Merge branch 'main' into john/ladder

Author: john-b-yang

.env.example

@@ -4,6 +4,13 @@
 # Required: GitHub token with repo access for cloning game repositories
 GITHUB_TOKEN=your_github_token_here
 
-# Optional: LLM Provider API Keys (configure the ones you plan to use)
+# LLM provider API keys — set the ones for the models you run (see configs/models.yaml).
+# Models are resolved by litellm from their provider-prefixed names.
+ANTHROPIC_API_KEY=
 OPENAI_API_KEY=
-ANTHROPIC_API_KEY=
+GEMINI_API_KEY=
+XAI_API_KEY=
+DASHSCOPE_API_KEY=
+
+# Optional: only needed for legacy configs that still set `model_class: portkey`.
+PORTKEY_API_KEY=

.github/mlc_config.json

@@ -18,6 +18,18 @@
     {
       "pattern": "https://docs\\.codeclash\\.io"
     },
+    {
+      "pattern": "https://robotrumble\\.org/boards/2"
+    },
+    {
+      "pattern": "https://robocode\\.sourceforge\\.io.*"
+    },
+    {
+      "pattern": "https://huskybench\\.com/.*"
+    },
+    {
+      "pattern": "https://corewar\\.co\\.uk"
+    },
     {
       "pattern": "https?://(.*\\.)?twitter\\.com/.*"
     },
@@ -26,6 +38,9 @@
     },
     {
       "pattern": "https://www\\.contributor-covenant\\.org/version/2/1/code_of_conduct\\.html"
+    },
+    {
+      "pattern": "https://join\\.slack\\.com/t/swe-bench/shared_invite/.*"
     }
   ]
 }

.github/workflows/check-links.yaml

@@ -2,16 +2,29 @@ name: Check Markdown links
 
 on:
   workflow_dispatch:
-  push:
   pull_request:
   schedule:
     - cron: "0 0 1 * *"
 
 jobs:
-  markdown-link-check:
+  markdown-link-check-pr:
+    if: github.event_name == 'pull_request'
     runs-on: ubuntu-latest
     steps:
-      - uses: actions/checkout@master
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+      - uses: gaurav-nelson/github-action-markdown-link-check@v1
+        with:
+          config-file: '.github/mlc_config.json'
+          check-modified-files-only: 'yes'
+          base-branch: 'main'
+
+  markdown-link-check-full:
+    if: github.event_name == 'workflow_dispatch' || github.event_name == 'schedule'
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
       - uses: gaurav-nelson/github-action-markdown-link-check@v1
         with:
           config-file: '.github/mlc_config.json'

.github/workflows/pre-commit.yaml

@@ -0,0 +1,23 @@
+name: Pre-commit
+
+on:
+  push:
+  pull_request:
+
+jobs:
+  pre-commit:
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        shell: bash -l {0}
+    steps:
+      - uses: actions/checkout@v5
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.11"
+      - name: Install uv
+        run: curl -LsSf https://astral.sh/uv/install.sh | sh
+      - name: Install dependencies
+        run: uv sync --extra dev
+      - name: Run pre-commit
+        run: uv run pre-commit run --all-files --show-diff-on-failure

.pre-commit-config.yaml

@@ -15,7 +15,7 @@ repos:
       - id: debug-statements
 
   - repo: https://github.com/crate-ci/typos
-    rev: v1
+    rev: v1.46.0
     hooks:
       - id: typos
         files: \.(py|md|rst|yaml|toml)

README.md

@@ -95,6 +95,19 @@ Critically, *LMs don't play the game directly*.
 Their code serves as their competitive proxy.
 The winner is the LM agent who wins the most rounds.
 
+## 🧩 Available Arenas
+
+CodeClash includes competitive programming games and simulation-backed arenas, including BattleSnake,
+CoreWar, CybORG, Halite, HuskyBench, RoboCode, RobotRumble, and SCML.
+
+SCML is a supply-chain negotiation arena based on the ANAC Supply Chain Management League OneShot
+track. Agents edit a Python `scml_agent.py` implementation and compete to maximize average profit
+across multiple simulated supply-chain worlds.
+
+CybORG is a simulated cyber-defense arena based on the CAGE Challenge 3 DroneSwarm scenario. Agents
+edit a Python `cyborg_agent.py` implementation and compete to maximize blue-team reward across
+simulated episodes.
+
 ## 🚀 Get Involved
 
 - Check out our [docs](https://docs.codeclash.ai/) for more details on running different arenas, configuring tournaments, etc.

codeclash/agents/minisweagent.py

@@ -3,12 +3,9 @@
 import traceback
 from collections.abc import Callable
 
-from minisweagent import Model
 from minisweagent.agents.default import AgentConfig, DefaultAgent
 from minisweagent.environments.docker import DockerEnvironment
-from minisweagent.models import get_model
-from minisweagent.models.test_models import DeterministicModel
-from minisweagent.run.utils.save import save_traj
+from minisweagent.models import Model, get_model
 
 from codeclash import REPO_DIR
 from codeclash.agents.player import Player
@@ -22,10 +19,8 @@
 
 
 class ClashAgent(DefaultAgent):
-    """
-    Slightly modified version of `DefaultAgent` from mini-SWE-agent
-    (https://github.com/SWE-agent/mini-swe-agent)
-    """
+    """`DefaultAgent` from mini-SWE-agent (https://github.com/SWE-agent/mini-swe-agent)
+    with per-player debug logging."""
 
     def __init__(
         self,
@@ -39,9 +34,11 @@ def __init__(
         super().__init__(model, env, config_class=config_class, **kwargs)
         self.logger = logger
 
