Skip to content

Settings reference

Everything you configure in RelyAI lives under one screen: open the AcelleMail sidebar group RelyAI → Settings. A left rail lists nine tabs. This page is a quick reference to each tab — what it controls, the main fields it holds, and the defaults you start with — and points you to the deeper guides when you need a full walkthrough.

The RelyAI Settings tabsRelyAI → Settings

The nine tabs at a glance

TabWhat it controls
Engines & KeysProviders, models, API keys
Feature FlagsEnable or disable AI capabilities
Quotas & LimitsPer-task and per-customer caps
PermissionsRead-only matrix of who can use which AI tool
Brand VoiceGlobal tone and forbidden words
AppearanceChatbox launcher position and theme
AdvancedDiagnostics and log retention
LicensePlugin license key and status
Integrity & UpdatesVersion, integrity checks, auto-update

Engines & Keys

Add the AI providers and models RelyAI calls. Each engine holds a Name (for example "OpenAI GPT-4o"), Provider, Model, Max Tokens, and an API key (placeholder sk-… / sk-ant-… / (leave blank for Ollama); editing shows "Leave blank to keep the existing key"). Use Test Connection to verify a key — it steps through "Testing…" to Connection successful — then mark one engine as Default; that engine handles all AI calls. You can also let customers bring their own key with Allow customers to use their own API key. See Connect an engine for the full flow.

Feature Flags

Every AI capability is an on/off switch here — the support chatbox, agent chat, coaches, rewrite, the subject A/B generator, the observability surfaces, and more. Several ship off by default (agent chat, the four coaches, campaign pause and template tools, automations) so you opt into them deliberately. Turning an observability surface off hides it but never stops the underlying logging. See Feature flags and permissions.

Quotas & Limits

Rate caps that keep AI cost and load predictable. Defaults are 20/minute and 500/hour per task, plus an optional per-customer daily cap (default 0 = uncapped). Provider-side rate limits are also respected. See Quotas, limits and credits.

Permissions

A read-only matrix of who can use which AI tool. It is introspection only — it reflects what AcelleMail already enforces and never changes it. Customers can open their own "View my permissions" view when Permission transparency is on.

Brand Voice

Global copy rules applied to rewrite and generated text: tone adjectives (default "warm, witty, confident, plain-spoken"), forbidden words (default "synergy, leverage, click here, free!!"), and a voice sample. Settings resolve per-user, then global, then a built-in default. See Brand voice and appearance.

Appearance

Style the chatbox launcher: launcher position (bottom-right by default, top-right, in the top bar, or beside search), color theme (pastel violet by default, teal, orange, green, indigo), launcher glyph (sparkles by default, plus chat bubble, magic wand, star, helper, bolt, brain, robot, concierge, forum), and unread-dot animation (pulse by default, static, or hidden).

Advanced

Diagnostics and log retention, agent runtime limits (max loop iterations, tool-call timeout, confirm/undo windows), and the URL policy for the hallucination defence (rewrite with a menu hint by default, strip, or mark unverified). Reasonable defaults ship, and most admins never change these.

License

Your plugin license key and status. Activate, re-activate, or deactivate the key, and see your current license type and updates entitlement. See Licensing.

Integrity & Updates

The Installed version, Check for updates and Update now controls, advisory file-integrity status, and the list of database snapshots taken before each update. See Updating, Backup and restore, and File integrity.

Documentation for the Relyweb app catalogue.