What’s New?
This release introduces significant new capabilities, performance improvements, and enhancements across the entire Clarity platform:
- Self-Service Custom Agents: Create, customize, manage, and share your own custom agents directly from the Clarity User Portal without administrator involvement. Create an agent by defining settings that will shape your agent’s personality and expertise by composing a system prompt; choosing the underlying LLM; and uploading files as knowledge sources.
- Managing Your Agent List: The Settings dialog gives you full control over which agents appear in your agent dropdown menu. The agent catalog includes Yale’s featured agents, agents shared by other users, and agents you create yourself. Marking an agent as “Enabled” in your settings dictates its inclusion in your agent dropdown menu.
- Automatic Conversation Naming: Every new conversation is automatically given a meaningful title based on your first exchange, so your conversation list stays organized without any effort on your part.
- Faster, More Reliable File Uploads: Files you upload are now fully processed during the upload itself — content is extracted, validated, and indexed before the upload completes. This enables you to know immediately if there are any issues with your file.
- Improved Shields on Data Pipelines: Data pipeline shields can now either block an entire file when it fails a safety check (like prompt injection detection) or block only the specific unsafe content while still loading the rest of the document. This means one problematic section won’t stop you from using the remaining safe and useful information, and unsafe content is kept out of an agent’s knowledge base.
- More Reliable Agent Responses: Improved agent responses when errors occur, providing clear and actionable messages instead of a cryptic failure notification.
- Model Upgrades: The following models will be upgraded with this release.
- GPT-4o upgraded to GPT-5.2
- GPT-4o mini upgraded to GPT-5 mini
- Claude Sonnet 4.5 upgraded to Claude Sonnet 4.6
- NOTE: Existing conversations that use GPT-4o, GPT-4o mini, or Claude Sonnet 4.5 can no longer be continued. When opening a previous conversation with one of these models, Clarity will prompt selection of an agent and will open a new chat with the selected agent.