Tutorials

End-to-end walkthroughs that take you from zero to a working evaluation - running your first evaluation, bringing your own data and scoring logic, and connecting agents from popular platforms.

Run Evaluation From AI Atlas

Download and run a ready-made evaluation package with lf init eval --key.

Convert Your Logs into Datasets

Turn external agent conversation logs into Trace datasets.

Write a Sample-Dependent Scorer

Apply per-sample scoring rules for agent benchmarks like tau-bench and SWE-bench.

Evaluation for Intended-Use Evasion

Run a multi-turn adversary against your agent and score whether it stayed in scope.

Measure RAG Answer Completeness

Score whether your RAG chatbot covers every fact a correct answer needs (recall).

LangSmith

Connect a stateful LangGraph agent on the LangGraph Platform.

Dify

Connect a Dify agent-chat app with server-side conversation state.

Claude Managed Agents

Connect a Claude Managed Agents deployment as a custom-inference model.

Azure Foundry

Connect an Azure AI Foundry agent while preserving conversation state.

AWS Bedrock

Connect an AWS Bedrock AgentCore Harness as a custom-inference model.