Models Manager

Browse, download, and manage AI models for your NodeTool workflows.

Desktop app panel for downloading and managing local models. For the model catalog see Supported Models; for provider keys see Providers.

The Models Manager helps you browse, download, and manage AI models available on your system.

Opening the Manager

The Models Manager is a full page at the /models route — open it from the app navigation. It shows all downloaded models, recommended models, and available models from configured providers.

Models Manager — Full View

Browsing Models

Model Type Filters

Filter by Type

The sidebar on the left filters by HuggingFace pipeline tag rather than simplified labels. Tags include, for example:

Pipeline tag Example Use
text-generation Text generation, chat, reasoning
text-to-image Text-to-image generation
image-to-image Image transformation
text-to-video Text-to-video generation
automatic-speech-recognition Transcription, dictation
text-to-speech Voice synthesis, narration
feature-extraction Embeddings / vector search

The available tags reflect what’s actually present in your model list; each shows a count of matching models.

Search and Sort

  • Search by name or repository to quickly find specific models
  • Favorites – Star frequently used models for quick access
  • Recent – See models you’ve used recently

Downloading Models

Download Progress

  1. Find the model you want in the browser
  2. Click Download to start fetching it to your local cache
  3. Track progress in the Downloads bar at the bottom of the screen

Download Details

  • Downloads continue in the background while you navigate the app
  • The bottom bar shows total progress, speed, and estimated time remaining
  • Click the Downloads bar to expand and see individual file progress
  • The download WebSocket reconnects automatically on connection loss (up to 5 attempts with exponential backoff); this reconnection is separate from any per-file download behavior

Storage Location

Downloaded models are stored in your local HuggingFace cache (~/.cache/huggingface/) or provider-specific locations (e.g., ~/.ollama/ for Ollama models).

Managing Models

Per-Model Actions

Model Card Actions

  • Download – Fetch a model to your local cache
  • Delete – Remove a model you no longer need to free disk space
  • Show in Explorer – Open the model folder on your computer
  • README – Read the model’s documentation on Hugging Face

Model README

Recommended Models Dialog

Many workflow nodes specify recommended or required models. The Models Manager highlights these under a Recommended section with direct install links, so you can quickly get the models your workflows need.

Model Selection Dialogs

Each property role has a type-aware picker:

  • Language Model — LLM selector with provider grouping
  • Image Model — image-generation models only
  • Video Model — video-generation models only
  • TTS / ASR Model — speech models only
  • Embedding Model — vector embedding models only
  • HuggingFace Model — search any HF repo

Language Model Selector

Cloud Provider Models

Models from cloud providers (OpenAI, Anthropic, Google, etc.) appear in the manager based on your configured API keys. These don’t require downloading – they run remotely when you use them in workflows.

Configure API keys in Settings > Providers. See Models & Providers for setup details.

Next Steps

Edit this page on GitHub