New feature alert!Storytell can now analyze structured data in the form of .csv and/or .xlsx files. If you haven't used this feature yet, we encourage you to give it a try. This new mode is triggered in SmartChat™ under the following conditions:
You are chatting with 5 or fewer files. This can be an individual document, a tag, or tag group
At least one of the files is a .csv / .xlsx file
You will know when you are using this new feature when a chat response starts with:" You are chatting with CSVs, switching to Analytics Mode... "So what exactly is different about this mode?In short, this new mode actually let's an AI model write and execute it's own code! Think of it like a personal Data Analyst that can do everything from describing datasets to identifying patterns/trends and even training it's own Machine Learning model for specific tasks. 🤯
Tips & TricksThis new feature is in Beta, and as a result can be fragile, so I've collected some tips and tricks here to help you get the most value possible out of this tool.
1. Write clear and detailed prompt:
This includes describing the file format and content (.csv with stock data), defining any terminology (like custom metrics and column names) that could be confusing, and providing guidance on how you would like the question to be answered.Here is example prompt I have used when analyzing stock data to give you a starting point:
Attached is csv file with daily opening and closing prices for TSLA (a publicly traded company. Please answer the following questions:
1. Which day had the gain? Gain is defined as the largest difference between the closing and opening price
2. Please forecasts the opening prices for the next 30 days using a SARIMA times series modeling approach
Note: Closing prices are in the column labeled "cl". Opening prices are in the column labeled "op"
** Don't worry if you don't know what SARIMA is, it's just a specific technique for forecasting data. I'm just using it to illustrate that you can give direction on how you want a problem to be approached.
2. Clean your data beforehand.
This new mode actually writes/runs code you in order to answer complex questions. For those that have done some data analysis before, you are likely familiar with how important formatting can be. If you provide messy, poorly formatted data, SmartChat™ will likely have difficulties reading the file in the first place.
3. Not all errors are created equal.You might notice an "error" mentioned when reading files or running an alaysis. This does not necessarily mean the request is complete and has failed. The model is actually able to debug / fix some errors on the fly and might just need more time to finish the request. You can always check if the chat is still in process but looking for the blinking cursor.
4. Chat History is now enabled!We've enabled chat history for up to 5 previous messages, so feel free to ask follow up questions!Here is what a typical output might look like:
What's coming next?
You tell us!
I hope this helps and feel free to reach out with further questions!