Add ability to truncate text messages by token limit set by openai

[SilasStokes]

Right now if you try to give gpt too much text history it will explode. We need a way to truncate both our requests to the api and then length of the response.