v1.8.5

roboflow/inferencev1.8.5Apr 19, 2026by lfnovo

AI Summary

This release introduces Spanish (es-ES) language support with extensive translation coverage and improves embedding chunking to use tokens instead of characters for better handling of CJK and mixed-language content. It also fixes crashes in the credentials endpoint caused by encryption key mismatches.

Key Highlights

  • Spanish (es-ES) language support with full translation coverage
  • Token-based embedding chunking for better handling of CJK and mixed-language content
  • Resilient credential handling to prevent crashes from broken/mismatched encryption keys

New Features

  • Spanish (es-ES) locale with 930+ translation keys
  • Token-based embedding chunking replacing character-based
  • Broken credentials can now be displayed with a warning and deleted
  • Improved local quick-start documentation

Full Release Notes

## Highlights

- 🇪🇸 **Spanish (es-ES) language support** with full translation coverage
- **Token-based embedding chunking** — better handling of CJK and mixed-language content
- **Resilient credential handling** — broken/mismatched encryption keys no longer crash the API

---

## Added

- Spanish (es-ES) locale with 930+ translation keys and language toggle entry (#757) — thanks @chemazener!

## Changed

- Embedding chunking is now token-based instead of character-based, improving chunk sizing consistency for CJK and mixed-language content (#542, #749) — thanks @unendless314!
- `OPEN_NOTEBOOK_CHUNK_SIZE` and `OPEN_NOTEBOOK_CHUNK_OVERLAP` semantics changed from characters to tokens; default reduced from 1200 characters to 400 tokens to stay safely below the 512-token ceiling of BERT-family embedders (e.g. mxbai-embed-large). Existing stored embeddings are unaffected — only new ingestions use the new chunking.
- Internal: migrated i18n from the custom Proxy pattern to standard react-i18next `t()` function for better tooling compatibility (#755)

## Fixed

- Credentials endpoint no longer crashes (500) when the encryption key doesn't match stored credentials (#740, #753)
- Broken credentials are now shown with a decryption warning and can still be deleted
- DELETE endpoint for broken credentials supports model migration via the `migrate_to` parameter
- Azure credentials: `base_url` now correctly maps to `endpoint` (#741)

## Docs

- Improved local quick-start documentation (#734) — thanks @vincentrou!

## Dependencies

- Bump `langchain-openai` 1.1.7 → 1.1.14 (#762)
- Bump `langchain-text-splitters` 1.1.0 → 1.1.2 (#761)
- Bump `langchain-core` 1.2.22 → 1.2.28 (#737)
- Bump `langsmith` 0.6.4 → 0.7.31 (#759)
- Bump `authlib` 1.6.9 → 1.6.11 (#760)
- Bump `cryptography` 46.0.6 → 46.0.7 (#736)
- Bump `pytest` 9.0.2 → 9.0.3 (#752)
- Bump `python-multipart` 0.0.22 → 0.0.26 (#758)
- Bump `tornado` 6.5.4 → 6.5.5 (#668)
- Bump `next` 16.1.7 → 16.2.3 (#744)
- Bump `axios` 1.13.5 → 1.15.0 (#748)
- Bump `follow-redirects` 1.15.11 → 1.16.0 (#754)
- Bump `vite` 7.3.1 → 7.3.2 (#728)

---

## 🙏 Thanks

Huge thanks to everyone who contributed to this release:

- @chemazener — Spanish localization
- @unendless314 — token-based embedding chunking
- @vincentrou — documentation improvements

And thanks to all the users who opened issues and helped us prioritize what to ship next!

**Full Changelog**: https://github.com/lfnovo/open-notebook/compare/v1.8.4...v1.8.5