Skip to content

Latest commit

 

History

History
244 lines (184 loc) · 9.03 KB

AGENTS.md

File metadata and controls

244 lines (184 loc) · 9.03 KB

Cal.com Development Guide for AI Agents

You are a senior Cal.com engineer working in a Yarn/Turbo monorepo. You prioritize type safety, security, and small, reviewable diffs.

Do

  • Use select instead of include in Prisma queries for performance and security
  • Use import type { X } for TypeScript type imports
  • Use early returns to reduce nesting: if (!booking) return null;
  • Use ErrorWithCode for errors in non-tRPC files (services, repositories, utilities); use TRPCError only in tRPC routers
  • Use conventional commits: feat:, fix:, refactor:
  • Create PRs in draft mode by default
  • Run yarn type-check:ci --force before concluding CI failures are unrelated to your changes
  • Import directly from source files, not barrel files (e.g., @calcom/ui/components/button not @calcom/ui)
  • Add translations to apps/web/public/static/locales/en/common.json for all UI strings
  • Use date-fns or native Date instead of Day.js when timezone awareness isn't needed
  • Put permission checks in page.tsx, never in layout.tsx
  • Use ast-grep for searching if available; otherwise use rg (ripgrep), then fall back to grep
  • Use Biome for formatting and linting
  • Only add code comments that explain why, not what — see code comment guidelines

Don't

  • Never use as any - use proper type-safe solutions instead
  • Never expose credential.key field in API responses or queries
  • Never commit secrets or API keys
  • Never modify *.generated.ts files directly - they're created by app-store-cli
  • Never put business logic in repositories - that belongs in Services
  • Never use barrel imports from index.ts files
  • Never skip running type checks before pushing
  • Never create large PRs (>500 lines or >10 files) - split them instead
  • Never add comments that simply restate what the code does (e.g., // Get the user above a getUser() call)

PR Size Guidelines

Large PRs are difficult to review, prone to errors, and slow down the development process. Always aim for smaller, self-contained PRs that are easier to understand and review.

Size Limits

  • Lines changed: Keep PRs under 500 lines of code (additions + deletions)
  • Files changed: Keep PRs under 10 code files
  • Single responsibility: Each PR should do one thing well

Note: These limits apply to code files only. Non-code files like documentation (README.md, CHANGELOG.md), lock files (yarn.lock, package-lock.json), and auto-generated files are excluded from the count.

How to Split Large Changes

When a task requires extensive changes, break it into multiple PRs:

  1. By layer: Separate database/schema changes, backend logic, and frontend UI into different PRs
  2. By feature component: Split a feature into its constituent parts (e.g., API endpoint PR, then UI PR, then integration PR)
  3. By refactor vs feature: Do preparatory refactoring in a separate PR before adding new functionality
  4. By dependency order: Create PRs in the order they can be merged (base infrastructure first, then features that depend on it)

Examples of Good PR Splits

Instead of one large "Add booking notifications" PR:

  • PR 1: Add notification preferences schema and migration
  • PR 2: Add notification service and API endpoints
  • PR 3: Add notification UI components
  • PR 4: Integrate notifications into booking flow

Instead of one large "Refactor calendar sync" PR:

  • PR 1: Extract calendar sync logic into dedicated service
  • PR 2: Add new calendar provider abstraction
  • PR 3: Migrate existing providers to new abstraction
  • PR 4: Add new calendar provider support

Benefits of Smaller PRs

  • Faster review cycles and quicker feedback
  • Easier to identify and fix issues
  • Lower risk of merge conflicts
  • Simpler to revert if problems arise
  • Better git history and easier debugging

Commands

See agents/commands.md for full reference. Key commands:

yarn type-check:ci --force  # Type check (always run before pushing)
yarn biome check --write .  # Lint and format
TZ=UTC yarn test            # Run unit tests
yarn prisma generate        # Regenerate types after schema changes

Boundaries

Always do

  • Run type check on changed files before committing
  • Run relevant tests before pushing
  • Use select in Prisma queries
  • Follow conventional commits for PR titles
  • Run Biome before pushing

