LangSmith
LangChain's observability and debugging platform for tracing LLM calls, evaluating outputs, and catching issues in production.
Think of it like
A debugger for your LLM app—set breakpoints on prompts, inspect tool outputs, replay failures.
Example
Your RAG app makes a weird answer. In LangSmith, you see the exact retrieval results, the prompt that was sent, the model response, and the tool calls—drill down to diagnose the bug.
How it actually works
LangSmith logs every LLM call and tool invocation in a trace tree. You can tag runs, write evals (pass/fail checks), and compare eval scores across commits. It integrates with LangChain but works with any app via their SDK. Strong for catching regressions before production.
For product teams
Reduces debugging time on production LLM issues; enables metric-driven improvements.
For engineers
Trace-based debugging; custom evaluation logic; replay and diff of runs.
Read anything AI without the jargon
Look up any term in plain English, or save terms as you read with the free Chrome extension.
Open DecoderAdd to Chrome