@kiwa-test/dapp

@kiwa-test/dapp

English • 日本語

_{Full kiwa overview (127s) — this package covers the dApp e2e and manual-write paths. Full-quality MP4 (2.9 MB).}

Headless E2E test fixture for dApps on anvil forks (Playwright + viem).

@kiwa-test/dapp is the runtime fixture used by kiwa's Playwright-based dApp E2E layer. It injects window.ethereum into the test page, starts an anvil-backed wallet fixture, handles the core EIP-1193 flows directly, and forwards the rest of the JSON-RPC surface to anvil.

You can use this package in two ways: (a) let Claude generate Playwright tests via /kiwa-play, or (b) import the fixture directly (import { dappE2eTest as test } from "@kiwa-test/dapp") and write the tests by hand. See kiwa README for the full 3-path picture (contract test via @kiwa-test/forge + dApp e2e via @kiwa-test/dapp + manual write).

Installation

pnpm add -D @kiwa-test/dapp @playwright/test viem

Bonus — Claude Code plugin

If you use Claude Code, install the kiwa skill chain in one command to get /kiwa:kiwa-play (Playwright e2e generation), /kiwa:kiwa-design, /kiwa:kiwa-forge, /kiwa:kiwa-hardhat, /kiwa:kiwa-vitest, /kiwa:kiwa-api, /kiwa:kiwa-review across every dApp project. (The /kiwa:kiwa-test one-shot orchestrator requires examples/ and is kiwa-monorepo-only.)

# In Claude Code:
/plugin marketplace add cardene777/kiwa
/plugin install kiwa@kiwa-marketplace
/reload-plugins

Plugin skills are namespaced by plugin name (/kiwa:kiwa-design, not /kiwa-design). See kiwa README — Option A for the full skill list.

Quickstart

import { expect } from "@playwright/test";
import { dappE2eTest as test } from "@kiwa-test/dapp";

test("window.ethereum is injected", async ({ page, dappE2e }) => {
  await page.goto("/");

  const chainId = await page.evaluate(() => {
    return (window as any).ethereum.request({ method: "eth_chainId" });
  });

  await dappE2e.waitForRpcIdle();
  expect(chainId).toBe("0x7a69");
});

Features

• Playwright fixture preset (dappE2eTest) that injects window.ethereum
• anvil lifecycle integration for isolated test runs
• 9 EIP-1193 JSON-RPC methods handled directly, with all other methods forwarded to anvil
• 4 EIP-1193 events triggerable from tests
• transaction broadcast helper for eth_sendTransaction
• approval-mode helpers for deterministic reject-flow testing
• EIP-6963 multi-wallet injection and smart-account support
• vitest helper (setupTestEnv / withAnvil) that boots a real anvil with --load-state for unit / integration tests
• anvil pool (createAnvilPool) with anvil_reset-backed 0 ms borrow / release for parallel test suites
• tunable transport timeout / retry on sendTransaction for fail-fast error-path tests

Vitest helper (mock + real anvil with state load)

@kiwa-test/dapp ships a vitest helper that lets the same test file run in either mock mode (no anvil process, default) or real anvil mode with a pre-built state file. The state file is produced once via kiwa anvil seed, then every test boots anvil with --load-state — equivalent to pasting the entire chain in one shot.

import { setupTestEnv } from "@kiwa-test/dapp";

// mock mode (default): no anvil process, viem mock transport
const env = await setupTestEnv();

// clean anvil: spawns anvil on a free port, chainId 31337
const env = await setupTestEnv({ anvil: true });

// anvil with pre-built state: instant setup, no per-test deploy
const env = await setupTestEnv({
  anvil: { loadState: "tests/fixtures/state.json", chainId: 31337 },
});

await env.stop(); // call in afterAll

Generate the state file once with the CLI seed command:

# tests/seed.ts uses process.env.ANVIL_RPC_URL to deploy + setup
pnpm exec kiwa anvil seed tests/seed.ts --out tests/fixtures/state.json

The seed script runs against a one-shot anvil; on exit, anvil's --dump-state writes the full chain state to the output path. Subsequent test runs read it back via loadState and reach a ready chain in ~300 ms without replaying any transactions.

Anvil pool (v0.2.0+)

For test suites that need real anvil across many cases, createAnvilPool pre-spawns N instances and lets each test borrow/release one. Releasing runs anvil_reset so the next borrower sees a clean chain — borrow + release is ~0 ms after the pool warms up.

import { createAnvilPool, setupTestEnv, type AnvilPool } from "@kiwa-test/dapp";

let pool: AnvilPool;
beforeAll(async () => {
  pool = await createAnvilPool({ size: 4 });
});
afterAll(async () => {
  await pool.stopAll();
});

it("borrows from pool", async () => {
  const env = await setupTestEnv({ pool }); // borrow + lease
  // ...
  await env.stop(); // releases back, anvil_reset is invoked
});

setupTestEnv cannot accept anvil and pool together — the call throws if both are passed. Polling/spawn was tuned in v0.2.0 so a fresh setupTestEnv({ anvil: true }) returns in ~32 ms (down from 107 ms).

Transport tuning for sendTransaction (v0.2.0+)

TxBroadcastCtx accepts optional transportTimeoutMs (default 5000) and transportRetryCount (default 0). Tests that exercise transport-error paths (e.g. invalid port → ECONNREFUSED) can shrink the wait from viem's 10s default to 200 ms:

await sendTransaction(
  { privateKey, chainId: 31337, anvilPort: brokenPort, transportTimeoutMs: 200, transportRetryCount: 0 },
  params,
); // rejects with code -32603 within ~200ms

Direct RPC Methods

@kiwa-test/dapp handles these methods directly:

• eth_requestAccounts
• eth_accounts
• eth_chainId
• net_version
• personal_sign
• eth_signTypedData_v4
• wallet_switchEthereumChain
• wallet_addEthereumChain
• eth_sendTransaction

Common read methods such as eth_call, eth_getBalance, eth_blockNumber, and eth_estimateGas are forwarded to anvil.

EIP-1193 Events

The injected provider supports these events:

• accountsChanged
• chainChanged
• connect
• disconnect

From tests, you can drive them with dappE2e.triggerEvent() or use higher-level helpers such as connect(), disconnect(), switchChain(), and waitForRpcIdle().

Related

• GitHub repository
• Full documentation
• Cookbook
• Errors reference
• Events reference
• RPC reference

Author

cardene — GitHub / X

License

MIT