Overview
Block, strip, or redact risky content; enforce allowlists for tools and models.
Prerequisites
None.
Permissions required
Downloads & Templates
Steps (3)
-
1
Enable detectors
Security → Guardrails: enable PII, credentials, jailbreak and toxicity detectors.
Tips
—
Validation
—
Success criteria
—
-
2
Set actions
Choose per-signal actions: block, redact, or warn; configure redaction masks for logs.
Tips
—
Validation
—
Success criteria
—
-
3
Model & tool allowlists
Permit only approved models and tools per queue/brand; cap budgets.
Tips
—
Validation
—
Success criteria
- Zero policy violations during smoke tests; logs show redaction.
About this guide
Chat Core decides what happens next in any live conversation. It stitches identity across channels, maintains durable sessions, detects intent and sentiment, calls tools and knowledge bases, and chooses between automated replies or human handoff with clear SLAs. Operators get routing/skills rules, cost & safety guardrails, experiments, and an orchestration trace for explainability.
Tightly integrated with PulseGate for transport, LiveConnect for real-time inbox, Agent Desk for tickets, FlowForge for automation, Chronicle for audit-ready timelines, and InsightLake for analytics—so you can ship omnichannel support that scales safely.