Agentic Programming · Lab 5: Testing New Features Hadar Wolf & Ely Grunfeld
A local two-player (or player-vs-computer) chess game in Python with a browser UI. We started from the course's chess workshop starter. Working on separate branches with coding agents, we added undo, save/load and redo, merged two supplied feature branches (captured pieces and pawn promotion), and covered every combination of features with tests.
✅ 112 tests pass · ruff check clean · ruff format clean · mypy --strict clean
| Feature | Notes |
|---|---|
| Replayable history | Session.moves is the single source of truth. Undo, redo, load and promotion all rebuild the position by replaying moves (Session._rebuild), so derived state like captured pieces stays correct. |
| Undo | Against the computer, one undo takes back the computer's reply and your move. In a two-player game it takes back one move. With no moves, it does nothing. |
| Save / load | A JSON file (*.save.json) stores the opponent type, the moves and the redo history. Loading into a fresh session restores everything, including undo and redo. |
| Redo | Replays exactly the moves that undo removed, including the computer's reply and any promotion choice. A new move clears redo. |
| Captured pieces + promotion | The supplied branches, merged in order and then tested together with undo and save/load. |
| Checks | The position became a typed dataclass. ruff and mypy are configured, the computer's tie-breaks are seeded so tests are deterministic, and CLAUDE.md tells agents to run all four checks. |
Team decisions are written down in NOTES.md, and the save/load spec is in SPEC.md.
Requires Python 3.10+.
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\Activate.ps1
pip install -r requirements-dev.txt
python server.py # open http://127.0.0.1:8000python -m pytest -q --cov=. --cov-report=term-missing
ruff check .
ruff format --check .
mypy --strict .- One feature = one branch = its own tests. The merge order was undo and save/load, then captured pieces, then promotion, then redo.
- After every merge we ran all the checks and added a test of the features together (for example
test_undo_save_load.pyandtest_promotion_combined.py). - Our working plan and checklists are in
docs/lab05-plan.md. - There are two agent-made reviews:
docs/code-review.mdcovers everything merged intomain, andlab05-review.mdis the lab review of777b904. - The original assignment and the starter's README are in
docs/ASSIGNMENT.md.
game.py: board, moves, check, checkmate, stalemate (castling and en passant aren't supported)session.py: game actions, history (undo, redo, save/load), opponent, notation, sidebarcaptured.py: captured-pieces trackingcomputer.py: a computer opponent that looks one move aheadserver.py: HTTP actions plus static filesstatic/: the browser UItests/: 112 tests, including combined-feature tests
Part of my Agentic Programming labs (HUJI, AICODE 67723): Lab 1 · CTF Arena · Lab 2 · LLM failure modes · Lab 3 · Coin Dash with several agents · Lab 4 · Traffic monitor spec · Lab 5 · Chess: testing new features · Lab 6 · Binary search explainer video