Batching / parallel API calls for GPT-4-vision

API
1

Hi OpenAI Community,

I’m currently using the GPT-4-Vision API to summarize images in documents. While the results are great, the processing time for large batches of documents is quite long due to the sequential nature of the API calls. I’m looking for ways to optimize and speed up the process through batching and parallel processing.

Current Workflow

  1. Extract all files from a directory and identify documents containing images.
  2. For each document with images, convert images to base64 and call the GPT-4-Vision API to generate summaries.
  3. Combine text and image summaries into new documents for further processing.

Does the GPT-4-Vision API support batching multiple images in a single request? If yes, what is the recommended way to structure such requests?
Or does anyone have experience implementing parallel API calls and ensure compliance with rate limits?