game-ai
Purpose
This skill develops AI algorithms for games, focusing on pathfinding (e.g., A* algorithm), decision trees for NPC behaviors, and machine learning integration (e.g., using TensorFlow for training models). It helps automate AI logic in game development workflows.
When to Use
- When implementing NPC navigation in games, such as finding optimal paths in a grid-based world.
- For creating decision-making systems, like enemy AI choosing actions based on game states.
- Integrating ML for adaptive AI, such as training models to predict player moves in real-time simulations.
Key Capabilities
- Pathfinding: Implements A* algorithm with configurable heuristics; supports grid-based maps up to 100x100 cells.
- Decision Trees: Builds trees from JSON config files, e.g., {"node": "if health < 50 then flee"}; evaluates in under 10ms per decision.
- Machine Learning Integration: Wraps TensorFlow APIs for model training; uses endpoints like
/api/ml/trainwith input vectors for reinforcement learning in games. - Optimization: Includes flags for performance tuning, such as
--optimize-memoryto reduce heap usage by 20% in pathfinding routines.
Usage Patterns
To use this skill, invoke it via OpenClaw's CLI or API, passing required parameters. Always set the environment variable $GAME_AI_API_KEY for authentication. For pathfinding, call a function with a start/end point and grid; for decision trees, load a config and evaluate inputs. Structure code to handle asynchronous responses, e.g., wrap API calls in try-catch blocks.
Common Commands/API
- CLI Command:
openclaw game-ai pathfind --start 0,0 --end 10,10 --grid '{"width":20,"height":20,"obstacles":[[5,5]]}'
- Code Snippet: import openclaw result = openclaw.run('game-ai pathfind', {'start': '0,0', 'end': '10,10'}) print(result['path']) # Outputs: [[0,0], [1,0],...]
- API Endpoint: POST to
/api/game-ai/decision-treewith JSON body {"tree": {"root": "if enemy_near then attack"}, "input": {"enemy_near": true}}
- Code Snippet: import requests headers = {'Authorization': f'Bearer {os.environ["GAME_AI_API_KEY"]}'} response = requests.post('https://api.openclaw.com/api/game-ai/decision-tree', json={'tree': {...}}, headers=headers) print(response.json()['decision']) # e.g., 'attack'
- Config Format: Use JSON for inputs, e.g., {"algorithm": "A*", "params": {"heuristic": "manhattan"}}; validate with
--validate-configflag to check for errors before execution.
Integration Notes
Integrate by importing the OpenClaw SDK and initializing with $GAME_AI_API_KEY. For game engines, add as a module in Unity (via C# scripts) or Unreal (via Blueprints). Ensure compatibility by matching versions, e.g., use OpenClaw SDK v2.5+. For ML, link to external libraries like TensorFlow by adding pip install tensorflow and configuring via env vars, e.g., $TF_MODEL_PATH=/path/to/model.h5. Test integrations in a sandbox environment to avoid game loop interruptions.
Error Handling
Always check for API errors by inspecting response codes (e.g., 401 for unauthorized, handled via retry with $GAME_AI_API_KEY). For invalid inputs, use CLI flag --debug to log details, e.g., openclaw game-ai pathfind --start invalid --debug. In code, catch exceptions like ValueError for malformed grids:
- Code Snippet:
try: path = openclaw.run('game-ai pathfind', params) except ValueError as e: print(f"Error: {e} - Fix grid format and retry")
Validate configs before use, e.g., with a pre-check function, and implement retries for network failures up to 3 attempts with exponential backoff.
Concrete Usage Examples
- Pathfinding in a 2D Game: To find a path for an NPC around obstacles, run
openclaw game-ai pathfind --start 1,1 --end 5,5 --grid '{"width":10,"obstacles":[[3,3]]}'. This returns a list of coordinates; integrate into your game loop by updating the NPC's position based on the path array. - Decision Tree for Enemy AI: Build a tree with
openclaw game-ai build-tree --config '{"root": "if player_health < 20 then heal"}', then evaluate in-game: Use the API to check decisions, e.g., POST to/api/game-ai/decision-treewith current game state, and trigger actions like healing if the response is "heal".
Graph Relationships
- Related Clusters: game-dev (direct parent for game-related skills).
- Related Tags: artificial-intelligence (shares ML components), pathfinding (core functionality overlap).
- Connections: Links to skills like "game-engine" for integration, and "ml-tools" for advanced training; forms a subgraph with "game-ai" as a central node for AI in gaming ecosystems.