search

lucagrippa / obsidian-ai-tagger

Simplify tagging in Obsidian. Instantly analyze and tag your document with one click for efficient note organization.
MIT License
52 stars 10 forks source link

Usage of langfuse.com #20

Closed A-wels closed 2 days ago

A-wels commented 2 days ago

Hello, can you tell me what langfuse.com is used for in the code? What data is send to their servers? Is any content of the notes included in the callback in llm.ts (l. 158)?

Also, your public and secret key for the API is included in the repo. It this intentional?

lucagrippa commented 2 days ago

Hi @A-wels, thank you so much for bringing this to my attention.

I just removed Langfuse in the latest release. I was using it to debug the LLM's outputs during testing and I accidentally left it in the code. The API keys have been deleted so even if users do not update none of the data will reach Langfuse's servers.

I never intend to collect any data on AI Tagger's users now or in the future.

Thanks again for pointing this out.

A-wels commented 2 days ago

Thanks for clarifying! I think we've all accidentally left debug code in our published work at some point...

I did not want to write this publicly, but did not manage to find any contact point, sorry for that.

marcklingen commented 2 days ago

Langfuse maintainer here

Just fyi, if you do not specify credentials / host for the Langfuse instrumentation, no data is logged. Thus you could keep instrumentation in production code without risking that data is logged.

Also, you can easily run Langfuse on your local machine via docker compose or add this to the documentation if someone wants to look at traces of what obsidian-ai-tagger is doing.

Example: Langflow which allows users to enable logging by via environment variables (docs).