Client Feedback: Issues with GPT-4o and Phoenix Image Prompting

A bit of feedback from a client. They are using Phoenix to trace multimodal calls. When using Gemini with images the prompt only shows the GCS image url so it's fine, however when using GPT-4o, the base64 encoded image is fully in the prompt. That is making Phoenix gets completely stuck because the prompts are too long and getting the platform to halt. Their solution right now is not use ChatGPT-4o 😅 Which doesn't sound like an optimal solution. It would be great to see some good solutions for this 🙂