# CLAUDE.md This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository. ## Common Development Commands ### Development Server - `bin/dev` - Start development server (Rails, Sidekiq, Tailwind CSS watcher) - `bin/rails server` - Start Rails server only - `bin/rails console` - Open Rails console ### Testing - `bin/rails test` - Run all tests - `bin/rails test:db` - Run tests with database reset - `bin/rails test:system` - Run system tests only (use sparingly - they take longer) - `bin/rails test test/models/account_test.rb` - Run specific test file - `bin/rails test test/models/account_test.rb:42` - Run specific test at line ### Linting & Formatting - `bin/rubocop` - Run Ruby linter - `npm run lint` - Check JavaScript/TypeScript code - `npm run lint:fix` - Fix JavaScript/TypeScript issues - `npm run format` - Format JavaScript/TypeScript code - `bin/brakeman` - Run security analysis ### Database - `bin/rails db:prepare` - Create and migrate database - `bin/rails db:migrate` - Run pending migrations - `bin/rails db:rollback` - Rollback last migration - `bin/rails db:seed` - Load seed data ### Setup - `bin/setup` - Initial project setup (installs dependencies, prepares database) ## Pre-Pull Request CI Workflow ALWAYS run these commands before opening a pull request: 1. **Tests** (Required): - `bin/rails test` - Run all tests (always required) - `bin/rails test:system` - Run system tests (only when applicable, they take longer) 2. **Linting** (Required): - `bin/rubocop -f github -a` - Ruby linting with auto-correct - `bundle exec erb_lint ./app/**/*.erb -a` - ERB linting with auto-correct 3. **Security** (Required): - `bin/brakeman --no-pager` - Security analysis Only proceed with pull request creation if ALL checks pass. ## General Development Rules ### Authentication Context - Use `Current.user` for the current user. Do NOT use `current_user`. - Use `Current.family` for the current family. Do NOT use `current_family`. ### Development Guidelines - Carefully read project conventions and guidelines before generating any code. - Do not run `rails server` in your responses - Do not run `touch tmp/restart.txt` - Do not run `rails credentials` - Do not automatically run migrations ## High-Level Architecture ### Application Modes The codebase runs in two distinct modes: - **Managed**: A team operates and manages servers for users (Rails.application.config.app_mode = "managed") - **Self Hosted**: Users host the codebase on their own infrastructure, typically through Docker Compose (Rails.application.config.app_mode = "self_hosted") ### Core Domain Model The application is built around financial data management with these key relationships: - **User** → has many **Accounts** → has many **Transactions** - **Account** types: checking, savings, credit cards, investments, crypto, loans, properties - **Transaction** → belongs to **Category**, can have **Tags** and **Rules** - **Investment accounts** → have **Holdings** → track **Securities** via **Trades** ### API Architecture The application provides both internal and external APIs: - Internal API: Controllers serve JSON via Turbo for SPA-like interactions - External API: `/api/v1/` namespace with Doorkeeper OAuth and API key authentication - API responses use Jbuilder templates for JSON rendering - Rate limiting via Rack Attack with configurable limits per API key - **OpenAPI Documentation**: All API endpoints MUST have corresponding OpenAPI specs in `spec/requests/api/` using rswag. See `docs/api/openapi.yaml` for the generated documentation. ### Sync & Import System Two primary data ingestion methods: 1. **Plaid Integration**: Real-time bank account syncing - `PlaidItem` manages connections - `Sync` tracks sync operations - Background jobs handle data updates 2. **CSV Import**: Manual data import with mapping - `Import` manages import sessions - Supports transaction and balance imports - Custom field mapping with transformation rules ### Provider Integrations: Pending Transactions and FX (SimpleFIN/Plaid) - Detection - SimpleFIN: pending via `pending: true` or `posted` blank/0 + `transacted_at`. - Plaid: pending via Plaid `pending: true` (stored at `extra["plaid"]["pending"]` for bank/credit transactions imported via `PlaidEntry::Processor`). - Storage: provider data on `Transaction#extra` (e.g., `extra["simplefin"]["pending"]`; FX uses `fx_from`, `fx_date`). - UI: "Pending" badge when `transaction.pending?` is true; no badge if provider omits pendings. - Configuration (default-on for pending) - SimpleFIN: `config/initializers/simplefin.rb` via `Rails.configuration.x.simplefin.*`. - Plaid: `config/initializers/plaid_config.rb` via `Rails.configuration.x.plaid.