Skip to main content

AI-Assisted Development

LightningAddon includes structured context for AI coding assistants in the ai/ directory. Attach these files to your AI sessions to improve accuracy and keep generated code consistent with the architecture.

Directory Structure​

ai/
├── contexts/   # Codebase summaries and architectural context
├── prompts/    # Reusable prompt templates for common tasks
├── rules/      # Project-specific coding rules and conventions
└── tasks/      # Step-by-step task guides for workflows

Contexts​

  • monorepo-context.md — High-level overview: entrypoints, core packages, shared domains
  • package-template.md — Pattern for creating new packages

Attach context files when asking the AI to modify or extend the codebase. They reduce hallucinations about project structure.

Prompts​

  • add-feature.md — Template for adding a new feature
  • fix-bug.md — Template for debugging and fixes

Use these as starting points when prompting an AI assistant. Customize them for your specific task.

Rules​

  • global.md — Project-wide conventions
  • frontend.md — Frontend-specific rules (React, styling, etc.)

Rules help the AI follow existing patterns—naming, file layout, import style.

Tasks​

  • add-module-checklist.md — Checklist for adding a new domain/module
  • release-checklist.md — Pre-release verification steps

Use task files when the AI is helping with multi-step workflows.

Best Practices​

  1. Attach contexts first — Before asking about architecture or adding features, attach the relevant context file.
  2. Reference specific files — Point the AI to actual source files (e.g. packages/core/src/runtime.ts) for precise edits.
  3. Use the Architecture docs — The Architecture Overview and related docs mirror what the AI contexts describe; both stay in sync.

Complementing docs/product​

The docs/product/ directory holds human-readable planning artifacts: phased roadmap, go-to-market strategy, release policy. These are for planning; the ai/ directory is for execution—giving the AI enough context to generate correct code.