Enhancing Testing, Fixing Bugs, and Building Transparency in AI Responses

In our most recent engineering demo, December 13, our engineering team showed us how they are improving testing and resolving bugs to make our platform more accessible. Here’s a breakdown of the key issues and how we’re addressing them.

Streamlining Backend Testing with the Corpus Generation Tool

We wanted to solve issues for file uploads and some confusing prompt responses. To address this, Aaron, our principal engineer, developed a corpus generation tool which:

• Enhances the reliability of prompt responses.
• Streamlines file upload testing for consistent performance.
• Empowers our Crew to contribute to automated testing, improving coverage and accuracy.

This tool strengthens our backend testing processes, helping us build a more scalable and adaptable platform.

Fixing Bugs to Enhance Usability

We've heard from our users about file upload issues and some UI glitches. Here’s what we improved:

• File Upload Status on "Getting Started" Screen:
  • Added a "Ready" status to indicate when file processing is complete.
  • Changed language on the bottom panel to clarify that users can chat with files once they're ready.
• Blank Screen Issue:
  • Fixed a bug causing a blank screen after transitioning from the "Getting Started" screen and attempting to start a chat.
• File Upload by Dragging to Homepage:
  • Resolved issues with the file drop area disappearing when hovering over certain sections.
  • Fixed a bug where dragging homepage links onto the page triggered the file drop area.
• Story Tile Navigation:
  • Fixed navigation issues in the drawer where the "Next" and "Previous" buttons were not changing the content.
• Planned Improvements:
  • Adding timestamps to chat responses:
    • Responses within 24 hours will display "X hours ago."
    • Older responses will show the date and time.

Introducing Interactive Citations for AI Responses

We’re working on a new feature to make AI-generated responses clearer and more trustworthy: interactive citations.

Here’s what this means for you:

• Clearer Sources:
  When you ask our AI a question, it searches through its stored information to find the most relevant answers. Now, it will include citations—notes showing exactly where the information came from.
• Interactive Exploration:
  Soon, these citations will be interactive. You’ll be able to click on them to see the original source, whether it’s a document, a text snippet, or another reference.
• Simple, Reliable Results:
  While the AI works behind the scenes to match and link information, the result is simple: clearer, more reliable answers, and the confidence of knowing where the AI got its information.

Become an Alpha or Beta Tester

Become an alpha or beta tester to tell us what you'd like us to build! It's a great opportunity to shape the future of Storytell and get a sneak peek at what's coming next. Sign up here: https://web.storytell.ai/early-access

‍