Open-source, lightweight tracer for debugging LLMs applications
Open-source, lightweight tracer for debugging LLMs applications
This allows you to set up logging traces with minimal effort on any python codebase and store it into a mongo db database or a local directory. I'm leveraging it mostly for developing on language models, and use the data for a downstream evaluation and debugging. Since I'm didn't want to rely on any of the opinionated frameworks (Langchain with Langsmith, etc), I thought a simpler plug-and-play solution would suffice. I have a lot of features in mind for the roadmap -- let me know if you'd like to add to it and feel free to drop in any suggestions. There is also a quick Streamlit app added to stare at the results; I'm not a Streamlit pro and preemptively apologize for any UI issues.
Share cardActual performance
Launch Intel predictions
Analyze your own launch →Incorrect prediction on native model
Similar products
Poozle – open-source Plaid for LLMs
Leaping – Open-source debugging with LLMs
Open-source tool for deploying applications to K8s/OpenShift
Swiftwave – Simple Lightweight Open Source PaaS
GAI, a Go-idiomatic, lightweight abstraction on top of LLMs
Grimoire – A lightweight, open-source WoW addon manager
Restfox – Open source lightweight alternative to Postman
Lightweight, open source FlatFile clone
Minimalistic, open source analytics for web applications
TensorZero – open-source data and learning flywheel for LLMs