What’s New?
This release introduces several major features that enhance the Clarity experience for end users:
- End User Agent Selection: An enhanced Settings dialog allows you to pick and choose which agents are displayed in your agent dropdown, giving you control over your agent list and enabling you to customize your workspace to show only the agents you use most frequently.
- Featured Agents in Chat Portal: A new “Featured agents” section in the agent dropdown highlights the agents available to all Clarity users, making it easy to see agents using all of the AI models available to users.
- Prompt Injection Detection for Knowledge Sources: A prompt injection is a malicious instruction designed to manipulate a Large Language Model to behave in unintended ways. When this shield is enabled for the data pipeline used by an agent, whenever you or a user uploads a file, it will automatically be scanned for prompt injections. If a prompt injection is detected, you will receive an error message and the file will not be uploaded, indexed, or available for subsequent conversations. This acts like an “antivirus” for knowledge sources, protecting your agents from malicious content embedded in files.