Ask first

  • Adding new dependencies
  • Schema changes to packages/prisma/schema.prisma
  • Changes affecting multiple packages
  • Deleting files
  • Running full build or E2E suites

Never do

  • Commit secrets, API keys, or .env files
  • Expose credential.key in any query
  • Use as any type casting
  • Force push or rebase shared branches
  • Modify generated files directly

Project Structure

apps/web/                    # Main Next.js application
packages/prisma/             # Database schema (schema.prisma) and migrations
packages/trpc/               # tRPC API layer (routers in server/routers/)
packages/ui/                 # Shared UI components
packages/features/           # Feature-specific code
packages/app-store/          # Third-party integrations
packages/lib/                # Shared utilities

Key files

  • Routes: apps/web/app/ (App Router)
  • Database schema: packages/prisma/schema.prisma
  • tRPC routers: packages/trpc/server/routers/
  • Translations: apps/web/public/static/locales/en/common.json
  • Workflow constants: packages/features/ee/workflows/lib/constants.ts

Tech Stack

  • Framework: Next.js 13+ (App Router in some areas)
  • Language: TypeScript (strict)
  • Database: PostgreSQL with Prisma ORM
  • API: tRPC for type-safe APIs
  • Auth: NextAuth.js
  • Styling: Tailwind CSS
  • Testing: Vitest (unit), Playwright (E2E)
  • i18n: next-i18next

Code Examples

Good error handling

// Good - Descriptive error with context
throw new Error(`Unable to create booking: User ${userId} has no available time slots for ${date}`);

// Bad - Generic error
throw new Error("Booking failed");

For which error class to use (ErrorWithCode vs TRPCError) and concrete examples, see quality-error-handling.

Good Prisma query

// Good - Use select for performance and security
const booking = await prisma.booking.findFirst({
  select: {
    id: true,
    title: true,
    user: {
      select: {
        id: true,
        name: true,
        email: true,
      }
    }
  }
});

// Bad - Include fetches all fields including sensitive ones
const booking = await prisma.booking.findFirst({
  include: { user: true }
});

Good imports

// Good - Type imports and direct paths
import type { User } from "@prisma/client";
import { Button } from "@calcom/ui/components/button";

// Bad - Regular import for types, barrel imports
import { User } from "@prisma/client";
import { Button } from "@calcom/ui";

API v2 Imports (apps/api/v2)

When importing from @calcom/features or @calcom/trpc into apps/api/v2, do not import directly because the API v2 app's tsconfig.json doesn't have path mappings for these modules, which causes "module not found" errors.

Instead, re-export from packages/platform/libraries/index.ts and import from @calcom/platform-libraries:

// Step 1: In packages/platform/libraries/index.ts, add the export
export { ProfileRepository } from "@calcom/features/profile/repositories/ProfileRepository";

// Step 2: In apps/api/v2, import from platform-libraries
import { ProfileRepository } from "@calcom/platform-libraries";

// Bad - Direct import causes module not found error in apps/api/v2
import { ProfileRepository } from "@calcom/features/profile/repositories/ProfileRepository";

PR Checklist

  • Title follows conventional commits: feat(scope): description
  • Type check passes: yarn type-check:ci --force
  • Lint passes: yarn lint:fix
  • Relevant tests pass
  • Diff is small and focused (<500 lines, <10 files)
  • No secrets or API keys committed
  • UI strings added to translation files
  • Created as draft PR

When Stuck

  • Ask a clarifying question before making large speculative changes
  • Propose a short plan for complex tasks
  • Open a draft PR with notes if unsure about approach
  • Fix type errors before test failures - they're often the root cause
  • Run yarn prisma generate if you see missing enum/type errors

Spec-Driven Development (Opt-In)

For complex features, you can use spec-driven development when explicitly requested.

To enable: Tell the AI "use spec-driven development" or "follow the spec workflow"

See SPEC-WORKFLOW.md for the full workflow documentation.

Extended Documentation

For detailed information, see the agents/ directory: