Mine is at Maktos.com
I made a lot of progress recently, getting it working, and even observed some good answers on rather "controversial" topics.
I switched which AI model I was using, and modified the "prompt" or special orders I give it behind-the-scenes.
The reasons I haven't made a huge red-carpet release yet (still "to be done"):
- I want to put in some logging (LangSmith) to see where any processing time is being spent, so I can discover where improvements can be made. Is looking up the docuмents taking a while? Is it the AI model itself? etc.
- I want to peek at the data my system is returning (remember, I'm handing the AI the "best material" from a selection of Catholic literature, before I have it answer the question. That's where the magic happens).
- Any possible improvements to maximize the speed?
That's about it.