AnkiVec - Vector Search for Anki

This anki addon creates vector embeddings for your cards, allowing you to search using natural language queries rather than keywords.

Features

• Vector Embeddings: Generate embeddings for all cards using local Ollama models
• Semantic Search: Find cards by meaning, not just keywords
• Fast Local Processing: Uses lightweight embedding models (nomic-embed-text by default)
• Persistent Storage: ChromaDB stores embeddings for quick retrieval

How to Use It

When you restart Anki after installing the add-on, a vector-database will be initialized for your deck. This process may take a few minute- you should see a loading dialog with a progress bar. From this point on, any changes to your deck will be automatically indexed.

To search the vector database, use the form "vec: [your query here]" in Anki's usual search bar. You can combine keyword searches with vector searches by putting the "vec" section at the end of the query: e.g. "keyword1 keyword2 vec: [natural language description]".

Prerequisites

1. Anki (version 2.1.45+)
2. Ollama installed and running locally

Install Ollama

Download from ollama.ai and install.

Pull the embedding model:

ollama pull nomic-embed-text

Configuration

There are three parameters that govern how AnkiVec operates, available in the add-on's configuration dialog:

• model_name: The name of the Ollama model to use for generating embeddings. The default is "nomic-embed-text", but you can specify any model supported by Ollama (for example, "kronos483/MedEmbed-large-v0.1" for a specialized medical model).
• search_results_limit: The maximum number of search results to return. Default is 20.
• ollama_host: The URL of the Ollama server. Default is "http://localhost:11434".

Future Plans

• Other Platforms: Currently, only Mac OS is supported. Windows support should be straightforward if I can get my hands on a Windows machine. Linux support is more tricky, as the uv binary bundled with Anki will be installed in different places depending on your distribution. I need to investigate how Anki is distributed in different package managers.

• Ordering by Relevance: Currently, Anki's search feature allows you to filter the cards you seen in the Browser, but not to order them by a specific relevance score. Implementing this as an add on might require obscene amounts of monkey patching, but I think it will be worth the effort.

• Debugging Ollama: Ollama can be buggy. A small fraction of Anki notes cause Ollama to throw an error when generating embeddings. This is a known issue, and I'm working on figuring out what's going on.