2.2 Configure LLM (Advanced)

No configuration needed by default. You can configure more powerful models separately to meet more personal preferences.

MiniTavern currently mainly uses LLM (Large Language Model) classification, so it's called Configure LLM in the Settings page.

Default Zero Configuration Available

MiniTavern currently provides free quota and default configured models for new users.

Now you can start conversations without configuring LLM, just import character cards. If you're not satisfied with the current default conversation model, you can select other models on the Configure LLM page.

Check Quota

In the settings page, you can check remaining quota in the user type and available quota area below the username.

Default Model

The default model configured by MiniTavern is the Grok mainstream model. In the available model list at the bottom of the Configure LLM page, the model with a checkmark on the right is the currently selected model.

Rate Explanation

Due to cost issues, we have set rate limits for each model. The (2x) in the model subtitle means that each request will consume 2x the request quota.

Add Your Own LLM

Step 1: Enter Configure LLM Page

• Click the third tab "Settings" in the bottom navigation bar
• Select "Configure LLM"

Step 2: Select API Provider

Click the "Provider" dropdown menu, available providers include:

• Nvidia
• OpenAI
• Anthropic
• Google
• Deepseek
• Tencent
• OpenRouter
• Other LLM (compatible with Ollama)

Step 3: Enter API Key

• Paste your API Key in the "API Key" input box

Step 4: Select Model

• After selecting provider and entering API Key, click the "Model" dropdown menu to load available model list
• Recommended models:
  • deepseek: Open source, low consumption
  • gemini: Good results, high consumption

Step 5: Test Connection

• Click the "Test Connection" button
• Wait for test results
• Success: Shows "API connection test successful"
• Failure: Check API key and network connection

Step 6: Save Settings

• After successful test, the "Save Settings" button becomes clickable
• Click save to complete configuration

Adjust Model Default Settings

Step: Switch to Default Settings Tab on Configure LLM Page

Temperature

• Default: 1
• Range: 0-2

Max Tokens

• Default: 4096
• Controls maximum length of AI single response
• Note: Higher values consume more API quota, different models have different maximum limits

Enable Streaming

• Default: Off
• When enabled, AI responses are displayed in real-time instead of waiting for complete response

Note

• After setting, each time you start a new conversation, these default settings will be enabled

Custom API Configuration (Advanced)

If you selected "Other LLM":

Enter LLM URL

• E.g.: https://api.openai.com/v1
• Must be OpenAI-compatible API interface

Auto Get Model List

• After entering API Key and URL
• Click model list button to get available model list

Privacy and Security

• API keys are only stored locally on your device
• All API requests are sent directly to your chosen AI service provider

Note

• Some service providers restrict certain geographic locations. For example, Gemini and Claude model access has geographic restrictions for certain countries. If your IP is on this list, the service will be unavailable