High volume of failures with Gemini models during graph creation [know issue, from googles side]

Should anyone experience a lot of failures and retries during graph creation, this is known issue being tracked in the googleapis/python-genai project. Currently Open AI is working flawlessly, generating 100k graphs without a single failure, so there is not much we can do from our side at present. 

https://github.com/googleapis/python-genai/issues/1815