Hello. I am new to ChatGPT, and there are so many sources of information that it’s easy to get lost. Thank you for being patient 
My goal: To create a personalized chatbot to help my users in terms of product searches, information about these products, technical support, FAQ. I would like to limit the responses to only the domain of the company. All this in multiple languages.
What I have tested:
- The custom chatbot: I used a web scraper to feed the beast, then I passed these data to it. It works not too badly, but it’s not integrable from a website. Moreover, it requires an account, etc…
- The assistant: I uploaded the files I had used in the custom chatbot. The responses are totally different from the custom chatbot, and whether using the playground or the API, the responses are even more different. Moreover, it is extremely slow and costly. It does not seem to be the best solution for me. I tried to reformat and optimize my JSON files, but it doesn’t change much, neither in terms of cost nor in terms of relevance.
So, I resolve to use the base API, which is very comprehensive, and I am a bit lost in terms of the development process, what technology to use, and where to start. According to my readings of “How to build an AI that can answer questions about your website,” the first thing to do would be to transform my data into a CSV file to limit the size of the “tokens.” Is this mandatory, or can a JSON file be tokenized?
Anyway, I’ll get back to my reading and would be very happy to read your advice.
Cheers.