How vectorinstitute/agent-bootcamp works
Project type and purpose
• This repository is an open-source collection of Python reference implementations (CLI and Gradio apps) for “Agent Bootcamp,” demonstrating retrieval-augmented generation (RAG) agents, evaluation pipelines, and orchestration using modern tools.
• It is not a traditional full-stack web app, but rather a modular toolset/library with multiple example modules (“basics,” “frameworks,” “evals”), shared utilities, CI/CD, docs, and tests.
File structure
Top-level directories and files:
• src/ – contains three main module groups:
– 1_basics: minimal proof-of-concept ReAct agents
– 2_frameworks: SDK-based agents (OpenAI Agents SDK, Langfuse integration, multi-agent, code interpreter)
– 3_evals: evaluation pipelines (LLM-as-judge, synthetic data)
– prompts.py: shared prompt templates
– utils/: shared helpers (async, data loading, env vars, Gradio messages, logging, Langfuse setup, tools for code interpreter, Weaviate, etc.)
• docs/, mkdocs.yml – documentation site for the project
• tests/ – unit/integration tests organized by data and tool tests
• .github/workflows/ – CI pipelines for linting, docs, tests
• pyproject.toml – project metadata and dependencies
README for additional insights
• Confirms the three module categories and their purposes.
• Highlights external services: OpenAI API, Langfuse for traceability, Hugging Face Datasets, Weaviate vector store.
• Describes how to run each demo via CLI or Gradio, and how to execute evaluation scripts.