feat: publish passkety-bundler

[matthiasmatt]

This pull request introduces a new GitHub Actions workflow for building and publishing the Passkey Bundler Docker image, updates the Docker base image for improved compatibility, enhances environment variable parsing, and adds new configuration-related tests. The changes improve the CI/CD pipeline, increase Docker image compatibility, and strengthen configuration validation.

CI/CD and Docker workflow improvements:

• Added a new workflow .github/workflows/passkey-bundler-docker.yml to build, push, and publish multi-architecture Docker images for Passkey Bundler, including logic to detect relevant changes and handle release tags.
• Updated .github/workflows/release.yml to ignore passkey-bundler/v* tags, preventing duplicate releases for Docker-specific tags.

Docker image compatibility:

• Changed the base image in passkey-bundler/Dockerfile from node:20-alpine to node:20-bookworm-slim for both build and runtime stages, improving compatibility with native dependencies.

Configuration handling and testing:

• Improved boolean environment variable parsing in passkey-bundler/src/config.ts to support more values (e.g., "y", "on", "n", "off") and handle whitespace/empty strings robustly.
• Added helpers and new tests in passkey-bundler/test/basic.test.ts to verify configuration loading, especially for testnet mode and authentication requirements. [1] [2]

[gemini-code-assist]

Code Review

This pull request introduces a new GitHub Actions workflow for publishing the Passkey Bundler, improves Docker image compatibility, and enhances configuration handling. The changes are well-structured and clearly described. My review includes a minor simplification for the boolean environment variable parsing and suggestions to improve test hygiene by using temporary directories for database files in the new tests, ensuring they are isolated and leave no artifacts.

[gemini-code-assist]

This test writes to a hardcoded database path (sqlite:./data/bundler.sqlite), which can cause side effects and leave artifacts after test runs. To ensure test isolation and cleanliness, it's best practice to use a temporary directory for such files and clean it up afterwards, similar to how other tests in this file are structured.

test("testnet mode defaults authRequired=true and requires API keys", () => {
  const dir = fs.mkdtempSync(path.join(os.tmpdir(), "passkey-bundler-"))
  try {
    const entryPoint = "0x" + "11".repeat(20)
    const privateKey = "0x" + "22".repeat(32)

    withBundlerEnv(
      {
        BUNDLER_MODE: "testnet",
        ENTRY_POINT: entryPoint,
        BUNDLER_PRIVATE_KEY: privateKey,
        DB_URL: `sqlite:${path.join(dir, "bundler.sqlite")}`,
      },
      () => {
        assert.throws(
          () => loadConfig(),
          /Testnet mode requires API keys \(set BUNDLER_API_KEYS\) or disable authRequired/,
        )
      },
    )
  } finally {
    fs.rmSync(dir, { recursive: true, force: true })
  }
})

[gemini-code-assist]

Similar to the previous test, this test writes to a hardcoded database path. This should be updated to use a temporary directory to ensure test isolation and prevent leaving artifacts.

test("testnet mode respects BUNDLER_REQUIRE_AUTH=false", () => {
  const dir = fs.mkdtempSync(path.join(os.tmpdir(), "passkey-bundler-"))
  try {
    const entryPoint = "0x" + "11".repeat(20)
    const privateKey = "0x" + "22".repeat(32)

    const config = withBundlerEnv(
      {
        BUNDLER_MODE: "testnet",
        ENTRY_POINT: entryPoint,
        BUNDLER_PRIVATE_KEY: privateKey,
        DB_URL: `sqlite:${path.join(dir, "bundler.sqlite")}`,
        BUNDLER_REQUIRE_AUTH: "false",
      },
      () => loadConfig(),
    )

    assert.equal(config.authRequired, false)
  } finally {
    fs.rmSync(dir, { recursive: true, force: true })
  }
})

[gemini-code-assist]

This check for an empty normalized string is redundant. If normalized is an empty string, it won't be found in either of the subsequent includes() checks, and the function will correctly return undefined at the end. Removing this line makes the function more concise without changing its behavior.