Better error handling, enhanced CSV processing, and our first agentic workflows

During this week's engineering demo, our team shared updates on enhanced CSV processing, clearer error messaging, prompt library upgrades, Chrome extension improvements, and our breakthrough agentic workflow system. These updates mark significant steps in making Storytell more powerful, transparent, and user-friendly.

Enhanced CSV processing breaks through data limits

Shelby, our principal software engineer, kicked off the demo with exciting news about our upcoming improvements to CSV processing. We are lifting the previous 1,000-row limit, allowing you to work with much larger structured datasets in Storytell.

Here’s how it works: When you upload CSV files, we create metadata and use SQL queries in an ephemeral in-memory database to handle much larger datasets—potentially millions of rows, depending on your questions. Instead of just stuffing data into context windows, we’re providing a smarter way to extract exactly what you need.

This means you’re no longer constrained by a small data cap when working with CSV files. Storytell can handle larger datasets smoothly, like having a powerful spreadsheet tool that can quickly find and analyze the data you need.

While some context window limitations remain for huge result sets, this update significantly expands what’s possible when working with data.

Clear error messaging improves user experience

Alex, Storytell’s DevOps Engineer, demonstrated our new approach to error handling, addressing a longstanding issue for users. Previously, if a streaming error occurred—whether from network disruptions, server problems, or model failures—Storytell would simply stop with no explanation.

Now, we show clear, helpful messages that explain what happened, why it might have occurred, and what you can do next. For example, if there’s a connection issue, you’ll see, “This can happen due to a temporary connection issue or high server load,” along with suggestion to report the issue.

Alex used Storytell itself to translate technical, system-generated error messages into more human-friendly language. This reflects our philosophy of using Storytell to build Storytell and tackle real-world challenges.

Prompt library improvements for better organization

Andi, our Sr. Frontend Engineer, showcased two key updates to the prompt library that make it easier to find and organize your prompts. First, you can now favorite prompts you use often; starred prompts appear at the top of your list automatically.

Second, you can filter prompts by type: All, Favorites, Personal (your personal library), and Shared (prompts from shared collections). This helps you focus on what’s most relevant for your current work.

These updates came directly from user feedback. 

Chrome extension improvements reduce intrusiveness

Andi also shared updates to our Chrome extension, following feedback about the text selection pop-up. We noticed more uninstalls after launching the pop-up feature, suggesting it felt intrusive.

The main update: You can now disable the text pop-up directly from the pop-up itself. If you choose “I don’t want this pop-up,” a tooltip explains how to re-enable it later through the side panel, then it disappears until you turn it back on.

We also improved the language in the pop-up items, making the options clearer and more useful based on earlier demo feedback.

Agentic workflows: a major breakthrough

Aaron, our Principal Software Engineer, showcased what might be our most significant advancement yet for V3: the agentic workflow system. Unlike traditional tools that only answer a single question, this system can understand your request, plan the best way to address it, and carry out the necessary steps—almost like a personal assistant that thinks ahead and works for you.

It doesn’t just react; it strategizes. When you ask a question, the system breaks it down and chooses the right tools like web search or document creation to deliver the most useful response. In the demo, Aaron showed how, for a question comparing streaming services for investments, the system didn’t just pull data from one place. It searched the web for updated information on Netflix, Disney+, and Paramount, then combined those findings into a clear, editable report, just like having a research team that can quickly and accurately find the information you need.

You can also see each step happening in real time, from initial idea to final result, giving you full visibility into how the system works. And if it realizes it doesn’t have enough information yet, it keeps going, automatically taking the next step, adding details, or calling in other tools until your question is fully answered.

What’s most exciting is that this system isn’t tied to one job. It’s designed to be flexible and domain-agnostic, adapting to whatever you ask—whether that’s analyzing financial data, drafting marketing copy, or exploring new ideas for your team.

This marks a major shift from the typical “ask a question, get an answer” approach. With the agentic workflow, we’re entering a new era of AI collaboration—where technology actively helps you accomplish your goals through intelligent, multi-step processes.

Looking ahead

These updates mark a major evolution in Storytell’s capabilities. We’re moving from a system that primarily searches and retrieves information to one that processes unlimited structured data, handles errors transparently, organizes knowledge more intuitively, integrates seamlessly with your workflow, and autonomously executes complex, multi-step tasks.

As language models continue to improve, this infrastructure will become even more powerful. We’re laying the foundation for AI that doesn’t just answer questions but actively helps you accomplish your goals.

These updates reflect our commitment to transparency, user feedback, and building genuinely useful AI tools. We’re excited to get these features into your hands and hear how they transform the way you work.‍

Become an Alpha or Beta Tester

Get early access to features as we release them by becoming an alpha or beta tester. Here's how to sign up: https://docs.storytell.ai/about/early-access