Input Variables
     
     
    Console Log
     

    You can copy and paste YouTube video URL, it will extract video_id and use YouTube Transcript API to get the video transcript in English. Support for other languages can be added, even on-the-fly translation is possible. If the transcript is not available, it will show an error message.

    You can also provide any Web URL, it will extract the text and summarize it using AsyncHtmlLoader and Html2TextTransformer

    PDF URLs are now supported! It uses PyPDFLoader and some custom logic to reduce documents to fit in the LLM context window.

    If the text to be summarized fits in the LLM context window it is directly summarized. If it is too long, it will be split into chunks, each chuck is summarized and then combined. This approach is called Map Reduce.

    I am using 🦜️🔗 LangChain to steer the process, SocketIO for streaming logs and LLM output in realtime.

    Your Summary