Significant technical updates and improvements
Task Revisions
Introduced version control for Tasks, enabling advanced testing and optimisation capabilities.
Implemented Task Revisions for version control
Prepared foundation for A/B testing of prompts
Enabled future model performance evaluation
Set up potential for model vendor load balancing
Retrieval-Augmented Generation (RAG) Support for Tasks
Enhanced Task capabilities with RAG support, allowing for more context-aware AI responses.
Added RAG support for Tasks
Enabled configuration with existing Rightbrain collections
Improved Task
output_formatfor more detailed, structured outputs
Gemini Model Integration
Expanded our model offerings with VertexAI-powered Gemini integration.
Integrated Gemini models into our LLM offerings
Broadened options for diverse use cases
OAuth Client Support
Improved API security with OAuth support for enhanced authentication and authorisation.
Introduced OAuth client support
Created new endpoint for OAuth client application registration
Implemented database linking between OAuth clients, authors, projects, and organisations
Anthropic Model Integration
Broadened model selection with Anthropic integration, providing more choices for language processing.
Added support for Anthropic's language models
Implemented dynamic LLM interface selection
New model_factory uses type definitions to determine LLM interface to return
Enhanced Audit Logging
Improved audit capabilities for enhanced security and compliance.
Enhanced audit logging with project-specific storage
Automatic creation of dedicated storage buckets for each new project
Implemented log routing to project-specific storage buckets
Added generic interface for logging pre-defined audit events
Conversational Guardrails Framework
Strengthened conversation safety with a new guardrails system.
Introduced conversational guardrails framework
Implemented
prompt_policyandpii_policyguardrailsDesigned for potential future integration with NVIDIA Nemo guardrails project
