Users, we've made some changes to improve the speed and efficiency of our file upload process. We understand how crucial time is, and we had some tweaks at work optimizing our system.
Previously, each file uploaded was broken into chunks, and each chunk required a separate call to open AI to create a summary. This process was both time-consuming and costly, but now, we've eliminated the need for separate summaries for each chunk. Instead, we now make a single call to summarize the entire document, drastically reducing the time and cost. This change alone has cut upload times by 60-70%!
And in our quest for optimal results, we're blending various language models. We use GPT-4 for private chats, but for full document summaries, we utilize Claude's large context window. We also use GPT 3.5 to generate AI tags and Hugging Face for vectors. It's a sophisticated dance, ensuring the best user experience!
We're excited about these improvements and can't wait for you to experience the enhanced speed and efficiency with Storytell! Here's a video from Connor explaining more about these updates.
Part I:
Part II:
Status: These changes are live in our staging environment and will be pushed to production by EOD on our next release date: Wed, Sept 27th