How can I know which rate limit I hit? Rate limit...
# proj-agent-canvas
p
How can I know which rate limit I hit? Rate limits are usually measured across three dimensions: • Requests per minute (RPM) • Requests per day (RPD) • Tokens per minute (input) (TPM) https://ai.google.dev/gemini-api/docs/rate-limits
l
I can't answer to that, unfortunately. You'd have to look at what the Gemini API exposes on its response, and whether LiteLLM can pass those values. If you keep hitting rate limits, I suggest to raise the 'retry' settings so that the total wait is over a minute. Use one of the online calculators for tenacity binary exponential backoff to figure out the right values (e.g. the OH defaults are just 18s, because the formula was different from what they expected = 65 seconds, as you can see from the comment in the code). Edit: I left a more detailed breakdown of the issue here a few hours ago.
👍 1
a
Based on my experience yesterday I’m guessing the requests per day… I started hitting resource exhausted around dinnertime and had the —yolo flag on so Gemini spent all our vibe coding credit by himself… Whoops. I was using the CLI directly, but the cadence of task steps felt similar to openhands with a non-local provider. Looking for their stats location now actually
👍 1