Worked for a while using OpenEputer although we hit some rate limits then it said it didn’t understand the request. We seem to be burning through Shakespeare credit really fast ($10 in about a couple of minutes). We’re working on the project/space-zapper
Discussion
“The AI provider did not understand the message / data it got. If it persists, try stating a nee conversation or using a different model”.
I almost exclusively use Opus 4.5 and Sonnet 4.5 via OpenRouter. I do not recall a rate limit, but I have seen an issue where OpenRouter had a network issue.
Yes, we're aware Shakespeare is expensive, but we're working on a way to reduce this with a smaller system prompts and other tricks.
I guess that’s around token management. It’s a great tool though so keep pushing on! The error we’re blocked with using OpenRouter now is:
“The AI provider did not understand the message / data it got. It it persists, try stating a nee conversation or using a different model.”
In the console there is a “AI generation error: BadRequestError: 400”
(A new chat does not resolve the issue)
Oh. Thanks Ben. I'll pass this on to nostr:nprofile1qqsqgc0uhmxycvm5gwvn944c7yfxnnxm0nyh8tt62zhrvtd3xkj8fhgpzamhxue69uhhyetvv9ujuurjd9kkzmpwdejhgtcpzemhxue69uhhyetvv9ujuerfw36x7tnsw43z7qgkwaehxw309aex2mrp0yhx6mmnw3ezuur4vghs0klf00
[Feature request] Ability to type as the agent is responding / working so I can line up the next request ready. nostr:npub1q3sle0kvfsehgsuexttt3ugjd8xdklxfwwkh559wxckmzddywnws6cd26p
Agree - those are my favourite models too.