FAQ

Quick answers to common questions about RelyAI for AcelleMail.

Which AI providers are supported?

RelyAI supports three provider options:

• Supported providers: OpenAI, Anthropic, and Ollama (self-hosted or Ollama Cloud).
• Default engine: Configure multiple engines and set one as Default for all AI calls.
• Popular models: GPT-4o mini, Claude Haiku 4.5, and Llama 3.1 8B.

See Connect an engine.

Do I need my own API key?

Yes. You provide API keys through either method:

• System key: Set one default key under Settings → Engines & Keys for all users.
• Customer keys: Enable Allow customers to use their own API key so customers add keys under Account → AI and use their own provider quota.

Does it work without a license?

No. All AI features require an active RelyAI license:

• Requirement: A license verified as Active for your domain.
• Unlicensed state: AI surfaces remain disabled and prompt administrators to enter a license key.

See Licensing.

Is the chatbox read-only, or can it act?

Both configurations are supported:

• Read-only (Default): Functions as a knowledge-base support bot that answers product questions.
• Agent mode: Enable Agent chat under Feature Flags to allow actions (creating lists, tagging subscribers, automations).
• Safety: Prompts for confirmation on destructive actions and provides an Undo button (or Cmd+Z).

See The chatbox and Agent tools.

Can the AI change things without asking?

No. RelyAI enforces strict guardrails:

• Confirmation required: Destructive actions require explicit confirmation before execution.
• Undo support: Most agent actions can be reversed within the undo window.
• Blocked fields: Passwords, card numbers, CVVs, OTPs, and API keys cannot be modified.
• Granular flags: Each tool is controlled by a dedicated feature flag (several disabled by default).

What data is logged, and can I audit it?

All AI activity is logged across the observability suite:

• Usage Dashboard: Request volume, token usage, and cost tracking.
• Conversations: Full session message replays.
• Audit Viewer: Detailed tool invocation logs with CSV export.
• Feedback & Self-Improve: User ratings, error clusters, and knowledge gap reports.

See Observability.

How are costs controlled?

Cost control is managed via three mechanisms:

• Quotas & Limits: Enforce per-task rate caps (default 20/min, 500/hour) and daily customer limits.
• AI Credits: Meter AI usage against AcelleMail subscription plans.
• Customer keys: Offload API costs by letting customers use their own provider keys.
• Local models: Self-hosted Ollama models incur zero per-token API cost.

See Quotas, limits and credits.

Does it modify AcelleMail core?

No. RelyAI installs as a standard AcelleMail plugin and mounts UI surfaces via AcelleMail's theme without modifying core files. See Installation.

How are updates delivered?

Updates run in-app from RelyAI → Settings → Integrity & Updates:

• In-app update: Click Update now for an automated download, verification, backup, and apply process with automatic rollback on failure.
• Manual update: Upload the new plugin package under Admin → Plugins → Upload Plugin.

See Updating.

What happens if my license lapses?

License lapse handling depends on your license type:

• Leased licenses: AI features pause until renewal and re-activation. AcelleMail data remains intact.
• Lifetime licenses: Installed versions operate permanently. An active Support & Updates plan is required only for new releases.

See Licensing.

Related pages

• Licensing — how the license gate works
• Connect an engine — OpenAI, Anthropic, Ollama
• Agent tools — what the agent can do, safely
• Observability — logging and audit
• Quotas, limits and credits — cost control
• Troubleshooting — when something isn't working