ai-code-battle/starters/java/README.md
jedarden 7a0de02059 feat(evolver): persist cross-pollination state to Postgres per §10.2
Add crosspoll_state table to persist per-island generation counters
across evolver restarts. Load state on startup and save after each
cross-pollination check. Add persistence pattern and translation
structure tests.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-04-22 16:04:15 -04:00

2.5 KiB

acb-starter-java

Java starter kit for AI Code Battle — a competitive bot programming platform.

Uses Javalin for the HTTP server, Jackson for JSON, and javax.crypto for HMAC authentication.

Quick Start

# Build
mvn package

# Run locally
BOT_SECRET=dev-secret java -jar target/starter-bot-1.0.0.jar

# Run with Docker
docker build -t my-bot .
docker run -e BOT_SECRET=your-secret -p 8080:8080 my-bot

Your bot listens on port 8080 and responds to POST /turn with move commands.

Register Your Bot

Once your bot is deployed and accessible via HTTPS:

curl -X POST https://api.aicodebattle.com/api/register \
  -H "Content-Type: application/json" \
  -d '{
    "name": "my-java-bot",
    "endpoint_url": "https://my-bot.example.com",
    "owner": "your-name",
    "description": "My awesome bot"
  }'

Save the bot_id and shared_secret from the response — the secret is shown only once.

Project Structure

src/main/java/com/acb/starter/App.java  # Server, auth, types, and strategy
src/main/java/com/acb/starter/Grid.java # Grid utilities (toroidal distance, BFS, neighbors)
pom.xml                                 # Maven build configuration
Dockerfile                              # Multi-stage container build

Grid Helpers

Grid.java provides static utility methods for the toroidal grid:

  • Grid.toroidalManhattan(r1, c1, r2, c2, rows, cols) — Manhattan distance with wrap-around
  • Grid.toroidalChebyshev(r1, c1, r2, c2, rows, cols) — Chebyshev distance with wrap-around
  • Grid.neighbors(row, col, rows, cols) — 8-directional neighbors with wrap
  • Grid.bfs(start, goal, passable, rows, cols) — BFS pathfinding, returns path or null

Customization

Edit computeMoves() in App.java to implement your strategy. The GameState object provides:

  • bots — all visible bots (yours and enemies)
  • energy — visible energy pickup locations
  • cores — visible core positions
  • walls — visible wall positions
  • youEnergy — your current energy count
  • youScore — your current score
  • config — match parameters (grid size, etc.)

Return a List<Map<String, Object>>, each entry with position (your bot's current position) and direction ("N", "E", "S", or "W"). Bots not included in the response stay in place.

Protocol

  • Endpoint: POST /turn — receives game state JSON, returns moves JSON
  • Health: GET /health — must return 200
  • Timeout: 3 seconds per turn
  • Auth: HMAC-SHA256 via X-ACB-Signature header