-    def add_message(self, role: str, content: str, **kwargs):
-        super().add_message(role, content, **kwargs)
-        self.logger.debug(f"[{role}] {content}", extra={"highlighter": None})
+    def add_messages(self, *messages: dict) -> list[dict]:
+        result = super().add_messages(*messages)
+        for m in messages:
+            self.logger.debug(f"[{m.get('role')}] {m.get('content')}", extra={"highlighter": None})
+        return result
 
 
 class MiniSWEAgent(Player):
@@ -51,26 +48,21 @@ def __init__(self, config: dict, environment: DockerEnvironment, game_context: G
         super().__init__(config, environment=environment, game_context=game_context)
 
     def run(self):
-        # temporary workaround around https://github.com/SWE-agent/mini-swe-agent/issues/477
-        if "DeterministicModel" not in self.config["config"]["model"].get("model_class", ""):
-            model = get_model(config=self.config["config"]["model"])
-        else:
-            model = DeterministicModel(outputs=self.config["config"]["model"]["outputs"])
+        model = get_model(config=self.config["config"]["model"])
         self.agent = ClashAgent(
             model=model,
             env=self.environment,
             logger=self.logger,
             **self.config["config"]["agent"],
         )
         exit_status = None
-        result = None
         exc_message = None
         try:
-            exit_status, result = self.agent.run(task="", **self.game_context.to_template_vars())
+            result = self.agent.run(task="", **self.game_context.to_template_vars())
+            exit_status = result.get("exit_status", "")
         except Exception as e:
             exit_status = str(e)
             exc_message = traceback.format_exc()
-            result = exc_message
             self.logger.critical(exc_message)
         finally:
             traj_path = (
@@ -79,22 +71,16 @@ def run(self):
                 / self.name
                 / f"{self.name}_r{self.game_context.round}.traj.json"
             )
-            save_traj(
-                self.agent,  # type: ignore
-                traj_path,
-                exit_status=exit_status,
-                result=result,
-                print_fct=self.logger.debug,
-            )
+            self.agent.save(traj_path)
             copy_to_container(
                 self.environment,
                 traj_path,
                 self.game_context.log_env / "edits" / traj_path.name,
             )
             self._metadata["agent_stats"][self.game_context.round] = {
                 "exit_status": exit_status,
-                "cost": self.agent.model.cost,
-                "api_calls": self.agent.model.n_calls,
+                "cost": self.agent.cost,
+                "api_calls": self.agent.n_calls,
             }
         if exit_status.lower().strip() not in ["", "submitted", "limitsexceeded"] and exc_message is not None:
             raise RuntimeError(f"Agent {self.name} failed with exit status: {exit_status} and exception: {exc_message}")

codeclash/analysis/llm_as_judge/hallucination.yaml

@@ -71,7 +71,7 @@ system_prompt: |
     by reasoning)
   - "My bot is working perfectly" (this is just a slightly overconfident statement, but not a concrete claim that can be corroborated or disproven)
     This violates 2 (the statement of fact is not concrete)
-  - Agent using an incorrect linenumber when referring to a code snippet (as long as the agent recovers later on and this doens't cause an edit to
+  - Agent using an incorrect linenumber when referring to a code snippet (as long as the agent recovers later on and this doesn't cause an edit to
     fail without being able to recover). This violates 6 (the incident is not relevant to the overall trajectory and objective of the agent)
   - Anything related to failed edits as long as the failure is spotted and corrected later on.
 

codeclash/arenas/__init__.py

@@ -6,6 +6,7 @@
 from codeclash.arenas.bridge.bridge import BridgeArena
 from codeclash.arenas.chess.chess import ChessArena
 from codeclash.arenas.corewar.corewar import CoreWarArena
+from codeclash.arenas.cyborg.cyborg import CybORGArena
 from codeclash.arenas.dummy.dummy import DummyArena
 from codeclash.arenas.figgie.figgie import FiggieArena
 from codeclash.arenas.gomoku.gomoku import GomokuArena
@@ -15,6 +16,7 @@
 from codeclash.arenas.huskybench.huskybench import HuskyBenchArena
 from codeclash.arenas.robocode.robocode import RoboCodeArena
 from codeclash.arenas.robotrumble.robotrumble import RobotRumbleArena
+from codeclash.arenas.scml.scml import SCMLOneShotArena
 
 ARENAS = [
     BattleCode23Arena,
@@ -24,6 +26,7 @@
     BridgeArena,
     ChessArena,
     CoreWarArena,
+    CybORGArena,
     DummyArena,
     FiggieArena,
     GomokuArena,
@@ -33,6 +36,7 @@
     HuskyBenchArena,
     RoboCodeArena,
     RobotRumbleArena,
+    SCMLOneShotArena,
 ]
 
 

codeclash/arenas/arena.py

@@ -12,7 +12,12 @@
 from codeclash.agents.player import Player
 from codeclash.constants import DIR_LOGS, DIR_WORK, GH_ORG, RESULT_TIE
 from codeclash.utils.aws import is_running_in_aws_batch, pull_game_container_aws_ecr
-from codeclash.utils.environment import assert_zero_exit_code, copy_between_containers, copy_from_container
+from codeclash.utils.environment import (
+    ClashDockerEnvironment,
+    assert_zero_exit_code,
+    copy_between_containers,
+    copy_from_container,
+)
 from codeclash.utils.log import get_logger
 
 
@@ -185,7 +190,7 @@ def get_environment(self, branch_name: str | None = None) -> DockerEnvironment:
             run_args = ["--rm"]
         else:
             run_args = []
-        environment = DockerEnvironment(
+        environment = ClashDockerEnvironment(
             image=self.image_name,
             cwd=str(DIR_WORK),
             env={