Presets
Presets in LibreChat allow you to save and reuse conversation configurations, including model selection, endpoint settings, and parameters. This saves time when starting new conversations with frequently used settings.What Are Presets?
A preset is a saved configuration that includes:- Endpoint - The AI service provider (OpenAI, Anthropic, Google, etc.)
- Model - The specific AI model to use
- Model Parameters - Temperature, top P, frequency penalty, presence penalty, etc.
- Endpoint Options - Provider-specific settings and features
- Conversation Settings - Additional configuration based on the endpoint
Accessing Presets
Opening the Presets Menu
- Look for the book icon (BookCopy) in the chat header
- Click to open the presets menu
- A popover displays your saved presets and options
Creating Presets
Save Current Conversation as Preset
-
Configure your conversation settings:
- Select your preferred endpoint
- Choose a model
- Adjust parameters (temperature, max tokens, etc.)
- Configure any endpoint-specific options
- Click the presets menu icon
- Select Save as Preset or use the save option in the menu
- Enter a descriptive name for your preset
- Click Save
Preset Naming Best Practices
- Be descriptive - “GPT-4 Creative Writing” vs “Preset 1”
- Include key settings - “Claude Code Review (Temp 0.3)”
- Use categories - “[Work] Analysis” or “[Personal] Brainstorm”
- Note special configs - “Gemini Long Context” or “GPT-4 Vision Enabled”
Using Presets
Applying a Preset to New Conversation
- Start a new conversation or open an existing one
- Click the presets menu icon
- Browse your available presets
- Click on a preset name to apply it
- The conversation updates with all saved settings
Applying a Preset to Existing Conversation
You can change the settings of an active conversation:- Open the presets menu
- Select a different preset
- All settings update immediately
- Your conversation continues with the new configuration
Editing Presets
Modifying Preset Settings
- Open the presets menu
- Click on the preset you want to edit
- The preset edit dialog opens with current settings
- Modify any settings:
- Preset Name - Update the title
- Endpoint - Change the AI provider
- Model - Select a different model
- Parameters - Adjust temperature, tokens, penalties, etc.
- Advanced Options - Configure endpoint-specific features
- Click Save to update the preset
Preset Edit Dialog
The edit dialog shows:- Preset name field - Edit the display name
- Endpoint selector - Dropdown to change provider
- Model settings - Based on selected endpoint
- Parameter controls - Sliders and inputs for model parameters
- Action buttons - Export or Save changes
Auto-Validation
When you change endpoints in a preset:- LibreChat automatically validates the model selection
- If the current model isn’t available on the new endpoint, it selects the first available model
- Settings incompatible with the new endpoint are adjusted or hidden
Managing Presets
Setting a Default Preset
Designate a preset as your default to automatically apply it to new conversations:- Open the presets menu
- Find the preset you want as default
- Click the options menu (three dots) for that preset
- Select Set as Default
Deleting Presets
- Open the presets menu
- Hover over the preset you want to delete
- Click the options menu (three dots)
- Select Delete
- Confirm the deletion in the dialog
Clearing All Presets
To remove all saved presets at once:- Open the presets menu
- Look for the Clear All Presets option (usually at the bottom)
- Confirm you want to delete all presets
Importing and Exporting Presets
Exporting a Preset
Save a preset as a JSON file for backup or sharing:- Open the presets menu
- Click on the preset you want to export
- In the edit dialog, click Export
- A JSON file downloads with the preset configuration
Importing a Preset
Load a preset from a JSON file:- Open the presets menu
- Click Import or use the file upload option
- Select the preset JSON file from your computer
- The preset is added to your presets list
- Verify the imported settings are correct
Sharing Presets
Preset JSON files are portable and can be:- Shared with team members for consistent settings
- Backed up to cloud storage
- Version controlled for different use cases
- Distributed in documentation or tutorials
Preset Configuration Details
Endpoint-Specific Settings
Different endpoints have unique configuration options:OpenAI
- Model selection (GPT-3.5, GPT-4, etc.)
