Skip to content

Latest commit

 

History

History
50 lines (41 loc) · 3.32 KB

AGENTS.md

File metadata and controls

50 lines (41 loc) · 3.32 KB

AI Agents Configuration

Recommended instructions for AI agents working in this repository.

Project Overview

  • Stack: Node.js (most recent), Metarhia tech stack including Impress: Metacom, Metaschema, Metautil, etc.
  • Entry: server.js + static app in application/static/ (PWA with Service Worker)
  • Run: node server, static on ./application/static, ports from application/config/server.js

Codebase Layout

  • application/config/ - App configs for server, ports, cache, log, scale, sessions, etc.
  • application/domain/ - Business logic; object-literal modules, accessed as domain.<name>.<method>
  • application/schemas/ - Metarhia schema definitions (entities, types in .types.js)
  • application/static/ - Frontend: ES modules
  • application/api/ - Server-side API accessible with Metacom RPC
  • application/lib/ - Common backend libraries
  • application/test/ - Tests
  • types/global.d.ts - TypeScript declarations for tooling

Conventions (Follow Existing Patterns)

  • Follow common codebase style
    • Do not add comments, better make code self-descriptive but short and clear as well
    • Do not use jsdoc comments, we use .d.ts typings
  • Server code: application/**/*.js except application/static/*.js
    • Module format: Object-literal export: ({ method1() {}, method2() {} }) do not use 'use strict' (ESLint has strict: 'off'), sourceType: 'module'
    • Domain code (e.g. domain/chat.js). Use domain.<domain>.<method> (e.g. domain.chat.getRoom(name)). State lives in the same object (e.g. rooms: new Map())
    • Config: Pure data objects in config/*.js (host, ports, timeouts, cors, etc.)
    • Schemas: Metarhia schema syntax: 'Type', '?Type' (optional), { many: 'Type' }, { type: 'boolean', default: false }, etc. See application/schemas/ and .types.js for custom types (e.g. datetime, json, ip)
  • Frontend code: application/static/*.js
    • Module format: ES modules (import/export)
    • Worker: worker.js
  • Tests (application/**/*.test.js)
    • Format: Object with name, options, async run(t). Use t.test('description', async () => { ... }) and node.assert for assertions
  • Style, alignment, tooling
    • Editor: LF, UTF-8, final newline, max 80 chars per line, trim trailing whitespace, 2 spaces for JS/TS/JSON/YML
    • Lint/format: ESLint (eslint-config-metarhia) + Prettier. Run npm run lint / npm run fix before committing
    • Types: npm run types (TypeScript check). Use types/global.d.ts and project types for editor support.

Instructions for AI Agents

  1. Preserve architecture. Prefer existing patterns (object-literal domains, Metarhia schemas, Metacom + EventTransport in static). Don’t introduce a new framework or module system in application/ without explicit request.
  2. Match existing code style and alignment. Use the same naming/style as surrounding code.
  3. Respect boundaries. Domain logic in domain/, schema definitions in schemas/, config in config/, static assets and client/worker code in static/.
  4. Keep changes minimal and testable. Prefer small, focused edits.
  5. Don’t break the public API or config.
  6. Security and env. Don’t hardcode secrets or assume env vars that aren’t documented.
  7. Dependencies: do not add additional packages unless we directly ask you.