Problem

The logs are currently detached from the inputs, so while there is a UUID setup for associating the inputs to the the similarity log entry, there is no log of inputs anywhere. I imagine that as data analysts use this tool, it would be important to see the specific inputs that are causing drift on either side of the scale, user and AI.

Solution

We can solve this problem by adding the input raw text to some sort of additional file, like a JSONL or even just JSON file that has the UUID and input text, this would make the log entry for any specific drift violation have a UUID as a pointer to the JSONL that contains the input text.

Another solution: The function could be made to accommodate a UUID as a third parameter so that, in a production environment, any number of I/Os can be linked together in the log.

Additional information

No response

👨‍👧‍👦 Contributing

• 🙋‍♂️ Yes, I'd love to make a PR to implement this feature!