Loading documentation...
Understand Operator's safety boundaries, medical/legal disclaimers, and out-of-scope handling.
Operator enforces strict runtime guardrails:
1. Factual Grounding: Answers to policy or procedural questions must cite chunks from your uploaded Knowledge Base or seeded FAQs.
2. Prompt Injection Mitigation: User input is isolated inside formatted conversation turns, preventing malicious attempts to override system instructions.
3. SSRF Blocking: The website crawler blocks loopback (127.0.0.1), local subnets (10.0.0.0/8, 192.168.0.0/16), and cloud metadata APIs (169.254.169.254).
For healthcare and legal practices: - Operator will never diagnose illnesses, interpret laboratory results, or prescribe treatments. - Operator will never offer legal counsel or interpret legal agreements. - In medical emergencies, Operator instructs callers to hang up and dial emergency services (911 or local equivalent).
When an inquiry cannot be answered from your Knowledge Base: - The AI will politely state: *"I don't have that specific information in my records. Let me connect you with our team."* - It will NOT guess unlisted pricing, fake discounts, or non-existent services.
To maintain enterprise security, Operator will NEVER: - Modify staff calendar events that were created outside Operator. - Process bank transfers or unauthorized card charges without user checkout interaction. - Delete past conversation logs or customer records autonomously.