Model Presets
Model presets allow you to offer different AI configurations to users. Instead of locking everyone into a single model and settings, you can create presets that let users choose the right configuration for their specific task—whether that's a faster model for quick questions or a more capable model for complex analysis.
What are Model Presets?
Presets are saved configurations that bundle together a specific AI model, its parameters like temperature and max tokens, and a user-friendly name and description. When users start a conversation, they can choose from available presets based on what they need to accomplish.
Creating Presets
To create a preset, open your agent in the editor and navigate to the AI Presets section. Click Create Preset and configure a name that users will see, a description explaining when to use this preset, the AI model to use, and any model-specific parameters like temperature or max tokens.
Default Preset
Every agent has a default configuration that's used when no preset is explicitly selected. This default is defined in the agent's base settings and is always available to users. Design your default to work well for the most common use case—users shouldn't need to change it for typical interactions.
Preset Parameters
Temperature controls how random or creative the AI's responses are. Lower values (0.0-0.3) produce more focused, consistent, and predictable responses—good for factual tasks where you want the same question to get similar answers. Medium values (0.4-0.7) offer a balance between consistency and variety. Higher values (0.8-1.0) make responses more creative and varied, which works well for brainstorming or creative writing but may produce less predictable results.
Max Tokens sets the maximum length of responses. Set this based on how long you expect typical outputs to be. Higher values allow longer responses but also consume more resources, so find a balance that covers your use case without being wasteful.
User Experience
When users start a conversation, they see a preset selector with the default option pre-selected. They can click to choose a different preset if their needs differ from the typical case. The selected preset applies for the entire conversation.
Users can see each preset's name and description, helping them make informed choices. A visual indicator shows which preset is currently active.
Use Cases
Presets work well for several scenarios.
Speed vs. quality tradeoffs let users choose based on urgency. A "Quick Answers" preset might use a faster, smaller model for simple questions, while a "Detailed Analysis" preset uses a more capable model when accuracy matters more than speed.
Task-specific configurations optimize settings for different work. Creative writing benefits from higher temperature settings, code generation works better with lower temperature for more predictable output, and general chat falls somewhere in between.
Cost management can guide users toward appropriate resources. A "Standard" preset handles most tasks efficiently, while a "Premium" preset reserves more capable models for genuinely complex work.
Best Practices
Give presets clear, descriptive names that communicate their purpose. Names like "Quick Answers" and "Detailed Analysis" help users understand what to expect. Avoid technical names like "GPT-4" or generic labels like "Option 1" that don't convey meaningful differences.
Write descriptions that explain when to use each preset, what tasks it's best suited for, and any tradeoffs users should consider. Help users make informed choices without requiring them to understand AI model internals.
Set reasonable defaults that work well for most cases. If users constantly need to switch presets for basic tasks, reconsider your default configuration. Special presets should be for special needs, not everyday use.