I’m building a multi-turn app with the OpenAI Responses API and am trying to understand what state I need to send on each request. If I use previous_response_id, are the earlier instructions and tool definitions carried into the next response, or should my app include them again?

I’m especially interested in what happens after a tool call, since I want the assistant to keep following the same rules throughout the conversation.