A complete rewrite of the application. This is now split into two parts:
A small Python utility package called frogbase that contains all the backend logic for the UI. This can be used as a standalone package or integrated into other applications.
A slimmer Streamlit UI that provides a thin wrapper around the frogbase package built purely with self-hosted applications in mind.
Featuress
More content sources & formats
The use of pytube has been replaced with yt_dlp. This unlocks content download from a broad range of media platforms like YouTube (channels, playlists, videos), TikTok, Vimeo etc. (full list)
Local files can now be ingested from a directory instead of just a single file.
Sources can now be added in as a list of urls and/or local file paths.
Semantic Search
The search functionality now includes semantic search over transcript contents instead of a simple substring search.
This is done using sentence-transformers and hnswlib
Updated Streamlit UI
The UI now includes the concept of Libraries to further organize media downloads. Libraries are simply subdirectories within the main data directory.
Filter & search functionality have been simplified and made more intuitive.
⚠️ Breaking Changes
A complete rewrite of the application. This is now split into two parts:
frogbasethat contains all the backend logic for the UI. This can be used as a standalone package or integrated into other applications.frogbasepackage built purely with self-hosted applications in mind.Featuress
More content sources & formats
pytubehas been replaced withyt_dlp. This unlocks content download from a broad range of media platforms like YouTube (channels, playlists, videos), TikTok, Vimeo etc. (full list)Sources can now be added in as a list of urls and/or local file paths.
Semantic Search
Updated Streamlit UI