Open benthamite opened 1 week ago
Thanks. The recorded response bounds in syntax-errors.md
are wrong though, so I can't use it for testing. Do you have another file I could use?
I get this all the time with Claude, never with OpenAI. It's always mystifying, but I'll try to get a reproducible test for you!
This finally happened again, though this time it doesn’t go away when I reopen the buffer. After some testing, it appears that problem occurs only when gptel-mode
is activated. So it may be that the phenomenon I observed previously was due to the fact that closing the buffer and reopening the file had the effect of turning off gptel-mode
.
Relevant files attached.
Here’s another example of what I believe is the same phenomenon. I’m not entirely sure since the error message is slightly different:
Claude error: (HTTP/2 400) messages: final assistant content cannot end with trailing whitespace
[ETA: uploading a new set of files, which isolate the issue better]
Please update gptel first -- errors are often fixed by the time they're reported.
Bug Description
Continuing the discussion here: #442.
This is an instance of the phenomenon I described there as follows:
Backend
Anthropic
Steps to Reproduce
<
.<
and>
.gptel-send
.It may be hard to reproduce the issue since, as noted, reopening the file fixes it. Hopefully, the other attached files will provide additional hints as to what may be going on. These fils capture the output of a dry run either before or after the fix.
Additional Context
GNU Emacs 30.0.91 (build 1, aarch64-apple-darwin24.0.0, NS appkit-2566.00 Version 15.0.1 (Build 24A348)) of 2024-10 macOS Sequoia
Backtrace
No response
Log Information
No response
Archive.zip syntax-errors.md