Examples
Each example is a standalone uv project under
examples/
with its own pyproject.toml, rl_app.py, reward.py, and
evaluate.py.
- Strands Math Agent — GSM8K grade-school math. Start here; covers local vLLM, ACR deploy, and evaluation.
- Strands AppWorld Agent — multi-tool AppWorld benchmark with few-shot prompting and Docker build.
- Strands MigrationBench Agent — code migration reward + async evaluation pipeline.
- Strands OfficeBench Agent — OfficeBench tool-use harness with local test and benchmark scripts.
Choosing an example
Section titled “Choosing an example”- New to the toolkit? Start with the math agent — the smallest end-to-end example.
- Building a benchmark harness? Read the OfficeBench or AppWorld example.
- Designing a custom reward? The migration agent is the clearest template.