Vector Search for X (formerly Twitter)

About

Leverage the power of embeddings to find tweets you vaguely remember via keywords or concepts

- Saves every tweets you scroll by inside your browser’s IndexedDB, including media content
- Syncs across your devices via Google Drive appDataFolder (requires authorization). Turned off by default.
- Provides a Twitter-like interface for you to search tweets you’ve seen.
- Search results are displayed exactly as seen on Twitter. There is a “compact” mode option that reduces the size of the media elements so more focus is placed on text.
- All you need is an OpenAI key. You can generate embeddings via either the text-3-embedding-large or text-3-embedding-small. The cost is negligable, but there is a cost estimator inside the extension should you be curious about pricing (around 150k tweets per $ for the large model)
- Small amount of customization options, like the similarity score threshold for a tweet to appear in your search results.

Got an idea to improve this extension? Tweet at/DM me here: https://x.com/godardbuilds