Skip to content

AI_README.md

Latest commit

 

History

History
61 lines (46 loc) · 2.23 KB

AI_README.md

File metadata and controls

61 lines (46 loc) · 2.23 KB

SparkEngine AI Prompt Library

Standardized prompts for AI-assisted development with SparkEngine. Compatible with GitHub Copilot, GPT, Claude, Perplexity, and Gemini.

Quick Start

  1. GitHub Copilot: Automatically loads .github/copilot-instructions.md

    • Use #prompt:<name> to load specific prompt files
    • Example: #prompt:02_editor_ui_tasks

  2. Other AI assistants: Use the assistant-specific prompt files in .github/prompts/

    • OPENAI_PROMPTS.md for GPT
    • CLAUDE_PROMPTS.md for Claude
    • PERPLEXITY_PROMPTS.md for research queries
    • GEMINI_PROMPTS.md for Gemini

Prompt Categories

Core Setup

  • Project overview and architecture
  • Refactoring rules and code standards
  • Build system (CMake) guidance

Engine Systems

  • Core engine, graphics (DX11), audio (XAudio2), physics (Bullet)
  • Collision detection, post-processing, IBL lighting

Gameplay

  • Game logic, weapons, projectiles
  • Shader authoring (HLSL/GLSL)
  • Console commands and debug tools

Tools & Editor

  • ImGui editor panels
  • Asset pipeline and browser
  • Deployment and debugging workflows

Advanced

  • Networking, scripting (AngelScript), AI/NavMesh
  • Performance optimization, testing

SparkEngine Context

The prompt library covers SparkEngine's full stack:

  • Rendering: DirectX 11 pipeline with PBR, deferred/forward+, post-processing (bloom, tone mapping, FXAA), IBL, shadow mapping (PCF/VSM/CSM), particles, decals
  • Physics: Jolt Physics with rigid bodies, constraints, raycasting, collision callbacks, character controller, vehicles, ragdoll
  • Audio: XAudio2 3D spatial audio with object pooling
  • ECS: EnTT-based entity component system
  • Editor: ImGui with docking, gizmos (ImGuizmo), node graphs (imnodes)
  • Scripting: AngelScript with hot-reload and engine API bindings
  • Gameplay: FPS controller, weapons, vehicles, class system, HUD, terrain, mesh LOD
  • Tooling: Crash handler, debug console (200+ commands), profiler, scene serializer

Best Practices

  • Load the project overview prompt first for context
  • Choose task-specific prompts for focused work
  • Request complete, compile-ready code with error handling
  • Validate thread safety and performance impact
  • Include console integration for new systems