Targeted Knowledge Retrieval and More with Cognigy.AI 2025.23 Release

Cognigy.AI 2025.23 introduces smarter knowledge access for AI Agents and more flexible Webchat styling with new customizable elements.

AI Agent Knowledge Tool for More Efficient Workflows 

While Cognigy has long supported attaching a Knowledge Store to an AI Agent, this release brings a new Knowledge Tool that enhances precision and efficiency for that process.

With the Knowledge Tool, you can define the instruction for exactly when a specific Knowledge Store should be used, instead of querying it on every user input. This selective execution ensures that knowledge retrieval is executed only when it’s needed, improving accuracy and reducing latency. The Tool can be configured directly in the AI Agent Editor or referenced within an AI Agent Node or LLM Prompt Node in your Flow.

Previously, achieving similar functionality required chaining together a Tool Node, Knowledge AI (Search, Extract & Output) Node, and a Resolve Action Node. The new Knowledge Tool combines all three into one streamlined component, making your Flows cleaner and development faster.

Enhanced Webchat Styling with Launcher Icon Customization

The latest release also adds more flexibility to the styling and branding of your Webchat experience. The Webchat Icon settings now allow you to customize the displayed icon that users click to open the Webchat widget on your site.

You can:

• Choose from three default icons that align with common UI patterns
• Or upload your own custom icon to match the launcher appearance with your brand and visual identity.

This update contributes to a cohesive, on-brand chat experience without additional CSS or front-end tweaks.

Other Improvements

Cognigy.AI 

• Added the option to define a custom header to authenticate through OAuth 2.0 in the Azure OpenAI LLMs

Cognigy Live Agent

• Added translation logic for API availability messages. System messages such as User left the conversation and User joined the conversation are now displayed in the correct language

Cognigy AI Ops Center

• Limited the options in the subcomponent filter to only show component-relevant options