AI Model Configuration

IMPORTANT

This feature is only available to Super Administrators (platform-level access).

Overview

The AI Configuration panel allows super admins to manage which AI models are used for different tasks across the platform.

Accessing AI Config

1. Log in as a Super Admin
2. Navigate to Platform Admin → AI Configuration

Configuration Options

Each AI task can be configured independently:

Task Types

Task	Description
Content Generation	Creating social posts, blog content
Brand Analysis	Analyzing websites for brand extraction
Refinement	Quick refine actions on text
Chat Sessions	Multi-turn AI conversations
Sentiment Analysis	Analyzing inbox item sentiment
Script Generation	Video script creation

Model Settings

For each task, you can configure:

Setting	Description
Provider	AI provider (OpenAI, Anthropic, Google, etc.)
Model	Specific model (gpt-4o, claude-3-5-sonnet, gemini-2.0-flash, etc.)
Max Tokens	Maximum output length
Temperature	Creativity level (0.0 = deterministic, 1.0 = creative)
Input Cost per 1M	Cost tracking for input tokens
Output Cost per 1M	Cost tracking for output tokens
Enabled	Toggle individual tasks on/off

Updating Configuration

1. Select a task from the list
2. Modify settings as needed
3. Click Save
4. Changes take effect immediately (cache is cleared)

Price Synchronization

Keep pricing accurate with the Sync Prices feature:

1. Click Sync Prices
2. System fetches latest pricing from AI providers
3. Local configurations are updated automatically
4. A summary shows which models had price changes

TIP

Run price sync monthly to ensure accurate usage billing.

AI Gateway

The AI Gateway provides:

• Model Discovery: Browse available models from all providers
• Real-time Pricing: Current costs per model
• Fallback Handling: Automatic failover if primary model is unavailable

Viewing Available Models

1. Go to Platform Admin → AI Gateway
2. Browse models by provider
3. See capabilities, pricing, and status

Best Practices

1. Use appropriate models for tasks — Simpler tasks don't need the most expensive models
2. Set reasonable max tokens — Prevents runaway costs
3. Monitor usage — Check AI Usage dashboard regularly
4. Test before switching — Verify quality when changing models
5. Keep prices synced — Ensures accurate cost tracking

Troubleshooting

Model Not Responding

1. Check if model is enabled
2. Verify provider credentials are valid
3. Check provider status page for outages

Unexpected Costs

1. Review max tokens setting
2. Check usage logs for high-volume users
3. Consider model downgrades for non-critical tasks