Skip to main content

AI Functionality (experimental)

Starting with version 6.1.1, the TagSpaces desktop application allows you to connect to Ollama as an external AI service provider. Ollama is a locally installed offline software that enables running AI models (LLMs - large language models) directly on your computer.

tip

TagSpaces does not have its own AI engine or models but relies entirely on external AI software like Ollama for these features. All AI features are disabled by default.

Prerequisites​

  • A modern PC with a recent Nvidia/AMD graphics card or a Mac with an Apple Silicon processor. Ollama also works on regular CPUs, but performance may be slower.
  • Download and install the Ollama software.
  • At least 10 GB of free hard drive space for LLMs.

AI Configuration​

In the settings, a new tab called AI is available. Here, you can add AI engines and manage their models.

Adding an AI Engine​

After installing Ollama, you can add it by clicking the Add AI Engine button in TagSpaces.

Add Ollama as AI engine

If everything is set up correctly, you'll see a new section labeled "Ollama" with a green indicator, meaning Ollama is running in the background. If there are connection issues, the indicator will turn red.

Ollama connected in TagSpaces

By default, the configuration assumes the Ollama service is running on your local machine https://localhost:11434. However, it can also run on other computers in your network. You can add multiple Ollama configurations using the Add AI Engine button. Select the desired engine in the Default AI Engine dropdown.

Downloading Models​

The next step is to download a suitable model (LLM). This can be done via the Default model for text-based tasks dropdown.

Choosing AI models

In the "Example installable AI models" section, you can choose a suitable model. For text-based tasks, we suggest the llama3.2 model. For image-based tasks (e.g., image description), you can use llava or llama3.2-vision. While llama3.2-vision delivers better results, it is significantly slower.

Once you select a model, a progress dialog will show the download status. Note that models are typically several gigabytes in size, so patience is required.

If you can't find the desired model, you can install additional ones using Ollama's command-line tool.

AI Chat in Folders​

If configured correctly, a new button will appear next to the perspective switcher in the folder area. Clicking it opens the AI Chat tab in the folder properties.

Open folder AI chat

At the top of the tab, select the model for the AI chat. At the bottom, use the AI prompt to ask questions.

AI chat results

You can maintain a separate chat for each folder, enabling chat-based research or project organization directly in folders.

Add images to AI prompts
warning

All chat history, including dropped images, is saved in the ai folder within the .ts subfolder of the current folder. Deleting the ai folder or the current folder will delete the chat history.

AI Features for Files​

PRO

AI-related file features are part of the PRO version. These functionalities are currently available:

  • Generate image descriptions for JPG and PNG files.
  • Generate tags from image descriptions.
AI-generated image description
  • Generate tags for images in JPG and PNG formats.
AI-generated image tags
  • Generate PDF descriptions based on content.
AI-generated PDF description

Upcoming AI Features​

The following features are on the development roadmap for TagSpaces.

  • Text-based file summarization
  • Batch processing for multiple files
  • Integration with additional AI engines
  • Extraction of dominant colors in images
  • Translation of generated content
  • Language configuration for generated content

Any feedback and new ideas are welcomed in our forum.