Compute similarity score between two diffs using edit distance (0.0 = different, 1.0 = identical).

Helper for parallel similarity computation.

Copy round log to agent environment.

Format trajectory with step numbers and full agent output (not just actions).

Categorize loss reason hallucinations.

Helper function to load metadata for a single game folder in parallel

Negative log-likelihood for Bradley-Terry model with L2 regularization. Args: strengths: Array of player strengths (length n_play

Run a single simulation. Args: sim_meta: Simulation metadata with team assignments agents: List of agents (for name l

Run a single simulation. Args: sim_meta: Simulation metadata with team assignments agents: List of agents (for name l

Run a single Fastchess match between two engines. Args: agent1: First agent agent2: Second agent

Run a single halite simulation and return the output.

Run a single robotrumble simulation and return the output.

Run a single Bridge game simulation.

Run a single battlesnake simulation and return log and result outputs.

Run a single halite simulation and return the output.

Worker thread that processes tasks using a specific game worker.

Convenience function for backward compatibility.

Get line count analysis data for the current game

Create DummyArena instance with mocked environment.

Create DummyArena instance.

Create BridgeArena instance with mocked environment.

Create RoboCodeArena instance with mocked environment.

Create RoboCodeArena instance.

Create CoreWarArena instance with mocked environment.

Create CoreWarArena instance.

Create HaliteArena instance with mocked environment.

Create HaliteArena instance.

Create BattleCodeArena instance with mocked environment.

Create BattleCodeArena instance.

Create BattleCode23Arena instance with mocked environment.

Create BattleCode23Arena instance.

Create BattleCode23Arena instance.

Create RobotRumbleArena instance with mocked environment.

Create RobotRumbleArena instance.

Create HuskyBenchArena instance with mocked environment.

Create HuskyBenchArena instance.

Create FiggieArena instance with mocked environment.

Create ChessArena instance with mocked environment.

Create ChessArena-like instance with local logging directory.

Create BattleCode24Arena instance with mocked environment.

Create BattleCode24Arena instance.

Create BattleCode24Arena instance.

Create BattleSnakeArena instance with mocked environment.

Create BattleSnakeArena instance.

API endpoint to get AWS Batch jobs with caching (2 min timeout) Concurrent requests are coalesced to prevent multiple simultaneous AWS API calls.

AWS Batch job monitor page

Index page: every game with its winner, linking to its replay by game index. For a ladder (games carry a ``group`` matchup label) a Matchup colum

A complete, self-contained replay page (CSS + player + arena JS + data inlined). Built on demand by the server and served straight from memory —

Get number of completed rounds (excluding round 0 warmup)

Jaccard similarity on line-level tokens.

Game configuration from PvP metadata.

Create a mock player with specified file system contents.

Return SCML negotiation intents for the trusted runtime to validate. Return {} or None to use the trusted greedy fallback.

Return a discrete CybORG action for the trusted runtime to validate.

Delete an experiment folder

Download a file with proper security checks

Passthrough to `python -m codeclash.analysis.metrics.elo` (rich analysis + plots).

Simulate command execution based on configured outputs.

Execute a round: compile all agents, then run simulations. Handles failures gracefully: - If one agent fails to compile, the other wi

Execute a round of Bridge games.