# Implementation Plan: 001-pantry-meal-planner

*Created: 2026-08-18*

Turn the spec into concrete technical decisions. Stay minimal: prefer the
simplest structure that satisfies the spec, and justify anything that isn't.

## Technical Context

- Language: Python 3 (stdlib only)
- Dependencies: none at runtime; pytest for tests
- Storage: two JSON files in the project dir — pantry.json, recipes.json
  (starter set written on first run)
- Testing: pytest, TDD; planner logic tested against handwritten pantries

## Structure

```
pantry_planner/
  planner.py      # pantry CRUD, plan generation, shopping list derivation
  cli.py          # argparse: pantry add/remove/list, plan, shoplist
  recipes.json    # bundled starter set (10 recipes)
  tests/test_planner.py
```

## Decisions

- D1: Greedy day-by-day plan generation with stock simulation — because R2 only
  demands feasibility, not optimality; rejected: constraint solver (overkill).
- D2: JSON files over SQLite — because two flat collections with <1k entries
  need no queries; rejected: SQLite (adds schema/migration weight).
- D3: Shopping list = sum(plan needs) − stock, clamped at 0 per item (R3) —
  computed on demand, never stored; rejected: persisting derived data.