Dispatch Route Optimizer

React + TypeScript + Tailwind rewrite of the single-driver, ~10-stop route optimizer.

Setup

npm install
npm run dev

Open the printed localhost URL. Click the map to add a stop, drag any pin to move one, and the route re-solves automatically (debounced ~500ms).

Structure

src/
  lib/
    types.ts        Stop, DurationMatrix, SolveResult, ComplexityEstimate
    geo.ts           haversine fallback + time/distance formatting
    osrm.ts          OSRM /table (matrix) and /route (geometry) calls, with
                     graceful fallback to a haversine-based estimate
    tsp.ts           heldKarp() exact solver, bruteForce() reference solver,
                     estimateComplexity() for the growth-tracking panel
  components/
    RouteMap.tsx       Leaflet map, draggable pins, click-to-add
    RoutePanel.tsx      ordered stop list with per-leg time/distance
    MatrixView.tsx      duration matrix heatmap
    ComplexityPanel.tsx brute-force vs Held-Karp op counts as stops change
    HowItWorks.tsx       algorithm walkthrough + worked diagram
  App.tsx            state, tab switching, debounced solve pipeline

Notes

• Solver: heldKarp() in lib/tsp.ts is exact (not a heuristic), running in O(n²·2ⁿ) time. Practical up to ~13-15 stops — the Complexity tab shows this live as you add/remove stops, and estimateComplexity() flags when you're leaving that practical range.
• Real drive times: getDurationMatrix() calls the public OSRM demo server. It has no SLA — for production, self-host OSRM with a regional .osm.pbf extract and point OSRM_BASE in lib/osrm.ts at it. If the call fails for any reason, it transparently falls back to a haversine-based estimate and the UI badge reflects which one you're looking at.
• Scaling past ~15 stops or adding multiple vehicles, time windows, or capacity constraints: swap heldKarp() for a call to Google OR-Tools (typically via a small Python backend - which I aim to implement as an addition to learning this) — the same distance matrix feeds either solver.