<aside>
Description
The LLM Manager is the provider configuration interface for connecting and managing AI engines in SimpleT. The feature covers API key management, model selection, engine activation, and default engine configuration across OpenAI, Anthropic, and Google providers.
</aside>
What you can do:
- Connect multiple AI providers with secure API key configuration
- Switch between different models for various use cases
- Control costs by activating or deactivating engines
- Set default engines for a consistent user experience
Getting Started
1. Access LLM Manager
Navigate to Setup → LLM Manager from the main menu. You need the MANAGE_SETUP permission to access this feature.
2. Connect your first LLM
Select Connect new LLM to add your first AI engine:
- Name: Give your engine a descriptive name (for example, “Production GPT‑4”)
- AI type: Select your provider (OpenAI, Anthropic, or Google)
- API key: Enter your provider’s API key securely
- Model: Choose from the available models for your selected provider
- Status: Set to Active to enable the engine
- Default: Optionally set this engine as the default
<aside>
<img src="/icons/lock-keyhole_blue.svg" alt="/icons/lock-keyhole_blue.svg" width="40px" />
Security note
API keys are encrypted before storage and never exposed in logs or exports. Always use API keys with appropriate permissions and rate limits.
</aside>
Configuring AI Engines
Engine list view
The main interface lists all configured engines in a table that shows:
- Engine name and AI type
- Selected model
- Status (Active or Inactive)
- Default indicator
- Creation and last modification timestamps
Editing engines
Use the menu icon (⋮) on any engine row to:
- Edit: Update name, API key, or model
- Delete: Remove the engine (system default engines cannot be deleted)
- Set as default: Make this the primary engine
Best Practices
1. API key configuration
- Use separate API keys for development and production
- Rotate keys regularly (quarterly is recommended)
- Set appropriate spending limits with each provider
2. Model selection
- Use GPT‑4 for complex analysis and reasoning tasks
- Use Claude for creative content and long‑form generation
- Use GPT‑4o mini for simple, high‑volume tasks
- Consider cost versus performance for each use case
- New models are continuously tested and added as they prove reliable
3. Multi‑provider strategy
- Configure multiple providers for redundancy
- Use different providers for different features
- Monitor provider status pages for outages
- Test failover procedures regularly
Troubleshooting
API key issues
- “Invalid API Key” – Verify the key is correct and active with the provider
- “Rate Limit Exceeded” – Check the provider dashboard for current limits
- “Insufficient Quota” – Add billing information or increase limits with the provider
Model availability
- Model not showing – Ensure the API key has access to that model
- Model deprecated – Update your configuration to a newer version
- Region restrictions – Some models are only available in specific regions
<aside>
<img src="/icons/light-bulb_blue.svg" alt="/icons/light-bulb_blue.svg" width="40px" />
Pro tip
Default system engines are available as fallbacks but have limited capabilities. Always configure your own engines for production use.
</aside>
Integration with Other Features
Engines configured in the LLM Manager are automatically available in:
- Prompt Builder: Select engines when creating prompts
- SimpleT Salesforce Components: Use configured engines across all Salesforce AI components
Remember to configure appropriate engines before using AI features. If no engine is explicitly selected, the system uses the default engine.
<aside>
Quick View
End User Documentation