Version: Next

📝 Trustwise Changelog

Below is a summary of major changes, improvements, and fixes. Each version section is referenceable via anchor links.

Version 3.3.0 (2025-04-08)

Together AI Embedding Model Support
- Added support for Together AI embedding models, expanding the list of supported embedding providers.

Custom Reranker Model Registration
- Users can now register most reranker models available from supported providers (Cohere, Together AI, Azure).
Reranker Model Support in Scans
- Users can now apply registered reranker models in Optimize:ai scans to test and compare reranker performance for their specific use cases.
Tokenizer Upload
- Users can now upload custom tokenizer files. This unlocks compatibility with LLMs that previously lacked open tokenizer access.
Cost and Carbon Control Maps
- Introduced new Cost and Carbon maps that help users configure cost and carbon maps for their scans based on their own cost policies.
Airgapped Support
- Quickstart is now compatible with airgapped environments, allowing for fully offline deployments.
Custom LLM Gateway Support (beta)
- Added support for registering and routing through custom LLM gateways, enabling advanced enterprise configurations.

Cost Endpoint Release
- API endpoint that estimates usage costs based on the user's LLM and/or reranker configuration.

Various stability improvements and minor UI fixes across model registration and scan interfaces.

Expanded Embedding Model Support
Users can now register their own embedding models from supported providers, including OpenAI, Hugging Face, and Azure. Previously, users were limited to a predefined list of embedding models.
Azure LLM & Embedding Model Support
Added support for Azure LLMs and embedding models in addition to previously available providers.

Custom Embedding Model Registration
- Users can now register most embedding models available with our supported providers (OpenAI, Hugging Face, Azure).
- This removes previous limitations on embedding model selection, allowing for greater flexibility and customization.
Asynchronous Document Processing
- Document uploads are now processed asynchronously, allowing users to leave the page without interrupting processing.
- A new status button provides real-time updates on the progress of uploaded documents, including success and error states.
Support for Large Scans
- The previous limitation of 3 queries at a time has been removed. Users can now run large scans without query limitations.
Expanded Storage Provider Support
- Optimize:ai now supports additional storage providers, including MinIO, offering greater flexibility for storage solutions.
OAuth2 Support
- Added Oauth2 support for standardized authentication and seemless integration.

Carbon Endpoint for SCI Calculations
- Released Carbon endpoint for SCI calculations