[2.0] Add Generals.io bot arena task

2.0/README.md

@@ -39,6 +39,14 @@ This variant keeps the same SIFT1M-scale service contract and recall target as
 offline indexing strategies are more viable. Its problem ID is
 `vector_db_ann_relaxed`.
 
+## Generals.io Bot Arena
+
+This game-playing problem asks agents to improve a patch-based bot for a local
+Generals.io-style simulator. Its problem ID is `generals_io_bot`. The judge
+applies the submitted patch to a clean skeleton, runs a hidden arena against
+multiple baseline bot families, and scores by mean baseline win rate with a
+small faster-win tiebreak. The online generals.io service is not used.
+
 ## BBOPlace ISPD2005
 
 This VLSI placement problem asks agents to generate macro placement candidates

2.0/problems/generals_io_bot/config.yaml

@@ -0,0 +1,47 @@
+tag: games
+runtime:
+  language: patch
+  timeout_seconds: 10800
+  environment: "Generals.io bot patch; local generals-bots simulator arena"
+  apt_packages:
+    - bash
+    - ca-certificates
+    - git
+    - python3
+    - python3-pip
+  judge_apt_packages:
+    - bash
+    - ca-certificates
+    - git
+    - python3
+    - python3-pip
+  docker:
+    image: frontiercs/generals-io-bot-agent:experimental-c2b77bf
+    judge_image: frontiercs/generals-io-bot-judge:experimental-c2b77bf
+environment:
+  cpus: 4
+  memory_mb: 8192
+  storage_mb: 8192
+  build_timeout_seconds: 1800
+evaluation:
+  generals_bots_commit: "c2b77bf72812ec91fb2024d80d90112b961dfa7e"
+  arena_seed: 20260608
+  games_per_matchup: 1
+  async_start_method: spawn
+  max_eval_seconds: 240
+  truncation: 180
+  pool_size: 2
+  speed_weight: 0.25
+  grid_sizes:
+    - 10
+  baselines:
+    - random_low_split
+    - expander
+    - strongest_frontier
+    - hunter
+    - fast_pathing
+    - flobot_fast
+submission:
+  kind: file
+  path: /app/solution.patch
+  allow_empty: true

2.0/problems/generals_io_bot/docker/README.md

@@ -0,0 +1,14 @@
+# Generals.io Bot Images
+
+Build the task-specific Harbor images before running a local Harbor trial:
+
+```bash
+bash 2.0/problems/generals_io_bot/docker/build_images.sh
+```
+
+The images install `strakam/generals-bots` from pinned commit
+`c2b77bf72812ec91fb2024d80d90112b961dfa7e` plus explicit CPU `jax/jaxlib`
+dependencies used by the simulator.
+
+- Agent image: exposes `/app/generals_agent` as a git checkout for the agent.
+- Judge image: keeps a clean copy at `/opt/generals-agent-clean` for patch application.

2.0/problems/generals_io_bot/docker/agent/Dockerfile

@@ -0,0 +1,27 @@
+# syntax=docker/dockerfile:1.7
+FROM python:3.11-slim
+
+ARG GENERALS_BOTS_COMMIT=c2b77bf72812ec91fb2024d80d90112b961dfa7e
+
+ENV XLA_FLAGS="--xla_cpu_multi_thread_eigen=false intra_op_parallelism_threads=1" \
+    OMP_NUM_THREADS=1 \
+    OPENBLAS_NUM_THREADS=1 \
+    MKL_NUM_THREADS=1 \
+    NUMEXPR_NUM_THREADS=1
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends bash ca-certificates git && \
+    rm -rf /var/lib/apt/lists/*
+
+RUN pip install --no-cache-dir \
+    "git+https://github.com/strakam/generals-bots.git@${GENERALS_BOTS_COMMIT}" \
+    "jax[cpu]>=0.4.30" \
+    "jaxlib>=0.4.30"
+
+WORKDIR /app
+COPY harbor/app/generals_agent /app/generals_agent
+RUN git -C /app/generals_agent init -q && \
+    git -C /app/generals_agent config user.email frontier-cs@example.invalid && \
+    git -C /app/generals_agent config user.name "Frontier-CS" && \
+    git -C /app/generals_agent add . && \
+    git -C /app/generals_agent commit -q -m base

2.0/problems/generals_io_bot/docker/build_images.sh

@@ -0,0 +1,21 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/../../../.." && pwd)"
+TASK_DIR="$ROOT/2.0/problems/generals_io_bot"
+COMMIT="${GENERALS_BOTS_COMMIT:-c2b77bf72812ec91fb2024d80d90112b961dfa7e}"
+SHORT="${COMMIT:0:7}"
+AGENT_TAG="${AGENT_TAG:-frontiercs/generals-io-bot-agent:experimental-${SHORT}}"
+JUDGE_TAG="${JUDGE_TAG:-frontiercs/generals-io-bot-judge:experimental-${SHORT}}"
+
+docker build \
+  --build-arg GENERALS_BOTS_COMMIT="$COMMIT" \
+  -f "$TASK_DIR/docker/agent/Dockerfile" \
+  -t "$AGENT_TAG" \
+  "$TASK_DIR"
+
+docker build \
+  --build-arg GENERALS_BOTS_COMMIT="$COMMIT" \
+  -f "$TASK_DIR/docker/judge/Dockerfile" \
+  -t "$JUDGE_TAG" \
+  "$TASK_DIR"

2.0/problems/generals_io_bot/docker/judge/Dockerfile

@@ -0,0 +1,28 @@
+# syntax=docker/dockerfile:1.7
+FROM python:3.11-slim
+
+ARG GENERALS_BOTS_COMMIT=c2b77bf72812ec91fb2024d80d90112b961dfa7e
+
+ENV XLA_FLAGS="--xla_cpu_multi_thread_eigen=false intra_op_parallelism_threads=1" \
+    OMP_NUM_THREADS=1 \
+    OPENBLAS_NUM_THREADS=1 \
+    MKL_NUM_THREADS=1 \
+    NUMEXPR_NUM_THREADS=1
+
+RUN apt-get update && \
+    apt-get install -y --no-install-recommends bash ca-certificates git && \
+    rm -rf /var/lib/apt/lists/*
+
+RUN pip install --no-cache-dir \
+    "git+https://github.com/strakam/generals-bots.git@${GENERALS_BOTS_COMMIT}" \
+    "jax[cpu]>=0.4.30" \
+    "jaxlib>=0.4.30"
+
+COPY harbor/app/generals_agent /opt/generals-agent-clean
+RUN git -C /opt/generals-agent-clean init -q && \
+    git -C /opt/generals-agent-clean config user.email frontier-cs@example.invalid && \
+    git -C /opt/generals-agent-clean config user.name "Frontier-CS Judge" && \
+    git -C /opt/generals-agent-clean add . && \
+    git -C /opt/generals-agent-clean commit -q -m base
+
+WORKDIR /judge

2.0/problems/generals_io_bot/evaluate.sh

@@ -0,0 +1,3 @@
+#!/usr/bin/env bash
+set -euo pipefail
+python3 "$(dirname "$0")/evaluator.py" "$1"