hello-agent-system
👋 Hello Agent System — learn enterprise AI agent system design by building every production layer from scratch. 17 bilingual lessons (EN/中文) with tested exer
What is hello-agent-system?
👋 Hello Agent System — learn enterprise AI agent system design by building every production layer from scratch. 17 bilingual lessons (EN/中文) with tested exercises: tools, architectures, reliability, security, evals, distributed execution, cost, RAG, release ops. 企业级 Agent 系统设计训练营
Overview of hello-agent-system
What it is
Hello Agent System is an educational project that teaches enterprise AI agent system design by building every production layer from scratch. It provides 17 bilingual lessons (English/Chinese) with tested exercises covering tools, architectures, reliability, security, evaluations, distributed execution, cost, RAG, and release operations. The project does not rely on any agent frameworks and is designed to help learners progress from calling LLM APIs to designing production-grade agent systems.
Key features
- 32 lessons in English and Chinese, each with exercises that include automated tests
- Core implementation of agent runtime written from scratch (approximately 3800 lines) plus distributed modules (approximately 2000 lines)
- Covers: LLM essentials, tools, context, common architectures, orchestration, reliability (retry/circuit breaker/degradation), security (prompt injection defense, RBAC, human approval), observability, evaluations, high concurrency, distributed execution, cost optimization, permission-aware RAG, gray release, incident response, retrieval quality, memory systems, MCP, sandbox, agent data, evaluation methodology, prompt optimization, test-time compute, coding agents, and proactive agents
- All code is async from lesson 02 onwards, and real multi-process execution (with crash handover, zombie workers, graceful shutdown) from lesson 12 onwards
- Supports any OpenAI-compatible interface (OpenAI, DeepSeek, Qwen, vLLM, etc.)
- Replaceable adapter layer for production components: Postgres, Redis, Temporal, OpenTelemetry, LiteLLM, Cedar
Use cases
This project is designed for developers and engineers who already know how to call an LLM API and want to learn how to build reliable, secure, and observable agent systems that can be deployed in production. Example scenarios covered include: handling service rate limits (429 errors), breaking infinite loops with budget controls, defending against prompt injection attacks, isolating tenant data, evaluating prompt changes without regressions, recovering from process crashes during agent execution, and running multiple worker processes that hand over tasks when one process dies.
Getting started
Clone the repository and see the README for instructions on setting up the project, running the agent chat example, and completing the lessons in order.bashgit clone https://github.com/heaven999b/hello-agent-systemcd hello-agent-system# See README for setup and usage
License and pricing
This project is licensed under the MIT license. It is open source and free to use.
Based on the project's public README and repository metadata.