Seonglae ChoLangfuse focuses on operational logging and monitoring of LLM applications
LangSmith is a service provided by LangChain that supports experimentation, evaluation, debugging, and systematic development workflows for designing and iterating on LLM-based applications.
While LangFuse focuses on operational logging and monitoring of LLM applications, LangSmith is a dedicated end-to-end tool for LangChain that supports the entire development cycle of LLM applications.
LangFuse Usages
LangSmith
- [Project = Session] Sessions are also referred to as "Projects" in the UI
- Trace - an execution composed of a set of runs (UI-only, no fixed scheme)
- Run - a unit of LLM call or function
- Message - a unit of conversation request & answer
LangFuse
- (Session) optional trace grouping from user side
- Trace - a unit of interaction
- Observation - a unit of LLM call or function
MLFlow
- Run - an execution
- Trace
- Span - an action
LangSmith Run = LangFuse Observation
LangSmith Trace = LangFuse Session
Rate Limit
Are there any limits to the Langfuse API? - Langfuse
While the Langfuse API is extremely open and flexible, there are some limits to ensure the stability and performance of the platform.
https://langfuse.com/faq/all/api-limits
Open Source LLM Observability via OpenTelemetry - Langfuse
How to connect Langfuse with OpenTelemetry to observe LLM applications.
https://langfuse.com/docs/opentelemetry/get-started