Skip to main content

Release v1.8

K
Written by Kay Chen

We are thrilled to announce the latest update to our platform. This release brings new features, improvements, and backend enhancements. Detailed below are the specifics of the updates included in this release.

Composer

  • New attributes have been added to the API response for generative AI calls utilizing Retrieval Augmented Generation. These include a relevance score, expandedText, and expandedScore for the text chunks retrieved through the RAG process.

  • Introduced a parameter enabling API callers to define a relevance threshold. Only text chunks from the RAG that meet or exceed this relevance threshold will be utilized.

  • Support is now available for new Generative AI models including BioMistral, Llama3 70b, Llama3 8b, and Mixtral Unquantized

  • Support for additional quantization mechanisms, such as GPTQ, enhances model throughput performance.

  • Resolve an issue where the vector database occasionally encounters a limit on the number of rows returned.

Layar

  • Implemented a new feature to log all JSON requests and responses within the API for audit purposes.

Signal

  • Implemented a Notification Management Panel

  • Resolved an issue with the 'Find Related Terms' feature that occasionally failed to display results.

Did this answer your question?