Search your Anki notes by meaning, not just keywords. Find relevant cards with AI-powered semantic search and get answers generated from your own notes. Mainly used with AnKing Overhaul and AnkiHub syncing.
Features Semantic search — Finds notes by conceptual similarity (e.g. “causes of heart failure” matches related content even without exact words) AI answers — Get summaries and answers grounded in your selected notes Multiple providers — Use Ollama (100% local, no API keys), or Voyage/OpenAI/Cohere/Anthropic with API keys Flexible configuration — Choose note types, fields, decks, relevance mode, and optional re-ranking Privacy-friendly — Use Ollama for embeddings and answers to keep all data on your machine Quick Start (Ollama — free, local) Install Ollama and run: ollama pull llama3.2 and ollama pull nomic-embed-text In Anki: Tools → Anki Semantic Search → Settings → set embeddings and answers to Ollama (local) Tools → Anki Semantic Search → Create/Update Embeddings (run once or after adding many notes) Tools → Anki Semantic Search → Search Notes → enter a query and search Screenshots
Requirements Anki 25.09.2 (tested on this version) Embedding engine: Ollama (local) or Voyage/OpenAI/Cohere (API keys) AI provider: Ollama (local) or OpenAI/Anthropic (API keys) Optional Enhancements
| Feature | Install | |---------|---------| | Cross-encoder re-ranking (better results) | pip install sentence-transformers | | Anthropic streaming | pip install anthropic | | Spell check in search box | pip install pyspellchecker |
Credits
Developed by Zahi-Essam. Built with Cursor AI and API costs (~$50) to develop this add-on.
GitHub: https://github.com/zahiessam/anki-semantic-search
I’m not a developer—I was inspired by AnkiHub’s Smart Search and built this for myself since a monthly subscription isn’t feasible for me in Egypt, where currency devaluation has made it harder to afford. Sharing it in case it helps others in a similar situation.
Screenshots
Liên kết hỗ trợ
Reviews (1)
👍 2026-02-22
This is sick. So useful. Would be useful to be able to integrate into the anki browser search. Also, could benefit from other embedding models that are fine tuned for medicine, e.g: hf.co/cmiralop/BiomedNLP-PubMedBERT-Base-Uncased-Abstract-Fulltext-109M
