A summarization request exceeds a model's context window after many documents are added. What should the application do?

1. Disable all authentication headers
2. Retry the identical oversized request repeatedly
3. Increase only the response token limit indefinitely
4. Select relevant material or summarize in stages within supported limits