🐛 Messages from the user were placed after the assistant response. This was originally in place to handle when there were 429s or other request failures so that messages wouldn't append if OpenAI wasn't responding well. Turns out this broke the main working case.
🐛 Messages from the user were placed after the assistant response. This was originally in place to handle when there were 429s or other request failures so that messages wouldn't append if OpenAI wasn't responding well. Turns out this broke the main working case.