK 2.x by @Pringled in https://github.com/MinishLab/semble/pull/294
* feat: Parallelize chunking by @Pringled in https://github.com/MinishLab/semble/pull/295
* feat: Batch embeddings, upgrade model2vec by @Pringled in https://github.com/MinishLab/semble/pull/298
* feat: Add file-types command by @Pringled in https://github.com/MinishLab/semble/pull/299


**Full Changelog**: https://github.com/MinishLab/semble/compare/v0.6.2...v0.7.0

Related Repositories