Offline mode

[Arseni1919]

Describe the feature or potential improvement

The biggest drawback of Langfuse, in my experience, is the lack of offline support. You need an internet connection to access anything. Please correct me if I’m mistaken, but I was on a plane without internet and tried to debug an issue in my prompt—unfortunately, Langfuse couldn’t help in that situation.

Because of this, I’ve decided to switch to MLflow. I genuinely appreciate what Langfuse offers, and if offline mode becomes available in the future, I’ll seriously consider switching back.

Additional information

No response

[marcklingen]

Thanks for your feedback. You could run Langfuse locally on your machine using Docker Compose; it should work similarly to MLflow. Is this an option?

https://langfuse.com/self-hosting/docker-compose

[uplife1980]

Hi @marcklingen , I think "offline mode" is lighter than deploying Langfuse locally.

Sometime we want to deploy and run our Langchain project without Langfuse, but langchain still writes the log/state to a file. We need Langfuse to support "Importing" data from these raw log file so we can audit the agent sessions afterwards if needed.

Is it possible to have Langfuse to standardize a schema of logging format?