LangSmith
LangChain's platform for tracing, debugging and evaluating LLM applications — part of my observability and evals toolkit.
LangSmith is the hosted tracing, debugging and evaluation platform from the LangChain team. It records every step of a chain or graph run so prompts, tool calls and latencies can be inspected, and it supports datasets for regression-style Evals.
How I use it
LangSmith is part of my AI toolkit alongside LangChain and LangGraph. In my public projects I have favoured open alternatives: langgraph-langserve-starter exports traces to Langfuse, and Akashi ships an OpenTelemetry tracing exporter. None of my public projects is built specifically around LangSmith, so treat it as familiarity with the LangChain ecosystem's tooling rather than a showcase.
Why it matters
Tracing is how you debug non-deterministic systems: when an agent loops or picks the wrong tool, the trace shows which prompt and which intermediate result caused it. I apply that principle across my stack regardless of vendor — see Observability.