High level tips for creating a language learning AI tutor?

[lancejpollard]

I have a fantasy language which is made up. There are about 4000 words defined, written in ASCII with English definitions as seen here. I have a grammar for speaking/reading/writing/translating it as well, which is relatively straightforward to write down and teach to an AI it seems.

How can I create an AI assistant that does the following:

1. Remembers each definition of each word from that spreadsheet? So I need to somehow feed it structured data, perhaps as JSON.
2. Can learn how to read and write sentences in the language? I would have to teach it the grammar and tell it the rules for sentence formation. Perhaps I would also need to "work with it" a little, to correct it when it makes translation mistakes, or whatnot. Basically make sure it is reading and writing and translating to/from the fantasy language correctly.
3. Communicate with individuals using the app, giving them personalized lessons on the fantasy language?

What are the high level things I would need to look into to make this happen? What wouldn't be possible at this point in LangChain/AI/LLM development? Can I really teach it to read/write/translate a fantasy language? Could I have one conversation thread per user, remember that historically, but overall have one "person" AI assistant? Call them "Quinn". Could Quinn be the fantasy language AI tutor, but have conversations/memories unique to each user? What are the rough pieces of the puzzle to wire up, so I can direct my search when building such an app. I have been a JS/TS software engineer for many years, but this is all new to me so not really sure (a) what's possible / not possible, and (b) where to begin.

I saw there was a distinction between System/AI/Human. So I need to input in the system level the fantasy language word definitions and scope to limit the Quinn from being a general AI somehow?

Thanks for the help!

[lancejpollard]

I am following https://www.pinecone.io/learn/javascript-chatbot/ for now. Next question:

Can I create an AI assistant, Quinn, that has different "jobs" at different websites? This way, the "organization/company" has AI assistant Quinn, and say there are 5 websites, 1 on E-commerce, one on Language-learning, etc.. They all use Quinn, but Quinn is tailored to act certain ways in different contexts. Is that sort of thing possible? What would you need to do at a high level?

[lancejpollard]

A more specific question, focusing on a smaller slice of the puzzle:

High level overview of how to load dictionary definitions into LangChain?