AI Engineer from Scratch

What AI Engineer from Scratch does

The repository is organized as numbered topic folders, each mostly a standalone demo of .py or .ipynb files. Early folders cover function calling, ReAct, Plan-and-Execute, Reflexion, multi-agent patterns, RAG from simple to agentic, memory types, model routing and building an LLM from scratch. The second half turns to reading open-source agent source code and to engineering practice, including Loop Engineering, CI/CD and LangGraph.

Folder 08-hermes-agent holds the Hermes source-reading notes on memory, the main loop, eval, prompts, environment, cron and the gateway. Folder 12-hermes-agent-small contains waku, a local-first teaching assistant organized around the agent loop, memory and eval. Folder 13-pi-agent covers the Pi agent architecture. The README suggests reading the Hermes folder first to build a runtime mental model, then waku, then Pi for a TypeScript core and interactive split. The text is in Chinese, and the author links a Bilibili channel.

Key features

• Hermes source-reading notes on memory, main loop, eval, prompts, environment, cron and gateway
• waku, a small local-first teaching assistant in the 12-hermes-agent-small folder
• Code demos for ReAct, Plan-and-Execute, Reflexion and multi-agent architectures
• RAG examples from basic to agentic RAG, plus memory and model routing demos
• Separate folders for Pi agent, Loop Engineering, CI/CD and LangGraph

When to use it

• Learning how the Hermes Agent runtime is structured by reading annotated source notes
• Preparing for AI agent or RAG engineering interviews
• Comparing the Hermes architecture with the Pi agent's TypeScript design

Who it is for: Chinese-reading developers studying agent internals or preparing for agent engineering interviews who want to understand how Hermes Agent works.

How it fits with Hermes Agent

It is a learning resource rather than a Hermes tool: one folder of Hermes source notes and one small Hermes-inspired teaching agent sit inside a broader agent-engineering course.

Requirements: A Python environment; API keys are configured in a .env file at the repository root

Note: The README and notes are written in Chinese, and the repository has no license file.

FAQ