A public, fully runnable and tested TypeScript training repo for the GH-600 GitHub Agentic AI Developer course, module 3.0 - Memory, State and GitHub Context.
Built by Certy (Coded Vision Design). Part of the CertyPro course materials.
Memory is not "store everything". A capable agent decides what to remember, never stores secrets, survives across tasks, and resumes from checkpoints. In multi-agent workflows it locks shared state so two agents never write it at once.
The repo is deterministic: no API keys, no network, no model calls. Every behaviour is reproducible and covered by tests.
| Area | Files |
|---|---|
| Memory | src/memory/ - short-term buffer, episodic log, semantic recall, policy gate, JSON store |
| State | src/state/ - workflow status machine, serialiser, in-memory lock, checkpoints |
| Tests | tests/ - memory policy and recall, state locking, serialise round-trip |
| Docs | docs/ - architecture, serialisation, retention, privacy |
| Labs | labs/README.md - eight guided exercises |
| Examples | examples/ - sample episodic run records |
Requires Node.js 22 or newer.
npm install --no-fund --no-audit
npm test # run the test suite
npm run typecheck # strict TypeScript, no emit.github/workflows/ci.yml runs install, test and typecheck on every push and
pull request to main. .github/dependabot.yml keeps npm and GitHub Actions
dependencies up to date weekly.
- Certy: https://certy.pro
- Course content: https://github.com/CertyPro/certy-gh600-course-content
- CertyPro organisation: https://github.com/CertyPro
MIT - see LICENSE. British English is used throughout.