examples/README.md
This catalog highlights the examples shipped with Agent-lightning.
Community-contributed examples and recipes are available in the contrib directory.
| Example | Description | CI Maintenance |
|---|---|---|
| apo | Automatic Prompt Optimization tutorials covering built-in, custom, and debugging workflows. | |
| azure | Supervised fine-tuning with Azure OpenAI. | |
| calc_x | VERL-powered math reasoning agent training that uses AutoGen with an MCP calculator tool. | |
| chartqa | Vision-language ChartQA agent that reasons over charts with LangGraph and VERL plus multi-step self-refinement. | |
| claude_code | Claude Code SWE-bench harness that records Agent-lightning traces across Anthropic, vLLM, and OpenAI-compatible backends. | |
| minimal | Bite-sized programs that demonstrate how individual Agent-lightning building blocks behave in isolation. | |
| rag | Retrieval-Augmented Generation pipeline targeting the MuSiQue dataset with Wikipedia retrieval. | |
| spider | Text-to-SQL reinforcement learning training on the Spider dataset using LangGraph. | |
| tinker | Reinforcement learning with Tinker as the backend training service. | |
| unsloth | Supervised fine-tuning example powered by Unsloth with 4-bit quantization and LoRA. |
examples-* workflow statusCI status above avoids taking any workflow running with latest dependencies into account. That's why we reference the corresponding badge-* workflows instead. The following table displays the raw examples-* workflow status whenever the project is maintained by CI.