ContextWindowExceededError

[TSJasonH]

Feel free to mark this as a duplicate in the case that it's already being/been handled in #49 or #278

I hit the below error with freshly installed skyvern running in docker v0.1.54 and GPT4o

Prompt:

Open images.google.com and search for pictures of parades. Click on an image that contains balloons.

skyvern-1     | litellm.exceptions.ContextWindowExceededError: litellm.ContextWindowExceededError: litellm.BadRequestError: ContextWindowExceededError: OpenAIException - Error code: 400 - {'error': {'message': "This model's maximum context length is 128000 tokens. However, you requested 128191 tokens (111807 in the messages, 16384 in the completion). Please reduce the length ofthe messages or completion.", 'type': 'invalid_request_error', 'param': 'messages', 'code': 'context_length_exceeded'}}
skyvern-1     | 2025-02-03T23:49:20.544718Z [error    ] Unexpected exception in agent_step, marking step as failed step_id=stp_356238418793460668 step_order=1 step_retry=3 task_id=tsk_356235429496222434

Thanks for the awesome project!

[wintonzheng]

@TSJasonH

some websites have large amount of content which leads to this error.

if you could share the website and your prompt, I can see if there's improvement we can make to reduce the token size.

[TSJasonH]

Of course, the prompt was lofty anyway. It was:

"Open images.google.com and search for pictures of parades. Click on an image that contains balloons. "