- I’ve developed an integration to OpenAI using both the Assistant and the completions modes. initially tried assistant mode but it is taking far too long since it needs to first set up new/update assistant if the context has changed , set up a new/check for existing thread, send message and run, then check for response. each of the api calls that require response needed to be put on a timer loop to check for completion. minimum smart delay with manychat is 10 sec which means i’m losing far too much time in all these processes. conclusion: using the assistant mode is resulting in the response taking about 12-13 sec to come through.
alternatively, if i use completions, i get a response in about 4 sec but the downside is the entire prompt has to be sent each time which results in high token usage.
is there something i’m missing here about how these integrations could work?
- is there a method to remove or replace certain characters from text fields in manychat? for example if the user or chatgpt message contains “” then it messes up the api call as it reads that as the end of the content variable. I have not been able to find a way to remove or replace such characters berfore feeding them into the api body. so it’s likely going to happen at the very least if a user uses the characters in their response it will break the api call. is there a way around this?
