discourse/AI-AGENTS.md

AI Coding Agent Guide

Project-specific instructions for AI agents. MUST be loaded at conversation start.

Default Mode

• Architect mode enabled by default: detailed analysis, patterns, trade-offs, architectural guidance
• Stop and ask for context if unable to write code meeting guidelines

Development Rules

Discourse is large with long history. Understand context before changes.

All Files

• Always lint changed files
• Make display strings translatable (use placeholders, not split strings)
• Create subagent to review changes against this file after completing tasks

Toolset

• Use pnpm for JavaScript, bundle for Ruby
• Use helpers in bin over bundle exec (bin/rspec, bin/rake)

JavaScript

• No empty backing classes for template-only components unless requested
• Use FormKit for forms: https://meta.discourse.org/t/discourse-toolkit-to-render-forms/326439 (app/assets/javascripts/discourse/app/form-kit)

JSDoc

• Required for classes, methods, members (except @service members, constructors)
• Multiline format only
• Components: @component name, list params (this.args or @paramname)
• Methods: no @returns for @action, use @returns for getters (not @type)
• Members: specify @type

Testing

• Do not write unnecessary comments in tests, every single assertion doesn't need a comment
• Don't test functionality handled by other classes/components
• Don't write obvious tests
• Ruby: use fab!() over let(), system tests for UI (spec/system), use page objects for system spec finders (spec/system/page_objects)

Page Objects (System Specs)

• Located in spec/system/page_objects/pages/, inherit from PageObjects::Pages::Base
• NEVER store find() results - causes stale element references after re-renders
• Use has_x? / has_no_x? patterns for state checks (finds fresh each time)
• Action methods find+interact atomically, return self for chaining
• Don't assert immediate UI feedback after clicks (tests browser, not app logic)

Commands

# Ruby tests
bin/rspec [spec/path/file_spec.rb[:123]]
LOAD_PLUGINS=1 bin/rspec  # Plugin tests

# JavaScript tests
bin/rake qunit:test # RUN all non plugin tests
LOAD_PLUGINS=1 TARGET=all FILTER='fill filter here' bin/rake qunit:test # RUN specific tests based on filter

Exmaple filters JavaScript tests:

  emoji-test.js
    ...
    acceptance("Emoji" ..
      test("cooked correctly")
    ...
  Filter string is: "Acceptance: Emoji: cooked correctly"

  user-test.js
    ...
    module("Unit | Model | user" ..
      test("staff")
    ...
  Filter string is: "Unit | Model | user: staff"

# Linting
bin/lint path/to/file path/to/another/file
bin/lint --fix path/to/file path/to/another/file
bin/lint --fix --recent # Lint all recently changed files

ALWAYS lint any changes you make

Site Settings

• Configured in config/site_settings.yml or config/settings.yml for plugins
• Functionality in lib/site_setting_extension.rb
• Access: SiteSetting.setting_name (Ruby), siteSettings.setting_name (JS with @service siteSettings)

Services

• Extract business logic (validation, models, permissions) from controllers
• https://meta.discourse.org/t/using-service-objects-in-discourse/333641
• Examples: app/services (only classes with Service::Base)

Database & Performance

• ActiveRecord: use includes()/preload() (N+1), find_each()/in_batches() (large sets), update_all/delete_all (bulk), exists? over present?
• Migrations: rollback logic, algorithm: :concurrently for large tables, deprecate before removing columns
• Queries: use explain, specify columns, strategic indexing, counter_cache for counts

Security

• XSS: use {{}} (escaped) not {{{ }}}, sanitize with sanitize/cook, no innerHTML, careful with @html
• Auth: Guardian classes (lib/guardian.rb), POST/PUT/DELETE for state changes, CSRF tokens, protect_from_forgery
• Input: validate client+server, strong parameters, length limits, don't trust client-only validation
• Authorization: Guardian classes, route+action permissions, scope limiting, can_see?/can_edit? patterns

Knowledge Sharing

• ALWAYS persist information for ALL developers (no conversational-only memory)
• Follow project conventions, prevent knowledge silos
• Recommend storage locations by info type
• Inform when this file changes and reloads