SystemSculpt AI

Brain Module - AI Models and Selection

The Brain Module supports various AI models and provides flexible options for model selection and management.

Supported Model Types

  1. OpenAI models
  2. Groq models
  3. OpenRouter models
  4. Local models through Ollama or LM Studio

Model Selection Modal

The Model Selection Modal offers a quick and efficient way to switch between different AI models.

How to Use

  • Open the modal using the assigned hotkey (recommended: CMD+M)
  • Use the search bar to filter models in real-time
  • Navigate through models using keyboard arrows or Tab/Shift+Tab
  • Select a model by clicking, pressing Enter, or using keyboard shortcuts

Features

  • Fuzzy multi-term search
  • Real-time filtering
  • Keyboard navigation
  • Visual highlighting of search terms
  • Grouped display by provider
  • Automatic selection of the first model after search
  • Keyboard shortcuts for quick selection
  • Enhanced refresh functionality with visual feedback
  • Error handling with clear user messages

Model Provider Filtering

The Model Selection Modal now remembers your provider filter preferences:

  • Filter settings are saved between sessions
  • Toggle visibility for:
    • Favorited models
    • Local models
    • OpenAI models
    • Groq models
    • OpenRouter models
  • Changes take effect immediately and persist after closing

Your filter preferences will be restored each time you open the Model Selection Modal.

Model Fallback Behavior

When selecting a model, the plugin follows this order:

  1. Default model set in settings
  2. First available local model
  3. First available online model (OpenAI, Groq, or OpenRouter)

If no models are available, "No Models Detected" will be displayed.

Performance Considerations

  • Local models may require significant computational resources
  • Cloud-based models depend on internet connection speed
  • Adjust max output tokens setting to balance generation speed and output length

Best Practices

  1. Experiment with different models to find the best fit for your tasks
  2. Use hotkeys for quick model switching
  3. Regularly update the plugin to access new models and improvements

Model Response Types

The Brain Module supports both streaming and non-streaming models:

Streaming Models

  • Provide real-time response updates
  • Better for interactive conversations
  • Support token-by-token output

Non-Streaming Models

  • O1 models (o1-preview and o1-mini)
  • Better for reasoning tasks that need more time to think through your question

Token Management

  • Automatic token limit adjustment based on model capabilities
  • Smart handling of max output tokens
  • Optimized performance for both streaming and non-streaming responses