run-tests

name	run-tests
description	Run the opencode-litellm test suite with bun. Use when checking if tests pass, running unit tests, running integration tests, debugging a failing test, or verifying a change hasn't broken anything. Supports filtering by suite name.
argument-hint	[unit|integration|<filter>]
disable-model-invocation	true
allowed-tools	Bash, Read

Run Tests

Argument Parsing

Parse $ARGUMENTS:

• empty → run all unit + integration tests (bun test)
• unit → run unit tests only (bun test tests/plugin.test.ts)
• integration → run integration tests only (bun test tests/integration.test.ts)
• anything else → pass as a filter pattern (bun test --testNamePattern "$ARGUMENTS")

Test Structure

File	Type	What it tests
tests/plugin.test.ts	Unit	Plugin structure, auth hook (multi-prompt authorize), config hook (placeholder write + model fetch)
tests/integration.test.ts	Integration	Full round-trip against a real in-process mock LiteLLM HTTP server
tests/fixtures/litellm-server.ts	Fixture	startMockLiteLLM({ apiKey?, models? }) — Bun HTTP server on port: 0

Steps

1. Ensure dependencies are installed:

   bun install
   

2. Run tests based on argument:

   All unit + integration tests:

   bun test
   

   Unit only:

   bun test tests/plugin.test.ts
   

   Integration only:

   bun test tests/integration.test.ts
   

   By name filter:

   bun test --testNamePattern "$ARGUMENTS"
   

3. Report results. If tests fail, read the failing test file and src/index.ts to diagnose before suggesting changes.

Key Testing Patterns

Mocking fetch (unit tests):

function mockFetchSuccess() {
  const original = global.fetch
  global.fetch = mock(async () => ({
    ok: true,
    json: async () => ({ data: MOCK_MODEL_IDS.map((id) => ({ id })) }),
  })) as any
  return () => { global.fetch = original }
}
// restore in afterEach: restoreFetch?.()

Integration mock server:

const server = startMockLiteLLM({ apiKey?: string, models?: string[] })
// server.url  = "http://localhost:{os-assigned-port}"
// server.port = number
// server.chatRequests = [{ model: string, messageCount: number }, ...]
// server.stop() in afterAll

Config file cleanup: beforeEach/afterEach call clearPluginConfig() which wipes provider.litellm from ~/.config/opencode/opencode.json and the litellm credential from ~/.local/share/opencode/auth.json.

Error Handling

• If bun install fails: check package.json and network connectivity
• If unit tests fail with fetch errors: check that mockFetchSuccess() is being called and restored correctly
• If integration tests fail with ECONNREFUSED: the mock server likely stopped early — check beforeAll/afterAll ordering