Skip to main content
Version: 6

AI Configuration

All AI configuration in Blesta lives under Settings > System > AI. This page documents each field and the privacy acknowledgement workflow.

API configuration​

API Key​

Your Blesta AI service API key. Two ways to set it:

  • Manual entry: paste an existing key into the API Key field and click Update Settings. Validation runs against the Blesta AI service and you'll see "API key validated and saved" on success, or "Invalid API key" otherwise.

  • Fetch From My Blesta Account: click the Fetch From My Blesta Account button to open a modal, enter your account.blesta.com username, and click Authenticate. Blesta uses your installed license key plus the username to retrieve (or generate) an API key automatically.

    Notes on the fetch flow:

    • Your license key must already be configured under Settings > System > License — the fetch will fail with "No Blesta license key found" otherwise.
    • If a key has already been issued for your account, it is revoked and a new key is issued. Update any other Blesta installs that were using the old key.

After a valid key is saved, the AI Settings page also displays:

  • API Status — Active when the gateway is reachable.
  • Remaining Credits — credit balance on your Blesta AI account.
  • Credits Used (This Month) — usage running total.
  • Last API Call — timestamp of the most recent successful request, or "No calls yet".

Default model​

Default Model​

The model used by default for chat completions. Available models are populated from the Blesta AI service after a valid API key is saved. Choose any of the listed models from the dropdown.

The chatbot lets staff override the default per-conversation, so the Default Model setting only applies to the initial chatbot model and to non-chatbot AI features (email/HTML/package description generators).

Model parameters​

Temperature​

Controls how deterministic vs. creative the AI's responses are. Range: 0.0 to 2.0.

  • Lower values (e.g. 0.2) produce more focused, repeatable responses. Good for factual replies and structured generation.
  • Higher values (e.g. 1.2) produce more varied and creative responses. Good for marketing copy and brainstorming.
  • The default is balanced for general use.

Max Tokens​

The upper limit on the number of tokens the AI is allowed to generate in a single response. Lower values save credits but may cause longer responses to be truncated; higher values allow long-form output at higher cost per request.

If the AI returns a truncated response, the chatbot displays a notice with a link back to this setting.

System prompts​

Global System Prompt (Default)​

The default instructions sent to the AI model with every request — sometimes called a "system prompt." It defines the AI's persona, tone, and constraints across all features.

Most features (email templates, support ticket summarize, etc.) layer their own feature-specific prompt on top. The Global prompt sets the baseline.

Enabled features​

Toggle individual AI features on or off. Toggling a feature off here hides the related AI buttons throughout the admin UI without disabling AI globally.

  • Package Descriptions — Generate marketing-style descriptions on the Packages add/edit page.
  • Email Templates — AI-assisted edits and improvements in the email template editor (see Email Templates).
  • Chatbot — The AI chatbot in the top bar (see Chatbot).

Staff Group Access (Chatbot)​

Multi-select list of staff groups permitted to use the chatbot. Hold Ctrl (or Cmd on macOS) to select more than one group. Staff in groups not selected here will see "Access Restricted" if they navigate to the chatbot.

This is independent of the per-staff icon-bar visibility setting (below).

Chatbot icon visibility (per staff member)​

Each staff member can additionally hide the AI chatbot icon from their own icon bar under My Info — useful if you have access but rarely use the feature. This is purely a UI preference; the chatbot itself remains reachable by direct URL.

Email Template Context Settings​

Controls how much context Blesta sends to the AI when generating email template content (see Email Templates).

Relationship Depth​

How many levels of model relationships to traverse when assembling context for the AI. Range: 1 to 5. Default: 2.

Example for an invoice template:

  • Depth 1: Just the invoice fields.
  • Depth 2: Invoice + client.
  • Depth 3: Invoice + client + contacts.

Higher depths include more related data but use more tokens per request, increasing credit cost.

Include Schema Definitions​

When checked, includes field type information (e.g. "amount is a decimal currency value") in the AI context. Useful when generating templates that need precise formatting.

Include Example Data​

When checked, includes sample data values alongside field names, so the AI can see formats like "ABC-12345" or "2026-01-15" rather than guessing. Recommended for most cases.

Privacy notice​

Beta Feature Notice​

A persistent banner at the top of the AI settings page reminds you the feature is in beta and may produce unexpected results.

Privacy Notice​

Explains that AI requests are sent to third-party providers (OpenAI, Anthropic, etc.) for processing, that the Blesta AI service does not store conversation data, but that Blesta cannot control upstream provider data handling.

Privacy Acknowledgment​

You must check "I have read and understand the privacy notice above" before saving AI settings for the first time. The page records when you last acknowledged the notice — re-acknowledgement may be required if the notice text changes in a future release.

Trying to save without acknowledging produces an error: "You must agree to the privacy notice before saving."

Last updated on