Request too large

[ikbened]

Hi Many,

May be this is more a feature request than a bug, but I encountered this....

In my test I search for an article in a webshop and verify that a counter with the number of articles found is 1 or more. I made the css to that counter invalid. When running this test, I get this output:

Bol com search :: Search an existing product.                         Keyword 'Browser.Get Text' with arguments '['css=p[data-test="number-of-articlesX"]']' used on line 37 failed.

Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new
LiteLLM.Info: If you need to debug this error, use `litellm._turn_on_debug()'.

Attempt 1 failed: litellm.RateLimitError: RateLimitError: OpenAIException - Error code: 429 - {'error': {'message': 'Request too large for gpt-4o in organization org-XXXXXX on tokens per min (TPM): Limit 30000, Requested 253262. The input or output tokens must be reduced in order to run successfully. Visit https://platform.openai.com/account/rate-limits to learn more.', 'type': 'tokens', 'param': None, 'code': 'rate_limit_exceeded'}}

Although this page is not that complex/big, it seems that the request to openAI is too big. Is there a way to limit the request size, for now I only get these messages...

Cheers,
Ed

[manykarim]

Hi Ed,

thanks a lot for the feedback.
Yes, context lenght is a pain in the neck..

If I understand correctly, you are testing with https://www.bol.com (just guessing my your test step).
I can also play around with that one later.

One thing you can try in the meantime, is to enable an alternative approach to generate the locators:
https://manykarim.github.io/robotframework-heal/features/#cssxpath-generator
Add the argument use_llm_for_locator_proposals=false to the Library import.
By using that approach, the Library will no longer send the whole DOM Tree to the LLM.
But instead it will generate CSS Locator Proposals locally (without LLM) and only send those proposals to the LLM for a final selection.
It might take some more resource locally, but most likely the prompt to the LLM will be shorter.

Give it a try, in the meantime I will check on alternatives