Context continuity when implementing branching: `conversation_id` vs `previous_response_id`

Hi all,

I’m trying to implement the query edit and regeneration like on the ChatGPT UI, but stuck on how to implement the logic for context continuity. Previously just passed conversation_id to the Responses API in the linear chat and now, I am considering a setup like this:

1. At chat start, create conversation_id and pass to all API calls on the main branch
2. When a query is ‘edited’, pass only previous_response_id to the API call (I’ve read that you can’t pass both a conversation_id AND previous_response_id since they are mutually exclusive ways of context management)
3. For subsequent queries on this ‘edited’ branch, maintain passing previous_response_id only.

Would this mixing of conversation_id and previous_response_id work? I’ve heard that it wouldn’t since previous_response_id works by maintaining a chain of pointers and so if I start with previous_response_id at the edited question only, I’m losing all previous history and so the best option is to just trash conversation_id altogether and only pass previous_response_id all throughout. Any suggestions?