Settings
Every assistant has its own Settings tab. It holds the assistant's name, the controls for how the AI analyzer tags its conversations, the live chat and escalation setup, and assistant-level secrets.
Open an assistant and click Settings in the left sidebar to get there.
Changes here are not saved as you type. The header shows Unsaved changes while you have edits pending, and you commit them with Save changes.
General
Name How this assistant appears across the dashboard. It is for your own reference and is not shown to end users.
Conversation tagging
This is the part of the page worth understanding, because it decides what the AI analyzer records about every conversation, and those signals feed straight into the Dashboard and the Conversations filters.
Auto-tag frustrated conversations Applies the Frustrated tag when the analyzer's frustration score crosses the threshold below. You can still add or remove that tag by hand, and doing so does not change the setting.
Frustration sensitivity A slider from 0 to 1, shown with its current value in the label. Conversations are tagged Frustrated at or above the score you pick, so lower is more sensitive. It appears only when auto-tagging is on. If you are getting Frustrated tags on conversations that read as ordinary, raise it. If genuinely unhappy conversations are slipping through untagged, lower it.
Detect feature requests Identifies when a conversation contains a feature request. This is what powers the Feature requests figure on the Dashboard and the Top feature requests list, so turning it off empties both.
Analyzer instructions Optional free text, up to 2000 characters, applied on top of the built-in rules. Use it to tell the analyzer things it cannot infer: name your product so look-alike asks are classified correctly, or say that billing and refund disputes should count as high frustration. This is the highest-leverage field on the page if the automatic tagging does not match how your team reads its own conversations.
Live chat & support
Sets how a visitor reaches a human: the Provider choice between Yedric's built-in chat and an external handoff, business hours, and, with the built-in provider, staffing and routing. These are covered in full on Live Chat.
A Canned Responses section also appears here when the built-in live chat provider is selected.
Advanced: Secrets
Each assistant can hold its own encrypted secrets, listed under Advanced. Use these when different assistants need different credentials, for example when each one connects to a different customer account or environment.
Values are write-only. Once saved they are encrypted at rest and never returned to the dashboard, so to change one you enter a new value rather than revealing the old one.
Organization-level secrets are listed alongside them, read-only, so you can see which names an assistant-level secret would override. Assistant-level secrets win over organization secrets with the same name.
Reference a secret by name in two ways, depending on where it is going:
{{name}}
Your instructions, a tool's base URL, or a tool URL.
{{secret:name}}
Authentication headers and query parameters on a custom MCP connection.
Substitution happens server-side at chat time, so the raw credential never appears in your configuration. Single braces ({name}) are still accepted for backward compatibility, but prefer the double-brace form.
Note: Managing secrets requires the Owner or Admin role. A Member opening this section sees a permission message rather than an empty form.
Where other settings live
Not everything about an assistant is on this tab:
Its instructions are on Instructions.
Its model, first message, and knowledge base live on the assistant's own pages under Your Assistant.
Its widget appearance, escalation toggle, and mascot are on Widget.
Its actions and connections are on Actions.
Organization-wide settings are on Organizations.
Last updated