perf(ci): add a no-marker fast path to extract_model_prose

[seonghobae]

Summary

.github#1416 ("⚡ Bolt: extract_model_prose 빠른 반환 경로 추가") bundled three changes, but its branch was forked before two significant redesigns landed on main. Checked each of the three against current main before deciding what to carry over:

• extract_model_prose's fast path — still valid, applied here.
• max_tokens: REVIEW_MAX_OUTPUT_TOKENS -> 16 in the preflight request — superseded. main already has ADR-0005's own escalation design (REVIEW_PREFLIGHT_BASE_TOKENS = 16, escalating to REVIEW_PREFLIGHT_ESCALATED_TOKENS — i.e. REVIEW_MAX_OUTPUT_TOKENS — only when a response comes back truncated). Not carried over; Bolt's blunt always-16 version would have been a regression against a smarter mechanism already in place.
• curl --max-time 30 -> 60 in the sidecar preflight — superseded. A later main-side change removed the timeout from that curl call entirely, with an explicit comment citing ADR-0003 ("model inference forbids a wall-clock timeout"), replacing it with a bounded-attempt retry loop. Carrying this over would reintroduce exactly what that ADR forbids. Not carried over.

For the one idea that's still current: most model responses contain neither SENTINEL_PREFIX nor CONTROL_START, so scanning every line for both prefixes on every review is wasted work when neither is present at all. Bolt's own version of the fast path (return raw_output.strip()) wasn't quite behavior-preserving, though: str.splitlines()/"\n".join(...) normalizes CRLF and a handful of other line-separator characters that a plain .strip() leaves untouched, so a raw model response using non-LF line endings would come back subtly different from the un-optimized path. Fixed to "\n".join(raw_output.splitlines()).strip(), which is byte-for-byte identical to the slow path's output on any input, while still skipping its per-line prefix-matching loop for the common case. Added a test proving the two paths agree exactly, including on a CRLF input.

.github#1416 will be closed as fully addressed (its one still-current idea absorbed here, its other two explained as superseded) once this merges.

Test plan

• Full suite: coverage run -m pytest tests → 2772 passed, 1 skipped
• coverage report → 100%
• interrogate → 100%

🤖 Generated with Claude Code

[coderabbitai]

Warning

Review limit reached

Next included review available in 1 minute.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Team

Run ID: 8b9ccf7e-83ff-4212-ac4c-2518ce6c8ea0

📥 Commits

Reviewing files that changed from the base of the PR and between 0574df2 and 1839a53.

📒 Files selected for processing (2)

• scripts/ci/opencode_review_surfaces.py
• tests/test_opencode_review_surfaces.py

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}