AI Settings

The AI Settings page lets you view your AI feature status, monitor usage, manage training data, and maintain conversation history.

Accessing AI Settings

The quickest way in is through the AI chat assistant:

1. Open the AI chat widget (the purple chat bubble in the bottom-right corner of the admin dashboard).
2. Use the View usage link in the chat panel to open the AI Settings page. (You will also see a link to AI Settings if you ever reach your daily request limit.)

You must have an admin account with AI access on your plan to open this page.

Plan and Access Status

At the top of the page, you will see your current AI access status:

• AI Enabled -- Your plan includes AI features and they are active.
• AI Not Included -- Your current plan does not include AI. You will see upgrade options with pricing and feature details.
• AI Features Disabled by Administrator -- The platform administrator has temporarily disabled AI across the entire system. Contact your system administrator if you believe this is an error.

Upgrading for AI Access

If your plan does not include AI, the settings page displays:

• The name and price of the recommended upgrade plan
• A list of AI features included (AI Chat Assistant, AI Content Generation, daily request limit)
• A direct link to the Billing page to upgrade

Usage Overview

The usage section shows how your church is consuming AI resources:

Today's Requests

A visual progress bar (labeled Today's Requests) shows how many of your daily requests have been used. The bar changes color as you approach your limit:

• Purple -- Normal usage (under 70%)
• Amber -- Approaching limit (70-90%)
• Red -- Near or at limit (90-100%)

Below the bar, you can see exactly how many requests remain for the day.

30-Day Statistics

Four summary cards display:

• Total Requests (30d) -- How many AI requests were made in the last 30 days
• Total Tokens -- The combined input and output tokens consumed
• Est. Cost (30d) -- The approximate cost of your AI usage
• Conversations -- Total number of AI conversations stored

AI Training Data

Training data is the custom information you provide to help the AI give accurate, church-specific responses. This is especially important for the member-facing chat, which relies on this data to answer questions about your church.

Adding Training Data

1. Click the Add Context button in the AI Training Data section.
2. Fill in the form:
   • Category -- Select the type of information (see categories below)
   • Title -- A descriptive name for this entry (e.g., "Sunday Service Times")
   • Content -- The actual information the AI should know (up to 5,000 characters)
   • Active -- Check this box to make the entry active immediately
3. Click Save Context.

Training Data Categories

Managing Training Data

Each training data entry shows:

• The title and category
• Active/Inactive status badge
• A preview of the content

You can:

• Toggle active status -- Click the eye icon to enable or disable an entry without deleting it. Inactive entries are not used by the AI.
• Delete -- Click the trash icon to permanently remove an entry (you will be asked to confirm).

Tip: Start with the basics: service times, location, and 5-10 frequently asked questions. You can always add more training data later as you identify gaps in the AI's responses.

Feature Status

The Features panel on the right side shows which AI capabilities are currently enabled:

• Admin Chat -- The conversational AI assistant for admins (the purple chat bubble)
• Member Chat -- The public-facing AI chat for members and visitors
• Content Generation -- The inline "Generate" buttons on content creation forms

Each feature shows an "Enabled" or "Disabled" badge. These are controlled at the platform level and cannot be toggled individually by church admins.

Your Plan Details

The Plan panel shows:

• Your current plan name and monthly price
• Whether AI is included in your plan
• Your daily request limit
• Your monthly token allocation
• How many requests remain for today

Clearing Conversation History

At the bottom of the right column, you will find the Clear Old History button under Maintenance. This deletes conversation records older than 30 days from the server to free up storage.

• Only conversations older than 30 days are removed
• Recent conversations (within the last 30 days) are preserved
• This applies to both admin and member conversations for your church
• You will be asked to confirm before the deletion proceeds

Note: Clearing server-side history does not affect the conversation history stored in your browser's local storage. To clear local history, use the "Clear All History" option inside the chat widget itself.

Quick Tips

The settings page includes a Quick Tips panel with reminders:

• Add service times, beliefs, and FAQs to improve AI responses
• Use the chat widget (bottom-right corner) to ask questions
• Look for "Generate" buttons on forms to create AI content
• Always review AI content before publishing

For TimelyChurch operators (self-hosting / platform staff)

Most church admins do not need this section. These notes are for the operator who deploys and configures TimelyChurch.

Provider configuration

AI features are served by xAI Grok. Provider settings live in config/ai.php and read from environment variables:

Pricing reference

The "Est. Cost (30d)" card uses the rates configured in config/ai.php:

• Input tokens: $5.00 per 1M tokens
• Output tokens: $15.00 per 1M tokens

These reflect grok-3 standard pricing as of 2026. Override with the AI_PRICE_INPUT / AI_PRICE_OUTPUT environment variables (or edit pricing.input_per_million / pricing.output_per_million) if xAI's rates change.

Platform-wide kill switch

A Super Admin can disable AI features for the entire platform — overriding per-church plan access — via the ai_module_enabled System Setting (Super Admin panel → System Settings). When disabled, churches see a configurable disabled-message (ai_disabled_message).

Where the model knows about your data

Module-specific context is built in app/Services/AI/GrokPromptBuilder.php. The chat assistant pulls up to the last 20 messages of a conversation, then trims that history to the AI_MAX_CONTEXT_TOKENS budget (default 4,000 tokens, estimated with a chars/4 heuristic, oldest messages dropped first) before sending it to the model. After 24 hours of inactivity a fresh conversation is started rather than continuing the old one.