Hey folks. I am trying to do something with ChatGPT. I give it an initial input with instructions and tell it that I will share infomration in 3 parts and based on that it should help me with something. The first part of information is 4000 words long. After that input it asks me a couple relevant qustions. but after that it forgets basically the initial prompt. I tried to break it up in small jumks, but its the same thing. I read that with each prompt it sends the data of the full conversation as an API call. If the convo is too long (or prompts are too long) it starts to cut off previous data as it can not fit the data into the call. So it doesn't matter if you provide it with data in 1 big prompt or junk it up in 4 small prompts. Any idea on how to feed it data as big as that?

Product pictures: I have a client who creates shoes. Given she has a picture of a shoe, is there an AI that can change the color of the shoe? I thought this would be a feature within Adobe, but apparently it is not. So, all she needs is a picture of a shoe that is brown to change to a picture of a shoe that was made with red leather. Any ideas?

Has anyone had any experience with Chrome Extensions to load files into ChatGPT? I used ChatGPT File Uploader Extended to load in pdf files, but it looks like this is no longer possible, since my ChatGPT window no longer has a green File Upload button. Just happened this weekend.

My client is a very large NY based catering company. They want an LLM that will give their sales people the ability to ideate new proposals and offers to clients, drawing from an existing catering database (their own), Salesforce data, a recipe database and maybe from some other sources. I think they need a vector database (Pinecone?) that pulls from all the sources mentioned and through embeddings would learn to answer questions from this information. Am I doing this the right way? Anybody have any experience with similar situations?