New
- Grok 2 and Grok 2 Vision added to the LLM selector. We are anticipating a Grok 3 API release soon and will add it as soon as possible.
- Deepseek R1 added to the model selector. This is a reasoning model so response times will be in line with LLMs like o1. The model is hosted in AWS and differs from the deekseek[dot]com
- Upgrade packages inside the Hatz Platform and display multiple packages.
- Usage notifications showing credit overages.
- AI Help Center. Talk to our new assistant "Lark Meeper" about Hatz and how to use the platform.
Improved
- Improved Retrieval Augmented Generation. Our chunk ranking and retrieval strategy was improved to rank find more text blocks, set better similarity thresholds, and was optimized for faster responses. You should now see much better results when working with documents larger than the context window of the chosen model.
Fixed
- Notifications and logic to prevent inviting more users than licenses available.
