digital torrent
Hi there. I've been working a project that features a GPT4-driven chatbot that stores an entire conversation and dynamically responds based on the provided context, essentially giving it a "memory".
Regarding the API usage quota for this, would supplying a large table (of context) for the messages array while querying the v1/chat/completions API count towards the token cost per-each-message (despite the previous responses being already processed in the past)?