llm-engineer-toolkit Releases

This release focuses on dependency updates and infrastructure improvements, specifically upgrading the Go runtime version and fixing installation scripts for RHEL and Rocky Linux systems.

May 7, 2025

A maintenance release that addresses security vulnerabilities and introduces a security feature to prevent Node.js code injection within the sandbox environment.

Mar 24, 2025

A major version update for the Ultravox speech model, focusing on significant performance improvements across transcription accuracy, question answering, and translation, as well as a massive expansion of supported languages.

Feb 11, 2025

A maintenance and quality improvement release that upgrades the Whisper encoder to a larger model, adds support for six new languages, and fixes several training and evaluation bugs.

Nov 12, 2024

Oct 9, 2024

Ultravox 0.4 upgrades the Whisper encoder from small to medium and significantly expands multi-lingual support to include seven new languages. The release also introduces offline batch inference and a text/voice conversation demo, alongside substantial performance improvements in speech translation metrics.

Aug 27, 2024

Ultravox 0.3 introduces a frozen Llama 3.1 8B pre-trained core and utilizes a Knowledge Distillation approach for loss calculation to improve speech-to-text alignment. The model was trained on 2.5k hours of speech with synthetic data augmentation and offers early preview access to its APIs.

Aug 23, 2024