Agent Bootcamp
A collection of reference implementations of agentic retrieval-augmented generation, evaluation and orchestration workflows, built for Vector Institute's Agent Bootcamp using current Python agent frameworks.
Agent Bootcamp is a set of reference implementations demonstrating modern agentic RAG workflows in Python. It includes a basic, framework-free Reason-and-Act agent for knowledge retrieval, a version of the same agent built on the OpenAI Agents SDK to reduce boilerplate, and a multi-agent deep-research setup that trades off latency, cost and flexibility across an efficient and a verbose variant.
Its evaluation modules cover automated LLM-as-a-judge scoring and synthetic evaluation-data generation, both integrated with Langfuse for tracing agent behavior. The basic, no-framework implementations are intentionally verbose to show how agents work under the hood and are not meant as a starting point for production projects.