Can I upload custom Agent images?
You can upload any 16:9 image with the persona facing the camera and framed in the center, mouth closed (permission needed for real people). Voice cloning is not supported.
How is Agent behavior controlled?
- Prompt Editor – set personality, tone, goals.
- Deflection Topics – restrict sensitive areas.
- FAQs – override answers with FAQ information.
Prompt Examples:
| Goal | Prompt Snippet (example) |
| Friendly onboarding | “Respond as a friendly, upbeat guide helping first time visitors.” |
| Professional support | “Maintain a formal, concise style. Prioritize accuracy and compliance.” |
What logic does the FAQ use?
- Close match > FAQ based answer returned
- Partial match -> the answer returned is based on the website pages info, LLM's knowledge and the knowledge base (no specific priority).
How do I add FAQs and responses?
Under the FAQ tab in the UI, you can manually input questions and responses one at a time, or bulk upload them using a CSV file.
How does the Agent respond to inappropriate and/or out of scope topics?
Deflection behavior is built in. If triggered, the agent politely declines to respond and redirects.