# packages/stack — watch mode
cd packages/stack && pnpm test

# packages/cli — run once
cd packages/cli && pnpm test

Pattern: testing getters/mutations with the memory adapter

Use createMemoryAdapter + defineDb to spin up an isolated in-memory DB per test:

import { describe, it, expect, beforeEach } from "vitest"
import { createMemoryAdapter } from "@btst/adapter-memory"
import { defineDb } from "@btst/db"
import type { DBAdapter as Adapter } from "@btst/db"
import { myPluginSchema } from "../db"
import { listItems, getItemById } from "../api/getters"

const createTestAdapter = (): Adapter => {
  const db = defineDb({}).use(myPluginSchema)
  return createMemoryAdapter(db)({})
}

describe("my-plugin getters", () => {
  let adapter: Adapter

  beforeEach(() => {
    adapter = createTestAdapter()   // fresh DB per test
  })

  it("returns empty result when no items exist", async () => {
    const result = await listItems(adapter)
    expect(result.items).toEqual([])
    expect(result.total).toBe(0)
  })
})

Use vi.mock for external modules that don't exist in the test environment (e.g. @vercel/blob/server).

E2E tests (Playwright)

E2E setup

Playwright configuration lives in e2e/playwright.codegen.config.ts, command pnpm codegen:e2e. It starts the codegen-projects/ apps (ports 3006–3008) and tests the CLI codegen output — accurately reflecting what a real btst init user gets.

The codegen projects are built from scratch via scripts/codegen/setup-*.sh.

Location and naming

Tests live in e2e/tests/. Naming convention: smoke.{feature}.spec.ts

Examples: smoke.chat.spec.ts, smoke.blog.spec.ts

Setting up codegen projects (required before running codegen E2E)

The codegen-projects/ directory is not committed — build it from scratch:

# Set up a specific framework
bash scripts/codegen/setup-nextjs.sh
bash scripts/codegen/setup-react-router.sh
bash scripts/codegen/setup-tanstack.sh

# Or set up all frameworks at once
bash scripts/codegen/setup.sh

# Cleanup when you want a fresh start
bash scripts/codegen/cleanup.sh nextjs
bash scripts/codegen/cleanup.sh react-router
bash scripts/codegen/cleanup.sh tanstack

See scripts/codegen/README.md for full details on the E2E overlay file system and update workflow.

Run commands

cd e2e

# Run all codegen E2E tests
pnpm codegen:e2e

# Target a specific framework
pnpm codegen:e2e:nextjs
pnpm codegen:e2e:tanstack
pnpm codegen:e2e:react-router

# Specific test file
pnpm codegen:e2e:nextjs -- tests/smoke.blog.spec.ts

Playwright projects and ports

Project	Port
`nextjs:codegen`	3006
`tanstack:codegen`	3007
`react-router:codegen`	3008

Set BTST_FRAMEWORK=nextjs|tanstack|react-router to start only one server.

In-memory state between test runs

The in-memory adapter keeps data alive for the lifetime of the server process. If you re-run tests against an already-running server, earlier test data will still be present and can cause failures.

To reset: kill the server and let reuseExistingServer: true restart it, or rebuild with start:e2e:

kill $(lsof -ti:3006)
pnpm -F nextjs run start:e2e   # rebuilds .next and starts fresh

API key guard pattern

Features requiring external APIs (OpenAI, etc.) must skip gracefully when the key is absent:

test.beforeEach(async () => {
  if (!process.env.OPENAI_API_KEY) {
    test.skip()
  }
})

Check for the API key in test.beforeEach
Call test.skip() — don't throw
Document required env vars in a comment at the top of the spec file

Environment variables

# Env is loaded from codegen-projects/nextjs/.env automatically by playwright.codegen.config.ts

For CI, the codegen workflow (.github/workflows/codegen-e2e.yml) builds the codegen project from scratch and passes OPENAI_API_KEY from secrets.

btst-testing

BTST Testing

Unit tests (Vitest)

Location and naming

Run commands