6 Replies
Can you atleast gracefully not block the client. In this case I didnt even change my prompts so the logic where you are checking for updates fails I don't think it makes sense to crap out but to proceed gracefully!
Looks like its back.
are you on the latest version? this issue should not kill the prompt client but should cause to fallback on the stale cache
we currently make some infra changes which lead to a couple of seconds of these errors a day, will hopefully be resolved soon. both tracing and prompt management features in the latest SDK versions should gracefully handle these issues
sorry about the inconvenience, this is our top priority to fix right now
Thanks for getting back. I am on langfuse==2.35.2 on python. i checked the changelog https://langfuse.com/changelog and couldnt find anything. Did I miss something?
Changelog - Langfuse
The latest release updates from the Langfuse team.
?
Hi @AJ Ram - can you confirm whether this was an exception raised or simply an exception logged? The version you are on should already gracefully handle failed fetches on the prompt routes by returning the cache. Only in case there's no cache available on the first fetch, an exception would be raised. For that case, you may upgrade your SDK version to latest and make use of the
fallback
kwarg to mitigate these exceptions on first fetch. See docs here: https://langfuse.com/docs/prompts/get-started#fallbackPrompt Management - Langfuse
Manage and version your prompts in Langfuse (open source). When retrieved, they are cached by the Langfuse SDKs for low latency.