Your AI says "too many requests"

If your AI says Pennyweight returned "too many requests", it hit one of two rate limits. Both clear on their own. You don't lose any data, and nothing you already saved is undone. For how your AI talks to Pennyweight, see how Pennyweight works.

What you'll see

There are two messages, one per limit.

• Too many calls in a minute. Too many requests on this connection (120/min). Retry in 12s. The number of seconds varies.
• Too many bug reports or feature requests. Too many feedback operations this hour (max 10, previews included). Try again later. If it happens at the moment you approve a report, the message is Too many reports this hour (max 10). Try again later.

Both carry the error code RATE_LIMITED. Pennyweight returns the per-minute one as an HTTP 429 response, and your AI app decides how to show it, so you may see the app's own error wording instead.

Why it happens

• Per-minute limit. Each AI connection can make 120 calls to Pennyweight per minute. Every call counts, including reads. The count resets at the start of each clock minute, so the wait is never longer than a minute. A normal conversation stays far below this. It usually means your AI is repeating calls in a loop.
• Feedback limit. You can make 10 feedback operations per hour. Previews and sends both count, so one bug report you preview and then send uses 2. Bug reports and feature requests share the same 10, across all your AI connections. The hour starts with your first operation.

How to fix it

1. Wait. For the per-minute limit, wait the number of seconds in the message. For the feedback limit, wait up to an hour.
2. Ask your AI to carry on from where it stopped. If it was working through many items one at a time, ask it to do them in fewer, larger steps.
3. If the per-minute message keeps coming back during ordinary use, tell Pennyweight. See report a problem, or ask your AI to open the support page at pennyweight.ai/contact.

Good to know

• Each connection has its own minute count. Another AI app you've connected keeps working. You can see your connections in Settings: see manage AI access.
• Refused calls don't extend the wait. A call turned away by the per-minute limit isn't counted, so retrying doesn't push the reset back.
• A temporary problem can look like the feedback limit. If Pennyweight can't check your feedback count, it refuses the report with the same "too many" message. Try again in a few minutes.
• Your AI isn't told to file a report about it. Rate limits are treated as something to wait out, so Pennyweight doesn't suggest a bug report for them.