Improved LLM Ops Visibility and More with Cognigy.AI 2025.16 Release

Cognigy.AI 2025.16 delivers targeted enhancements across LLM orchestration, Live Agent routing, Salesforce MIAW integration, and developer tooling. This release strengthens operational reliability and enhances agent-user interaction fidelity.

Visual Warnings for Deprecated LLMs

To help developers stay aligned with rapid LLM release cycles and maintain production continuity, Cognigy.AI now displays explicit visual alerts for deprecated models in the LLMs Resource. This provides early visibility into upcoming changes, enabling timely migration to supported models.

In addition to these visual warnings, several deprecated models are being phased out of Cognigy.AI:

• September 24, 2025 – Google will retire gemini-1.5-flash and gemini-1.5-pro.
  Migration path:
  gemini-1.5-flash → gemini-2.0-flash
  gemini-1.5-pro → gemini-2.0-flash
  Support for the retired models will be removed from future Cognigy.AI versions.
• April 21, 2025 – Google retired gemini-1.0-pro and text-bison-001.
  Migration path: Gemini 2.0 Flash Lite
  These models have been removed as selectable options starting with Cognigy.AI 2025.16.
• July 21, 2025 – Anthropic retired claude-3-sonnet.
  Migration path: claude-sonnet-4
  Support for claude-3-sonnet is removed in Cognigy.AI 2025.16 and later.

Refer to the list of all upcoming removals on the Deprecations and Removals page.

Salesforce MIAW Integration Enhancements

The 2025.16 release expands Salesforce MIAW channel capabilities, delivering greater flexibility and feature parity.

Bidirectional Typing Indicator Support

Previously, Cognigy Webchat reflected typing activity from Salesforce agents, allowing end users to receive real-time responsiveness cues. With 2025.16, this capability is fully bidirectional. Human agents can now see when the customer is typing, enabling better engagement assessment and workflow optimization.

Attachment Handling Parity

Not supported with the previous Salesforce chat product, Salesforce MIAW now allows human agents to send file attachments directly from the messaging interface during handovers. These attachments are rendered for end users in Cognigy Webchat, ensuring parity with this Salesforce MIAW functionality.

Supported file types include PDFs, images, and other document types. Videos are not yet supported by Salesforce.

Other enhancements include:

• Added the ability to override the Flow Locale with a custom language for the Salesforce MIAW handover provider. To use this feature, go to the Handover to Agent Node, select Salesforce MIAW from the Handover Provider list, and add the language in the Conversation Routing Attributes fields using the format: "language": "es_MX"
• Added the possibility to handle routing errors while making handovers to Salesforce MIAW by selecting the Handover Status type in the Lookup Node
• Added a grace period to proactively terminate Salesforce MIAW chat sessions before the access token expires. This approach prevents unnecessary API calls to expired chat session

Refined Routing Behavior in Cognigy Live Agent

The Cognigy Live Agent routing logic for snoozed conversations (until next reply) has been optimized for high-volume environments to balance queues and reduce resolution latency:

• If an agent is at their conversation limit or their status is away or offline, the conversation is automatically unassigned.
• If the agent’s limit has not been reached and their status is online or busy, the conversation remains assigned.

This refinement ensures more efficient conversation distribution while preserving agent context where possible.

Other Improvements:

Cognigy.AI

• Deprecated the actions object in the Code Node. To achieve the same functionality, use api instead
• Introduced new versions of the call counter and conversation counter APIs
• Added the capability to retrieve all job configurations and the associated tools of AI Agents via the Cognigy.AI API endpoint GET /v2.0/aiagents/{aiAgentId}/jobs

Cognigy Webchat

• Updated Webchat v3 and Demo Webchat to v3.28.0

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