I am surprised there are no results that Google finds on the forum here or elsewhere. We recently updated out instance but existing workflows with claude attached to information extract continue to work (thankfully). Any ideas?
Crazy stuff. I have a self-hosted n8n for personal education. I have exactly the same version and I just copied the node to it + the same API credential it works fine.
This was very difficult to figure out as one cannot see any hint of what key has been used and the key itself seems to work everywhere except the n8n. Due to limited access to platforms in corporate environment, this took longer than it should have! But eventually I was able to A/B test to the trailing slash.
The credential was created in early autumn, that’s around 3 months ago. I am not sure if n8n or underlying library has changed something but it gone rogue. Existing workflows continued to work but new ones would gives this headache.
Since it’s a value set by default I highly highly doubt that I would have ever touched it. The connection test also successfully succeeds however everything else leads you to a 404 spiral hell.
It totally could be an idea to strip duplicate slashes or use better URL handling behind the scenes as this is such a funny/trivial bug in a nutshell. I
Perhaps it could also be an idea for more prompt support (two replies but not support was received) when we choose to pay for the cloud services - something tells me that if we were to self-host, looking at the logs this would have been far more straight forward.
