SUPERWISE v1.25.0: Advanced AI Agents, Guardrails, and Observability

Five new guardrails: competitor check (block discussion of named companies), correct language (allowed/disallowed languages), string check (reject expressions), jailbreak detection (prompt manipulation), and PII detection with dozens of categories including credit card, phone numbers, and identifiers.

Users can try to manipulate prompts to bypass guardrails. Jailbreak detection flags those attempts so you block or audit them. Governed AI requires that the platform defends the rules, not only the data.

SUPERWISE supports both chat-flow (single-agent) and agent-flow (multi-agent) Flowise builds. You register the implementation and get out-of-the-box dashboards, customized queries, and full guardrail coverage for each deployed agent.

Agents that handle user input can leak names, emails, or payment data. Built-in PII detection in SUPERWISE stops that at the platform layer so every integration benefits without custom code.