model2vec Releases

This release focuses specifically on fixing TLS router configurations to handle edge cases involving multiple entry points.

Jun 10, 2026

This release introduces significant new features centered around Text-to-Speech (TTS), including a new vLLM Omni Speech Adapter, a comprehensive Web UI for voice management, and an updated environment configuration template.

Jun 7, 2026

This release focuses on bug fixes and dependency updates, including improvements to Redis middleware configuration, WebUI libraries, K8s GatewayAPI status updates, and HTTP3/TLS support.

Jun 5, 2026

This release contains the same set of bug fixes and dependency updates as v3.7.4, targeting Redis middleware, WebUI libraries, K8s GatewayAPI, HTTP3, and TLS.

Jun 5, 2026

This release focuses on fixing multiword token handling, refactoring the training pipeline, and upgrading the underlying skeletoken library. It also introduces new options for training configuration and trusted publishers.

May 29, 2026

This update brings significant new capabilities including OCR support for embedded images and PDFs, alongside an Azure Content Understanding converter. It also resolves critical memory leaks and recursion errors in PDF and HTML processing.

May 26, 2026

The release introduces a comprehensive admin interface featuring a dashboard for configuration management and session monitoring. It also adds an avatar generation system with task submission and status tracking capabilities.

May 24, 2026

This release focuses on critical security updates, including a new experimental obfuscation layer and fixes for ACL bypass vulnerabilities. It also includes improvements to packet sniffing, outbound rules, and overall performance.

May 23, 2026

A patch release for Hysteria Realms that resolves connection issues for servers behind symmetric NAT and updates the `hysteria cert` command configuration.

May 10, 2026

Introduces Hysteria Realms, a feature enabling users to host a server without a public IP or port forwarding by utilizing NAT hole punching for P2P connectivity.

May 10, 2026

This release removes the deprecated Facebook (Meta) source and introduces the new Sub.md passive source. It also includes various bug fixes for transport errors, rate limits, and response handling across multiple data sources.

Apr 27, 2026

Updated the `sessionid` data type from integer to string to improve extensibility. Frontend API interfaces must be modified to accommodate this change.

Apr 11, 2026

Major refactoring introducing a plugin-based architecture for digital human models and TTS, reorganizing code into specific directories, and adding Aliyun QwenTTS support.

Mar 31, 2026

A minor patch release addressing Windows compatibility issues by using POSIX paths and bumping the version to 0.8.1.

Mar 27, 2026

Major update replacing the legacy tokenizer code with skeletoken, refactoring persistence methods, and deprecating Python 3.9 support. Includes updated documentation and MTEB results.

Mar 26, 2026

This release introduces URLScan.io as a new passive subdomain source and resolves a timeout issue with the Crtsh source. It also includes organizational updates like alphabetically sorting sources and updating the Dependabot working directory.

Mar 9, 2026

Enhancements to PDF table extraction capabilities include support for aligned Markdown and wide tables, alongside a fix for partially numbered lists. The release also removes a Windows pin for the ONNX Runtime dependency.

Feb 20, 2026

Added API key support for HackerTarget and Reconeer, added optional API key support for sources, and added request tracking to the -stats flag.

Jan 12, 2026

Introduced new data sources including Profundis, MerkleMap, and THC, alongside improvements to context cancellation and an update to Censys.

Dec 19, 2025

Maintenance release addressing security vulnerabilities in mammoth and pdfminer.six libraries.

Dec 1, 2025