Agent settings
Find the right settings for answers, the conversation interface and security.
Open copy menu
Choose settings according to what you want to change. Agent answers, knowledge sources and the conversation interface on your website are separate areas.
Open the agent and choose the area you want to change. Overview, Appearance and Actions share the same conversation preview.
| What you want to change | Start here | Detailed guide |
|---|---|---|
| Model or instruction text | Overview → Model or Instructions | Task and response style |
| Response length or behavior when information is missing | Settings → AI | Check the effect in a new test conversation. |
| Reply language, address form or terminology | Settings → Languages | Languages and localization |
| Widget content and colors | Appearance | Appearance and welcome messages |
| Information used in answers | Data sources | Data source management |
| Contact form or API action | Actions and the action’s settings | Lead collection , HTTP actions |
| Allowed domains or identity verification | Settings → Security | Publishing , identity verification |
| Support team availability | Settings → Human support | Human support setup |
The primary colour is used by the header, user messages, and primary buttons. The launcher can have its own colour, and its icon contrast is selected automatically. A white header therefore does not force the launcher to be white.
Current page recognition
Section titled “Current page recognition”The Understand the visitor’s current page switch in AI settings helps the chatbot interpret referential questions. For example, “What does this include?” on a pricing page can be associated with that knowledge-base page.
The feature is off by default. It can be enabled when the chatbot has at least one ready web page in its knowledge base. You do not need to change the embed code or configure the customer site. The chat widget checks the page for each message, so navigation in a single-page application is picked up automatically.
Recognition is used only when the browser page matches exactly one web page in the knowledge base. If there is no match or the match is ambiguous, the chatbot uses its normal knowledge search. Page-aware search does not displace the strongest result from the normal search.
Privacy and boundaries
Section titled “Privacy and boundaries”- The chat widget removes query parameters and the fragment, then derives a fixed identifier from the page origin and path in the browser.
- The chat widget does not send Aihio the raw page URL, page title, text, headings, forms, input fields, or other DOM data.
- The identifier is not stored with the conversation or written to application logs. It is used only to scope knowledge retrieval for that message.
- Keep the feature disabled on sites where the URL path identifies a person or contains sensitive information. If it is enabled in such an environment, assess the processing and describe it in the site’s own privacy notice.
The feature is optional and does not affect normal chatbot behaviour while disabled.
Test recognition with the chat widget published on the website. The dashboard preview does not have the customer page URL, so it uses normal knowledge search.