awesome-agentic-ai-zh is a trilingual learning roadmap that guides developers through agentic AI concepts and implementation.
The project addresses the challenge of learning agentic AI systems in a structured way. It defines an AI Agent as a system that autonomously judges and takes action toward human-defined goals, using tools as needed and adapting based on results. The roadmap organizes learning into three progressive phases: understanding LLM fundamentals like prompts and tokens, building functional systems where models call tools and run agent loops, and finally ensuring reliability through permissions, evaluation, and observability. Rather than creating a comprehensive encyclopedia, the project curates existing resources, links to official documentation and established learning materials, and provides small executable exercises that learners can run either in the cloud or locally.
The roadmap suits developers at different starting points. Those with no programming background begin at Stage 0 with foundational setup guidance. Developers already comfortable with Python, Git, and APIs start at Stage 1 with LLM basics. The project offers two main tracks: Track A for users of CLI agents like Claude Code who want to accomplish tasks without building systems, and Track B for those writing their own agents and workflows. A separate branch serves everyday users who want to use AI safely without programming. The project introduces technical terminology in plain language before using formal English terms, and maintains a glossary for reference.
The project maintains active engagement with its learning materials through regular updates to its roadmap structure and resource curation. The repository demonstrates ongoing development of its educational content across multiple stages and tracks, with continuous refinement of how concepts are presented to different audience levels. The project sustains a trilingual approach across Traditional Chinese, Simplified Chinese, and English, requiring consistent maintenance across all three versions.
