Although I would recommend just fine tuning it in open ai and then using it directly from open ai to reduce jumps between servers.