[knqiufan]

Summary

Closes #26

setup.sh previously treated missing Chrome/Chromium as a fatal error at step [4/7] AEC, calling exit 1 and aborting the entire installation. This blocked steps 5-7 (build, launcher, install) from running, making it impossible to install step on:

• Headless Linux servers without Chrome
• CI/Docker environments
• Any system where Chrome is not installed, even if the user only needs text mode (step exec)

Additionally, the --skip-chrome-install flag was only checked on macOS and completely ignored on Linux.

Changes

• Graceful degradation: When Chrome is not found, the AEC step now warns the user and runs step aec off instead of aborting. The setup continues normally through build, launcher, and install.
• macOS auto-install preserved: On macOS without --skip-chrome-install, the script still attempts brew install --cask google-chrome, but no longer aborts if Homebrew is missing or the install fails.
• Platform-specific guidance: Shows install instructions for the detected OS so users can enable AEC later with step aec on.
• Updated header comment: The step 4 description now reflects the new behavior ("gracefully skip otherwise").

Before / After

Scenario	Before	After
Linux, no Chrome	exit 1 at step 4 — setup aborted	Warns, runs step aec off, continues
Linux, no Chrome, --skip-chrome-install	exit 1 at step 4 (flag ignored)	Same as above (flag is macOS-only)
macOS, no Chrome, --skip-chrome-install	exit 1 at step 4	Warns, runs step aec off, continues
macOS, no Chrome, no Homebrew	exit 1 at step 4	Warns, runs step aec off, continues
Chrome present (any OS)	step aec on ✓	step aec on ✓ (unchanged)

Test plan

• Run bash scripts/setup.sh on Linux without Chrome — verify setup completes and step --version works
• Run bash scripts/setup.sh on Linux with Chrome — verify AEC is enabled
• Run bash scripts/setup.sh --skip-chrome-install on macOS without Chrome — verify setup completes with AEC disabled
• Run bash scripts/setup.sh on macOS without Chrome (with Homebrew) — verify Chrome auto-install is attempted, AEC enabled on success

[ZouR-Ma]

Thanks for the contribution!

To get this ready to merge:

• Reference an issue in the description with Closes #<number> (our link-check CI requires it; open one first if needed).
• Make sure pnpm check passes locally (lint, type-check, dependency/dead-code guards, formatting, and tests — also enforced in CI).

Full conventions:
https://github.com/stepfun-ai/Step-Realtime-CLI/blob/main/CONTRIBUTING.md

[knqiufan]

Thanks @ZouR-Ma for the review guidance!

• Updated PR description to use Closes setup.sh fails on Linux without Chrome even with --skip-chrome-install; AEC should be optional #26.
• Rebased onto latest main (ee6862f).
• Ran checks locally: lint, dep-guard, deadcode, tsc, format:check all pass.
• Full pnpm check on Windows previously had 13 unrelated test failures (symlink EPERM + shell timeout). Tracked in [chore] align utils tests with TESTING.md for Windows local pnpm check #74 and addressed in test(utils): align path and shell tests with TESTING.md on Windows #75 — local pnpm check is now green on Windows after that fix.

Ready for review when you have a moment.

[ZouR-Ma]

Thanks for this PR — the problem is well-identified: headless servers and CI environments used to die at step [4/7] with exit 1, blocking installation even for text-mode-only users, and the fix also covers --skip-chrome-install being ignored on Linux. We agree with the approach: the README already documents that voice falls back to system audio drivers when AEC is off and that AEC isn't needed with headphones, so "warn + step aec off + continue" brings the installer in line with the product's existing degradation design — and it degrades loudly (platform-specific install guidance plus a pointer to step aec on), not silently. One thing must be fixed before this lands:

1. (Must) The script still aborts when the brew install fails. scripts/setup.sh runs under set -euo pipefail, so the bare brew install --cask google-chrome call in the macOS branch kills the whole script on any failure (network hiccup, cask error, cancelled authorization) — which contradicts the "no longer aborts if … the install fails" promise in the PR description. A one-line fix does it: brew install --cask google-chrome || warn "Chrome install failed — continuing without AEC"; on failure AEC_AVAILABLE stays 0 and the flow lands on the aec-off branch as intended. Acceptance criterion: simulate a brew failure on macOS and confirm the script completes all 7 steps ending with AEC off.

Also, all four checkboxes in the test plan are still unchecked — setup.sh isn't covered by CI, so this change rides entirely on manual verification. Please run at least the headline "Lin

[knqiufan]

Fixed in 1aacd4a.

• brew install --cask google-chrome now handles a non-zero exit with a warning, so under set -euo pipefail the installer continues with AEC_AVAILABLE=0 and executes step aec off.
• Added scripts/setup.test.ts. On Bash-capable platforms it simulates Darwin with a failing brew, then asserts the flow reaches step [7/7] with AEC skipped and records step aec off.

Validation:

• pnpm test -- scripts/setup.test.ts passes on Windows (the Bash simulation is skipped because this machine has neither Bash nor WSL).
• lint, dependency guard, dead-code, TypeScript, and formatting checks pass.
• Full pnpm check remains blocked locally by 10 existing Windows symlink tests that fail with EPERM; the updated installer test passes within that run.

[ZouR-Ma]

Re-review passed — this is ready to land. The brew guard is exactly right: on failure AEC_AVAILABLE stays 0 and the flow lands on the aec-off branch, no longer fighting set -euo pipefail. What stands out is scripts/setup.test.ts: this repo previously had zero automated coverage for setup.sh, and now there's an end-to-end regression test (a shimmed brew that force-fails, asserting the script completes all 7 steps and runs aec off) that permanently locks down the exact scenario from the last round. The file lands inside vitest's existing include pattern and the bash-dependent case is correctly skipped on Windows — well thought through.

One small wording issue we'll take care of directly on this branch (maintainer edit) so it doesn't cost you another round: after a failed brew install the script still re-runs Chrome detection, and that branch prints "Chrome install succeeded but binary not detected" — misleading right after brew reported a failure. We'll move the re-detection inside the success branch and adjust the test assertion accordingly.

Thanks for the fast turnaround across both rounds — merging once CI reruns on that touch-up.

[ZouR-Ma]

Re-review passed: the brew guard fix is verified, the new setup.test.ts end-to-end regression locks down the failure scenario, and the wording touch-up landed on this branch. All required checks are green on the merged state. Approving.