Thanks for your response, Anthony P..
In my tool call definition, I referenced one parameter within the description of another, using it as a condition. According to several sources, OpenAI processes parameters from top to bottom. If a key is referenced before it's defined, it can lead to hallucinations and incorrect responses. I tested this by rearranging the parameter order, and when defined correctly, it worked as expected. In my case, the value of certain keys should depend on another key鈥攅ither being null or an object鈥攂ut it always returns null regardless.