*`. - Pending transactions are fetched by default and handled via reconciliation/filtering. - Set `SIMPLEFIN_INCLUDE_PENDING=0` to disable pending fetching for SimpleFIN. - Set `PLAID_INCLUDE_PENDING=0` to disable pending fetching for Plaid. - Set `SIMPLEFIN_DEBUG_RAW=1` to enable raw payload debug logging. Provider support notes: - SimpleFIN: supports pending + FX metadata (stored under `extra["simplefin"]`). - Plaid: supports pending when the upstream Plaid payload includes `pending: true` (stored under `extra["plaid"]`). - Plaid investments: investment transactions currently do not store pending metadata. - Lunchflow: does not currently store pending metadata. ### Background Processing Sidekiq handles asynchronous tasks: - Account syncing (`SyncJob`) - Import processing (`ImportJob`) - AI chat responses (`AssistantResponseJob`) - Scheduled maintenance via sidekiq-cron ### Frontend Architecture - **Hotwire Stack**: Turbo + Stimulus for reactive UI without heavy JavaScript - **ViewComponents**: Reusable UI components in `app/components/` - **Stimulus Controllers**: Handle interactivity, organized alongside components - **Charts**: D3.js for financial visualizations (time series, donut, sankey) - **Styling**: Tailwind CSS v4.x with custom design system - Design system defined in `app/assets/tailwind/maybe-design-system.css` - Always use functional tokens (e.g., `text-primary` not `text-white`) - Prefer semantic HTML elements over JS components - Use `icon` helper for icons, never `lucide_icon` directly - **i18n**: All user-facing strings must use localization (i18n). Update locale files for each new or changed element. ### Internationalization (i18n) Guidelines - **Key Organization**: Use hierarchical keys by feature: `accounts.index.title`, `transactions.form.amount_label` - **Translation Helper**: Always use `t()` helper for user-facing strings - **Interpolation**: Use for dynamic content: `t("users.greeting", name: user.name)` - **Pluralization**: Use Rails pluralization: `t("transactions.count", count: @transactions.count)` - **Locale Files**: Update `config/locales/en.yml` for new strings - **Missing Translations**: Configure to raise errors in development for missing keys ### Multi-Currency Support - All monetary values stored in base currency (user's primary currency) - `Money` objects handle currency conversion and formatting - Historical exchange rates for accurate reporting ### Security & Authentication - Session-based auth for web users - API authentication via: - OAuth2 (Doorkeeper) for third-party apps - API keys with JWT tokens for direct API access - Scoped permissions system for API access - Strong parameters and CSRF protection throughout ### Testing Philosophy - Comprehensive test coverage using Rails' built-in Minitest - Fixtures for test data (avoid FactoryBot) - Keep fixtures minimal (2-3 per model for base cases) - VCR for external API testing - System tests for critical user flows (use sparingly) - Test helpers in `test/support/` for common scenarios - Only test critical code paths that significantly increase confidence - Write tests as you go, when required - **API Endpoints require OpenAPI specs** in `spec/requests/api/` for documentation purposes ONLY, not test (uses RSpec + rswag) ### Performance Considerations - Database queries optimized with proper indexes - N+1 queries prevented via includes/joins - Background jobs for heavy operations - Caching strategies for expensive calculations - Turbo Frames for partial page updates ### Development Workflow - Feature branches merged to `main` - Docker support for consistent environments - Environment variables via `.env` files - Lookbook for component development (`/lookbook`) - Letter Opener for email preview in development ## Project Conventions ### Convention 1: Minimize Dependencies - Push Rails to its limits before adding new dependencies - Strong technical/business reason required for new dependencies - Favor old and reliable over new and flashy ### Convention 2: Skinny Controllers, Fat Models - Business logic in `app/models/` folder, avoid `app/services/` - Use Rails concerns and POROs for organization - Models should answer questions about themselves: `account.balance_series` not `AccountSeries.new(account).call` ### Convention 3: Hotwire-First Frontend - **Native HTML preferred over JS components** - Use `` for modals, `
` for disclosures - **Leverage Turbo frames** for page sections over client-side solutions - **Query params for state** over localStorage/sessions - **Server-side formatting** for currencies, numbers, dates - **Always use `icon` helper** in `application_helper.rb`, NEVER `lucide_icon` directly ### Convention 4: Optimize for Simplicity - Prioritize good OOP domain design over performance - Focus performance only on critical/global areas (avoid N+1 queries, mindful of global layouts) ### Convention 5: Database vs ActiveRecord Validations - Simple validations (null checks, unique indexes) in DB - ActiveRecord validations for convenience in forms (prefer client-side when possible) - Complex validations and business logic in ActiveRecord ## TailwindCSS Design System ### Design System Rules - **Always reference `app/assets/tailwind/maybe-design-system.css`** for primitives and tokens - **Use functional tokens** defined in design system: - `text-primary` instead of `text-white` - `bg-container` instead of `bg-white` - `border border-primary` instead of `border border-gray-200` - **NEVER create new styles** in design system files without permission - **Always generate semantic HTML** ## Component Architecture ### ViewComponent vs Partials Decision