Can you check out that sheet and see if it makes sense?
Hi, i think the accessories on the same column of brand will create confusion like that, maybe try to create a new dedicated google sheets with just accessories inside
1 Like
Thanks for this. Is there a limit on the number of URLs we can train a bot on?
Hi Kyle, there are no limits, but I must say that at the current stage, GPT is not designed for serving URLs. So, when you add a lot of URLs, if the match of the keywords is not right, you can get some made-up URLs. We are working to solve this problem. This will likely get better when OpenAI releases an update.
If you have a few links and the user’s keywords match the description on the link, it should work just fine.
1 Like
Thanks for the reply. I understand, but what about the content from the URLs? How much can we train them on? I have an organisation that has around 600 posts…
Oh ok Sorry I misunderstood your question, ok so, there is no limits in terms of how much content you can train, but what we understood is that formatted content works much better than raw content from websites, also too much content will likely create confusion/hallucinations
Ah ok thanks. Is there any way to scrape websites and format the data into a particular style to improve this?
I found that one of the best way is to create q&a of the website content. Check the video above
.