Notiondipity tackles the problem of workspaces that grow faster than their owners can remember them. It reads the pages you grant access to, converts them into vector embeddings with OpenAI's text-embedding-ada-002 model, and compares them using cosine similarity. From that index you get three things: a list of pages similar to the one you are reading, an explanation of how they relate, and ideas for projects that combine them.
Because retrieval is semantic rather than keyword-based, search works even when you do not remember the exact words you used. Typing a topic returns pages that mean the same thing, which is useful for research notes, meeting archives and personal wikis where vocabulary drifts over time.
The extension relies on a backend hosted on Fly.io that talks to the Notion API. According to the developer, that backend stores the generated embeddings along with titles, URLs, encrypted content and a hashed user identifier, and no third party other than the developer's own service has access. Your Notion API token stays in browser storage and is sent with each request. The developer labels the release a public Alpha, so bugs and rough edges should be expected.
