🤖 refactor: eliminate TEST_INTEGRATION env var and per-file boilerplate

[ammar-agent]

Summary

Eliminate the TEST_INTEGRATION env var and shouldRunIntegrationTests() per-file boilerplate from 72 files. Phase 3 of the integration test restructuring (Phase 1: #2370, Phase 2: #2389).

Background

Every integration test file repeated the same pattern:

import { shouldRunIntegrationTests } from "...";
const describeIntegration = shouldRunIntegrationTests() ? describe : describe.skip;

This conflated "integration test" with "needs real AI API keys" — two different concerns. Tests in tests/ are integration tests by definition; whether they need API keys should be expressed differently.

Implementation

Infrastructure (4 files)

File	Change
tests/testUtils.ts	Removed shouldRunIntegrationTests(). Simplified validateApiKeys()/getApiKey() to not depend on env var.
tests/ipc/setup.ts	createTestEnvironment({ seedDummyKeys?: boolean }) — explicit parameter replaces env var check. Default true seeds dummy keys for UI tests; setupProviders() overwrites for real-key tests.
tests/setup.ts	Removed conditional preload block. Tests needing AI SDK preloading call preloadTestModules() in their own beforeAll.
jest.config.js	Fixed 120s default timeout (replaces TEST_INTEGRATION ternary). AI tests set jest.setTimeout(600_000) individually.

Test files (65 files)

• 22 UI tests: Removed shouldRunIntegrationTests import + describeIntegration wrapper → bare describe()
• 38 IPC tests: Same removal. 26 AI-calling tests got jest.setTimeout(600_000).
• Compound-condition tests: anthropicCacheStrategy, ollama, system1BashCompaction now check only key/service availability (not TEST_INTEGRATION)
• src/node/services/streamManager.test.ts: Replaced env var check with key-availability gate (describeWithApi)
• tests/run/smoke.integration.test.ts: Same pattern removal + jest.setTimeout
• Removed stale committed tests/testUtils.js + .js.map artifacts
• Cleaned up stale // Skip all tests if TEST_INTEGRATION is not set comments and Docker skip messages

Developer documentation & CI (3 files)

File	Change
.mux/skills/tests/SKILL.md	Updated test layout docs to reflect subfolder structure from Phases 1-2. Removed TEST_INTEGRATION=1 / shouldRunIntegrationTests() guidance. Added guidance on jest.setTimeout(600_000) and validateApiKeys() for AI tests. Updated example test paths.
.github/workflows/pr.yml	Removed all TEST_INTEGRATION=1 prefixes from CI jest invocations (4 occurrences) and TEST_INTEGRATION env var from Windows smoke test job.
Makefile	Removed TEST_INTEGRATION=1 from test-integration target.

Net impact: -379 lines across 72 files

Validation

• ✅ make typecheck clean
• ✅ make lint clean
• ✅ make fmt-check clean
• ✅ 35/35 UI test suites pass without TEST_INTEGRATION=1
• ✅ Non-AI IPC tests pass (workspace/create, workspace/remove, doubleRegister, windowTitle)
• AI IPC tests require real API keys — validateApiKeys() fails fast when keys missing

Risks

• jest.config.js timeout raised from 10s → 120s for all tests. This is safe because unit tests in src/ run under bun test (not jest), and integration tests in tests/ universally need >10s for service container setup.
• tests/runtime/ files retain their own local shouldRunIntegrationTests() and TEST_INTEGRATION check — these are self-contained, Docker-gated tests with separate infrastructure. Out of scope for this change.
• src/ files that check TEST_INTEGRATION for telemetry suppression or network-calling unit test gating are unrelated concerns and left untouched.

---

Generated with mux • Model: anthropic:claude-opus-4-6 • Thinking: xhigh • Cost: $25.47

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e6d66bbfdd

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Require explicit opt-in for real-API StreamManager test

This change gates the real Anthropic test on ANTHROPIC_API_KEY alone, so it now runs during bun test src whenever that env var is present. Because make test/make test-unit executes bun test src (Makefile lines 340-343), developers or CI environments with a key exported will unexpectedly hit live API calls in what is supposed to be a unit-test path, causing slow/flaky/costly runs and potential failures from transient API issues.

Useful? React with 👍 / 👎.