NemoClaw Knowledge Wiki

New Ollama UI - Matt Williams channel

3 min read

https://www.youtube.com/watch?v=prrWESXl7wg Here is a Markdown summary of the video transcript regarding the release of Ollama’s native UI.

Ollama Native UI Release

Ollama, traditionally known for its command-line interface (CLI) and API, has released a native User Interface (UI). While currently in a “first version” state, it is described as polished and a significant step toward making local AI more accessible.

📥 Installation

  • Availability: Currently available via the Releases page on the Ollama GitHub repository (look for pre-releases).
  • Rollback: Users can easily revert to the standard installation if they encounter issues.
  • **Linux Command:**curl -fsSL https://ollama.com/install.sh | OLLAMA_VERSION=0.5.7 sh  (Note: Version number may vary based on the latest release).

🖥️ UI & User Experience

The interface is designed to be minimal and simple, contrasting with the more complex/cluttered UI of competitors like LM Studio.

  • Design: Simple chat box with a model selector in the bottom right corner.
  • Model Selection:
    • Lists all locally installed models.
    • Shows some downloadable models (though not an exhaustive list).
    • Filtering: Supports filtering by typing the beginning of the model name (prefix matching only, no fuzzy search yet).
    • Auto-Download: If you select a model you don’t have, it initiates a download upon the first prompt.
  • Quantization: The UI hides quantization details. All downloads default to Q4 (4-bit quantization), which is considered the “sweet spot” for performance vs. quality.

✨ Key Features

  • Vision Capabilities: Supports drag-and-drop for images. The vision models work seamlessly within the chat.
  • RAG (Retrieval Augmented Generation): Users can drag and drop text files, PDFs, etc., to ask questions about the documents.
  • Chat History:
    • A sidebar lists previous chats.
    • Shortcut: Cmd + N (on Mac) opens a new chat.
    • Limitation: No current keyboard shortcut to toggle the sidebar.
  • Standalone App: There is no synchronization between devices; chats are local to the machine.

⚙️ Settings

The settings menu has been expanded slightly from previous versions:

  • Model Location: Set where models are stored.
  • Network: Toggle to expose Ollama to the local network.
  • Context Length: A global slider to set context length.
    • Critique: The speaker notes this is a poor design choice, as different models have different maximum context limits, and setting this globally can cause errors or garbage output.
  • Sign In: A new option to sign into an Ollama account (functionality to be announced).

🛑 Missing Features / Wish List

The speaker highlighted several areas for improvement:

  1. Keyboard Shortcuts: The CLI heavily relies on shortcuts, but the UI currently requires significant mouse usage (selecting models, opening sidebar).
  2. Advanced Filtering: Ability to filter models by parameters (e.g., searching “8b” or “vision”).
  3. Model Management: No UI option to delete models, import models, or edit Modelfiles.
  4. Observability: Lack of access to metrics, logs, or a Prometheus endpoint for tools like Datadog.
  5. Visuals: No text resizing or UI zoom options.

🏁 Conclusion

The new UI is a fantastic entry point for new users who want to avoid the command line. However, power users will likely stick with third-party front-ends like Misty or Open WebUI for now due to the lack of advanced features and shortcuts.

Graph View