Advanced Knowledge AI Parser, Atmosphere Sounds, and More with Cognigy.AI v4.79

From advanced content extraction in Knowledge AI to optimizing VUX with contact center soundscapes, here is what awaits you in Cognigy.AI v4.79.

Semantic Content Extraction in Knowledge AI

In Cognigy Knowledge AI, a Knowledge Source is broken down into digestible, self-contained text units called Chunks. This approach enhances the system’s ability to match and retrieve the right information for any user query. The latest release introduces an Advanced Parser to optimize the knowledge extraction process.

Why It’s Better

Leveraging optical character recognition (OCR) technology and vision models, the Advanced Parser splits uploaded files into semantic Chunks, each in a meaningful Markdown format to improve comprehension and response formulation. Markdown allows the AI Agent to recognize other elements besides regular text like headers, tables, images, and links, aiding in better understanding and processing content structures.

Benefits

• High-Quality Responses: Meaningful Chunks facilitate accurate knowledge understanding and optimal answer generation by AI Agents.
• Broader Supported Content Formats: Using OCR to identify text and layout structures, Knowledge AI now also supports new file formats like .pptx, .png, .jpeg and more! (see demo)
• Source Referencing: Page numbers can now be automatically added as metadata for each chunk, allowing for precise tracking and reference of the answer source in lengthy documents.

How It Works

Advanced Parser is enabled by default in trial, private, and shared SaaS environments. For on-prem deployments, you can set up a dedicated connection with your own API key. See our documentation for details.

Watch the excerpt below from our last Tech Update where Alex, Cognigy’s Product Manager, explains the new Advanced Parser. See the complete deep dive here.

Hyper-Realistic VUX with Atmosphere Sounds

Cognigy.AI v4.79 also unveils the new Atmosphere Sounds feature, taking VUX to the next level. Audio files like contact center soundscapes can be added throughout voice interactions to create a more engaging, realistic, and user-friendly experience that effectively mitigates the adverse effects of latency. It’s like your AI Agents are sitting in a real contact center.

Benefits:

• Seamless Interaction: Atmospheric sounds fill the silence during response delays, making the interaction feel continuous rather than interrupted.
• Defeating Latency: This background noise gives users the impression that the AI Agent is actively processing their request, reducing perceived latency.
• Natural Experience: Adding a contact center soundscape creates a more realistic and familiar environment for users, mimicking the experience of engaging with a human agent.

How it Works

You can configure and activate Atmosphere Sounds using the Voice Gateway Set Session Config Node.

Watch the video below where Dilek, Cognigy’s Senior Product Manager demonstrates this feature in our last Tech Update. See the full video here.

Other Improvements

Cognigy.AI

• Added the API Post /v2.0/projects/{projectId}/nlu/scores request for matching and scoring sentences with NLU Intents
• Redesigned the Cognigy.AI interface to create a clearer brand identity and deliver a seamless user experience throughout the product
• Added support for readiness events in socket-based Endpoints (Webchat and Socket.io). This improvement enables socket-based Endpoints to trigger readiness events, indicating when they are prepared to send or receive data
• Added support for multiple languages in addition to English
• Removed the Cognigy Journey feature from the Cognigy.AI UI

Cognigy Voice Gateway

• Audio Pre-Caching is enabled by default. It pre-synthesizes and caches subsequent prompts during playback for seamless transitions.
• Improved by centralizing the definition of the call_terminated_by values
• Updated the vg-speech-utils service to support VG_TTS_TRIM_SILENCE_DISABLE_BY_ACC_SID
• Added a pre-commit hook that performs npm audit and npm run lint:fix
• Improved by support SemVer-tagged release images
• Added the maximum waiting time for AI responses through the Voice Gateway. The default value is 30 seconds. It can be adjusted in the Helm charts or Flux Fleet via the environment variable ENDPOINT_VG2_TIMEOUT_FOR_AI_RESPONSES_IN_SECONDS

Cognigy AI Copilot

• Added the capability to generate Adaptive Cards via Generative AI in the Copilot: Adaptive Card Tile Node

Cognigy Live Agent

• Improved the performance of the inbox members API endpoints (POST, DELETE, and PATCH) by optimizing them to achieve lower response times and better overall efficiency

For further information, check out our complete Release Notes here.