Skip to main content

Cognigy.AI

Planned on the Alice TypeScript SDK, because Cognigy runs custom code in Node.js sandboxes. An Endpoint Transformer will guard every message in and out of an endpoint with handleInput and handleOutput. A custom Extension node will drop into a single flow and let the builder branch on the verdict. Both evaluate synchronously and support block and mask in each direction.

Status: Coming soon · Evaluates: Prompts, Responses · Vendor: NICE

User input via evaluatePrompt, bot output via evaluateResponse. Cognigy sessionId and userId pass through for tracing.

Setup​

Configuration: Alice API key (Cognigy connection), app_id map per endpoint

  1. Endpoint Transformer: attach handleInput and handleOutput to the endpoint for an org-wide baseline.
  2. Extension node: import the extension, store the API key in a Connection, place the node where you want a verdict to branch on.
  3. Map each Cognigy endpoint to an Alice app_id so the right policies apply.

Good to know​

Transformers allow one outbound request per execution, so the SDK runs with maxRetries: 0. SDK errors fail open by default and can be set to fail closed.