Hello Agents — Build Intelligent Agents from Zero to Production

Introduction

Hello Agents is an open-source educational project by Datawhale that teaches you how to build AI agents from scratch. Rather than relying on high-level frameworks, it walks through the core primitives—tool calling, memory management, planning, and multi-agent coordination—so you understand what is actually happening under the hood.

What Hello Agents Does

• Provides a structured curriculum covering agent fundamentals, tool integration, memory systems, and RAG pipelines
• Includes runnable Jupyter notebooks for every chapter so you can learn by doing
• Demonstrates multi-agent patterns including delegation, debate, and supervisor architectures
• Covers both Python SDK usage and the underlying prompt engineering that drives agent behavior
• Teaches how to evaluate agent performance and handle failure modes gracefully

Architecture Overview

The project is organized as a series of progressive chapters, each building on the last. Early chapters cover the agent loop (observe → think → act → reflect), then introduce tool definitions and function calling, followed by retrieval-augmented generation and vector stores. Later chapters tackle multi-agent systems and production deployment patterns. All examples use standard Python libraries and can run with any major LLM provider.

Self-Hosting & Configuration

• Clone the repo and install dependencies with pip; no special infrastructure required
• Each chapter is a standalone notebook—run them in Jupyter, VS Code, or Colab
• Configure your LLM API key via environment variables (supports OpenAI, Anthropic, and local models)
• Chapters on RAG require a vector database; examples default to Chroma or FAISS
• The project supports both cloud-hosted and fully local model setups via Ollama

Key Features

• Zero-to-production curriculum designed for developers new to agent engineering
• Framework-agnostic approach that teaches principles rather than locking you into one SDK
• Bilingual documentation in Chinese and English
• Active community with regular updates as agent patterns evolve
• Covers advanced topics like self-evolving agents and human-in-the-loop workflows

Comparison with Similar Tools

• LangChain docs — framework-specific tutorials; Hello Agents is framework-agnostic and teaches fundamentals
• DeepLearning.AI courses — video-first and paywalled; Hello Agents is text-first and fully open source
• AutoGen tutorials — focused on Microsoft's multi-agent framework; Hello Agents covers multiple approaches
• OpenAI Cookbook — recipe-oriented for OpenAI APIs; Hello Agents builds understanding from first principles
• 12-Factor Agents — focuses on production principles; Hello Agents provides the step-by-step learning path to get there

FAQ

Q: Do I need ML experience to use this? A: Basic Python proficiency is sufficient. The curriculum starts from fundamentals and builds up progressively.

Q: Which LLM providers are supported? A: The examples work with OpenAI, Anthropic, local models via Ollama, and most OpenAI-compatible APIs.

Q: Is this a framework I install? A: No. It is a learning resource with runnable code. You take the patterns you learn and apply them in your own projects.

Q: How often is the content updated? A: The project follows an active contribution model with regular chapter additions as the agent ecosystem evolves.

Sources

• https://github.com/datawhalechina/hello-agents
• https://datawhalechina.github.io/hello-agents/