agentbbs/internal/games/bot.go
Anthony Ettinger cbc9069964
M3: AgentGames — agent-vs-agent games, ELO ladder, replays (#7)
A Gym-style game engine (PRD §5.2) with two transports sharing one
matchmaker, so an SSH agent and a WebSocket agent can be paired together.

Engine (internal/games):
  - Game/State contract (immutable positions); registry/catalog.
  - Phase-1 games: Tic-Tac-Toe (ttt) and Connect 4 (c4).
  - ELO (K=32, start 1500), a generic win/block/random GreedyBot.
  - Transport-agnostic NDJSON protocol + match driver: hello → state →
    move → result. We run no agent code — illegal move / per-move timeout /
    disconnect all forfeit (strict validation in place of a sandbox).
  - Matchmaker: per-game queue, bounded queue-wait; never abandons a match
    that started racing the wait timeout.

Transports:
  - SSH route game@ (ssh game@host ttt | join message), registered key,
    no PTY.
  - WebSocket /play (wss), bearer API token (agentbbs mint-token <user>);
    loopback behind Caddy.

Store: game_ratings (ELO ladder) + game_matches (full move log for replay)
+ api_tokens; Rating/SaveMatch satisfy games.Store; TopRatings/RecentMatches/
MatchByID/MintAPIToken/UserByToken. Banned accounts blocked.

Hub: plugins/agentgames — browse ladders, watch move-by-move replays, and
practice vs the bot (off the rated ladder).

Tests: engine (win/draw/legality), ELO, bot, full match via matchmaker with
replay, transport (deadline/closed), store round-trips. Verified live over
SSH (agent-vs-agent), WebSocket↔SSH cross-transport, forfeit-on-illegal-move,
and the hub ladder/replay views. Docs in docs/agentgames.md (the canonical
protocol spec, to mirror to logicsrc.com); README M3 → done.

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-14 03:30:12 -07:00

71 lines
1.7 KiB
Go

package games
import "math/rand"
// Bot picks a move for the player to move in a position. It powers the in-BBS
// human-vs-bot practice mode (bots never play rated agent matches).
type Bot interface {
Name() string
Move(s State) string
}
// GreedyBot is a light heuristic that works for any game implementing State:
// it takes an immediately winning move if one exists, blocks the opponent's
// immediate win if forced, and otherwise plays a random legal move. It is a
// fine sparring partner without being game-specific.
type GreedyBot struct{ R *rand.Rand }
func (GreedyBot) Name() string { return "greedy-bot" }
func (b GreedyBot) Move(s State) string {
legal := s.Legal()
if len(legal) == 0 {
return ""
}
me := s.ToMove()
// 1) Win now if we can.
for _, m := range legal {
if ns, err := s.Apply(m); err == nil {
if over, w := ns.Terminal(); over && w == me {
return m
}
}
}
// 2) Block an opponent move that would let them win next turn.
for _, m := range legal {
ns, err := s.Apply(m)
if err != nil {
continue
}
if over, _ := ns.Terminal(); over {
continue
}
if !opponentCanWin(ns) {
// This move does not hand the opponent an immediate win; prefer it.
return m
}
}
// 3) Otherwise random.
return legal[b.pick(len(legal))]
}
// opponentCanWin reports whether the player to move in s has an immediate win.
func opponentCanWin(s State) bool {
opp := s.ToMove()
for _, m := range s.Legal() {
if ns, err := s.Apply(m); err == nil {
if over, w := ns.Terminal(); over && w == opp {
return true
}
}
}
return false
}
func (b GreedyBot) pick(n int) int {
if b.R != nil {
return b.R.Intn(n)
}
return rand.Intn(n)
}