Making **Use ViewComponents when:** - Element has complex logic or styling patterns - Element will be reused across multiple views/contexts - Element needs structured styling with variants/sizes - Element requires interactive behavior or Stimulus controllers - Element has configurable slots or complex APIs - Element needs accessibility features or ARIA support **Use Partials when:** - Element is primarily static HTML with minimal logic - Element is used in only one or few specific contexts - Element is simple template content - Element doesn't need variants, sizes, or complex configuration - Element is more about content organization than reusable functionality **Component Guidelines:** - Prefer components over partials when available - Keep domain logic OUT of view templates - Logic belongs in component files, not template files ### Stimulus Controller Guidelines **Declarative Actions (Required):** ```erb
``` **Example locale file structure (config/locales/en.yml):** ```yaml en: components: transaction_details: show_details: "Show Details" hide_details: "Hide Details" amount_label: "Amount" date_label: "Date" category_label: "Category" ``` **i18n Best Practices:** - Organize keys by feature/component: `components.transaction_details.show_details` - Use descriptive key names that indicate purpose: `show_details` not `button` - Group related translations together in the same namespace - Use interpolation for dynamic content: `t("users.welcome", name: user.name)` - Always update locale files when adding new user-facing strings **Controller Best Practices:** - Keep controllers lightweight and simple (< 7 targets) - Use private methods and expose clear public API - Single responsibility or highly related responsibilities - Component controllers stay in component directory, global controllers in `app/javascript/controllers/` - Pass data via `data-*-value` attributes, not inline JavaScript ## Testing Philosophy ### General Testing Rules - **ALWAYS use Minitest + fixtures** (NEVER RSpec or factories) - Keep fixtures minimal (2-3 per model for base cases) - Create edge cases on-the-fly within test context - Use Rails helpers for large fixture creation needs ### Test Quality Guidelines - **Write minimal, effective tests** - system tests sparingly - **Only test critical and important code paths** - **Test boundaries correctly:** - Commands: test they were called with correct params - Queries: test output - Don't test implementation details of other classes ### Testing Examples ```ruby # GOOD - Testing critical domain business logic test "syncs balances" do Holding::Syncer.any_instance.expects(:sync_holdings).returns([]).once assert_difference "@account.balances.count", 2 do Balance::Syncer.new(@account, strategy: :forward).sync_balances end end # BAD - Testing ActiveRecord functionality test "saves balance" do balance_record = Balance.new(balance: 100, currency: "USD") assert balance_record.save end ``` ### Stubs and Mocks - Use `mocha` gem - Prefer `OpenStruct` for mock instances - Only mock what's necessary ## API Development Guidelines ### OpenAPI Documentation (MANDATORY) When adding or modifying API endpoints in `app/controllers/api/v1/`, you **MUST** create or update corresponding OpenAPI request specs: 1. **Location**: `spec/requests/api/v1/{resource}_spec.rb` 2. **Framework**: RSpec with rswag for OpenAPI generation 3. **Schemas**: Define reusable schemas in `spec/swagger_helper.rb` 4. **Generated Docs**: `docs/api/openapi.yaml` **Example structure for a new API endpoint:** ```ruby # spec/requests/api/v1/widgets_spec.rb require 'swagger_helper' RSpec.describe 'API V1 Widgets', type: :request do path '/api/v1/widgets' do get 'List widgets' do tags 'Widgets' security [ { apiKeyAuth: [] } ] produces 'application/json' response '200', 'widgets listed' do schema '$ref' => '#/components/schemas/WidgetCollection' run_test! end end end end ``` **Regenerate OpenAPI docs after changes:** ```bash RAILS_ENV=test bundle exec rake rswag:specs:swaggerize ``` ### Post-commit API consistency (issue #944) After every API endpoint commit, ensure: 1. **Minitest behavioral coverage** — Add or update tests in `test/controllers/api/v1/{resource}_controller_test.rb`. Use API key and `api_headers` (X-Api-Key). Cover index/show, CRUD where relevant, 401/403/422/404. Do not rely on rswag for behavioral assertions. 2. **rswag docs-only** — Do not add `expect(...)` or `assert_*` in `spec/requests/api/v1/`. Use `run_test!` only so specs document request/response and regenerate `docs/api/openapi.yaml`. 3. **Same API key auth in rswag** — Every request spec in `spec/requests/api/v1/` must use the same API key pattern (`ApiKey.generate_secure_key`, `ApiKey.create!(...)`, `let(:'X-Api-Key') { api_key.plain_key }`). Do not use Doorkeeper/OAuth in those specs so generated docs stay consistent. Full checklist and pattern: [.cursor/rules/api-endpoint-consistency.mdc](.cursor/rules/api-endpoint-consistency.mdc). To verify the implementation: `ruby test/support/verify_api_endpoint_consistency.rb`. To scan the current APIs for violations: `ruby test/support/verify_api_endpoint_consistency.rb --compliance`.