- Temperature
- Top P
- Frequency Penalty
- Presence Penalty
- Max Tokens
Anthropic (Claude)
- Model selection (Claude 3 variants)
- Temperature
- Top P
- Top K
- Max Tokens
Google (Gemini)
- Model selection
- Temperature
- Top P
- Top K
- Max Output Tokens
Model Parameters Explained
Temperature (0.0 - 2.0)
Controls randomness in responses:- 0.0-0.3 - Focused, deterministic (good for code, factual answers)
- 0.4-0.7 - Balanced creativity and coherence
- 0.8-1.0 - Creative, varied (good for brainstorming, writing)
- 1.1-2.0 - Very creative, potentially chaotic
Top P (0.0 - 1.0)
Controls diversity via nucleus sampling:- Lower values - More focused, predictable
- Higher values - More diverse, creative
- Common setting: 1.0 (disabled) or 0.9
Frequency Penalty (-2.0 - 2.0)
Reduces repetition of token sequences:- Positive values - Discourage repetition
- Negative values - Allow more repetition
- Common setting: 0.0 to 0.5
Presence Penalty (-2.0 - 2.0)
Encourages topic diversity:- Positive values - Encourage new topics
- Negative values - Stay focused on current topics
- Common setting: 0.0 to 0.5
Common Preset Use Cases
Code Assistant
Configuration:- Model: GPT-4 or Claude 3 Opus
- Temperature: 0.2-0.3
- Top P: 0.9
- Purpose: Accurate, focused code generation and debugging
Creative Writing
Configuration:- Model: GPT-4 or Claude 3 Sonnet
- Temperature: 0.8-1.0
- Top P: 0.95
- Purpose: Imaginative storytelling and content creation
Technical Analysis
Configuration:- Model: GPT-4
- Temperature: 0.3-0.5
- Frequency Penalty: 0.2
- Purpose: Detailed technical documentation and analysis
Brainstorming
Configuration:- Model: Claude 3 Sonnet or GPT-4
- Temperature: 0.9-1.2
- Top P: 0.95
- Presence Penalty: 0.3
- Purpose: Diverse ideas and creative problem-solving
Research Assistant
Configuration:- Model: GPT-4 with browsing or Claude with extended context
- Temperature: 0.4-0.6
- Max Tokens: High (for comprehensive responses)
- Purpose: In-depth research and information synthesis
Tips and Best Practices
Creating Effective Presets
- Start with defaults - Begin with standard settings and adjust based on results
- Test thoroughly - Verify preset behavior with sample prompts before regular use
- Document your presets - Add notes about optimal use cases in the preset name or externally
- Update regularly - Refine presets as you learn what works best
Organizing Presets
- Use naming conventions - Consistent prefixes or categories
- Limit total presets - Keep only actively used presets (export others for backup)
- Group by purpose - Work, personal, creative, technical, etc.
- Archive old versions - Export outdated presets before deleting
Performance Considerations
- Higher temperature = more variability - Results may be less consistent
- Max tokens affects cost - Higher limits use more credits per response
- Model selection impacts speed - Larger models are typically slower
Security and Privacy
- Review before sharing - Exported presets may contain metadata
- Don’t include sensitive data - Presets shouldn’t contain personal information
- Verify imported presets - Check settings before using presets from others
Troubleshooting
Preset Not Loading
- Verify the endpoint is available
- Check that the model exists and you have access
- Try re-importing if it’s an imported preset
Settings Not Saving
- Ensure all required fields are filled
- Check for validation errors in the edit dialog
- Try refreshing the page and saving again
Import Failed
- Verify the JSON file is valid and not corrupted
- Check the file was exported from a compatible LibreChat version
- Ensure the file format matches expected preset structure
Model Not Available
- Confirm your account has access to the selected model
- Check if the model has been deprecated or renamed
- Try selecting a different model for the same endpoint
Keyboard Accessibility
Preset management is fully keyboard accessible:- Tab - Navigate through preset options and menu items
- Enter/Space - Select preset or activate buttons
- Escape - Close preset menu or dialogs
- Arrow keys - Navigate dropdowns and select options