notebook.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 911 Dispatch Triage — Walkthrough Notebook\n",
    "\n",
    "This notebook walks through all three difficulty levels of the environment.\n",
    "Run each cell in order after starting the server (`uv run server`)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Setup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dispatch_triage.client import DispatchEnv\n",
    "from dispatch_triage.models import DispatchAction\n",
    "\n",
    "env = DispatchEnv(server_url='ws://localhost:8000')\n",
    "print('Connected.')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Helper — pretty-print observation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def show(obs):\n",
    "    print('\\n' + '='*60)\n",
    "    print(obs.message)\n",
    "    print()\n",
    "    print('INCIDENTS')\n",
    "    for inc in obs.incidents:\n",
    "        status = f'→ unit {inc.assigned_unit_id} (rank {inc.dispatch_rank})' if inc.resolved else 'WAITING'\n",
    "        deps   = f'  [depends on: {inc.depends_on}]' if inc.depends_on else ''\n",
    "        print(f'  [{inc.id}] {inc.type:<16} sev={inc.severity}  {inc.location:<12} {status}{deps}')\n",
    "    print()\n",
    "    print('UNITS')\n",
    "    for u in obs.units:\n",
    "        avail = 'available' if u.available else 'deployed'\n",
    "        print(f'  [{u.id}] {u.type:<12} {avail}')\n",
    "    print()\n",
    "    print(f'Score so far: {obs.score_so_far:.3f}   Done: {obs.done}')\n",
    "    print('='*60)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "## EASY — 3 incidents, 3 units\n",
    "\n",
    "Optimal strategy: dispatch by descending severity with matching unit types."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "obs = env.reset(difficulty='easy')\n",
    "show(obs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 1: cardiac_arrest (sev=9, id=0) → ambulance (id=0)\n",
    "result = env.step(DispatchAction(incident_id=0, unit_id=0))\n",
    "show(result.observation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 2: car_crash (sev=5, id=1) → police (id=1)\n",
    "result = env.step(DispatchAction(incident_id=1, unit_id=1))\n",
    "show(result.observation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 3: fire (sev=3, id=2) → fire_truck (id=2)\n",
    "result = env.step(DispatchAction(incident_id=2, unit_id=2))\n",
    "show(result.observation)\n",
    "print(f'\\nFINAL SCORE (easy, optimal): {result.observation.score_so_far:.3f}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "## MEDIUM — 5 incidents, 3 units\n",
    "\n",
    "Only 3 units available for 5 incidents.  \n",
    "Optimal: cover the 3 highest-severity incidents with matching units."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "obs = env.reset(difficulty='medium')\n",
    "show(obs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 1: cardiac_arrest (sev=9, id=0) → ambulance (id=0)\n",
    "result = env.step(DispatchAction(incident_id=0, unit_id=0))\n",
    "show(result.observation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 2: fire (sev=8, id=1) → fire_truck (id=1)\n",
    "result = env.step(DispatchAction(incident_id=1, unit_id=1))\n",
    "show(result.observation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 3: gas_leak (sev=7, id=2) → police (id=2)  [mismatch! fire_truck is correct]\n",
    "# This illustrates the penalty; ideally you'd have a second fire_truck.\n",
    "result = env.step(DispatchAction(incident_id=2, unit_id=2))\n",
    "show(result.observation)\n",
    "print(f'\\nFINAL SCORE (medium, with one mismatch): {result.observation.score_so_far:.3f}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "## HARD — 7 incidents, 3 units + cascade dependency\n",
    "\n",
    "**Key insight:** Incident 0 (cardiac_arrest, sev=7) spikes to sev=11 if the adjacent  \n",
    "gas leak (incident 1, sev=6) is not resolved first.  \n",
    "Greedy ordering (dispatch fire at sev=8 first) is sub-optimal here."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "obs = env.reset(difficulty='hard')\n",
    "show(obs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 1: gas_leak (sev=6, id=1) → fire_truck (id=1) — PREVENT the cascade first\n",
    "result = env.step(DispatchAction(incident_id=1, unit_id=1))\n",
    "show(result.observation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 2: cardiac_arrest (sev=7, id=0) → ambulance (id=0) — no cascade now\n",
    "result = env.step(DispatchAction(incident_id=0, unit_id=0))\n",
    "show(result.observation)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Step 3: fire (sev=8, id=2) → police (id=2) — only unit left, mismatch penalty\n",
    "result = env.step(DispatchAction(incident_id=2, unit_id=2))\n",
    "show(result.observation)\n",
    "print(f'\\nFINAL SCORE (hard, cascade-aware): {result.observation.score_so_far:.3f}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Score Summary\n",
    "\n",
    "| Difficulty | Optimal approach | Expected score |\n",
    "|---|---|---|\n",
    "| easy | Greedy by severity + correct units | ~1.0 |\n",
    "| medium | Greedy top-3 severity + correct units | ~0.85 |\n",
    "| hard | Resolve gas leak first, then greedy | ~0.80 |"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.11.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}