feat: make QuickJS the default plugin engine everywhere

[steipete]

Summary

• make QuickJS the default provider-plugin engine on every platform
• retain JavaScriptCore as an Apple-only rollback engine and in A/B CI coverage
• isolate the portable TypeScript transpiler and preserve true in-engine interruption, heap, and stack limits
• add committed engine benchmark instrumentation and document the decision

Benchmark

August 8, 2026 baseline from a Swift debug build on an Apple M3 Ultra. Fetch values are 50 fixture-backed iterations reusing one context; memory is the rough macOS physical-footprint delta per retained context.

Engine	Poe (50)	OpenRouter (50)	Crof (50)	Rough peak delta/context
JavaScriptCore	321.237 ms	36.424 ms	23.917 ms	806.4 KiB
QuickJS	1000.066 ms	90.610 ms	38.119 ms	117.3 KiB

QuickJS is slower in relative terms, but the absolute fetch deltas remain below 10 ms per fetch at minutes-cadence polling. That trade is preferable to JavaScriptCore's roughly 7x larger measured context footprint, especially because QuickJS also provides true in-engine interrupt safety instead of abandoning an uninterruptible evaluation thread after timeout.

Rollback

On Apple platforms, set CODEXBAR_PLUGIN_ENGINE=jsc or enable the JavaScriptCore rollback in Settings → Debug → Provider Plugins, then restart CodexBar. An explicit environment selection overrides the persisted Debug setting.

Proof

• full suite: 822/822 under the QuickJS default
• full suite: 822/822 under the JavaScriptCore default
• focused A/B engine runs: 64/64 for QuickJS and 64/64 for JavaScriptCore
• post-rebase local gate: make check, make build, full make test, and swift build --target CodexBarLinuxTests
• JavaScriptCore remains retained in A/B CI to detect drift
• autoreview clean
• coordinator live proof: no-flag Poe fetch succeeded with the new QuickJS default

🤖 Generated with Claude Code

[clawsweeper]

Codex review: found issues before merge. Reviewed August 8, 2026, 5:13 PM ET / 21:13 UTC.

ClawSweeper review

What this changes

Changes provider plugins to use QuickJS by default, retains JavaScriptCore as an Apple rollback, and adds safety, test, CI, settings, benchmark, and documentation updates.

Merge readiness

⚠️ Needs maintainer review before merge - 4 items remain

This owner-authored PR remains useful but is not merge-ready: it contains unresolved changelog conflict markers and still documents a 2 MiB QuickJS limit while the implementation uses 1 MiB.

Priority: P2
Reviewed head: 0f26b232a8b7a43346ee8fed29ccee394d7f2aca
Owner decision: Required. See Decision needed.

Review scores

Measure	Result	What it means
Overall readiness	🦐 gold shrimp (3/6)	Real-behavior evidence and coverage are useful, but concrete merge and documentation defects remain.
Proof confidence	🐚 platinum hermit (4/6)	Sufficient (live_output): The PR body reports an after-fix no-flag Poe fetch under the QuickJS default, supplemented by QuickJS and JavaScriptCore A/B suites.
Patch quality	🦐 gold shrimp (3/6)	2 actionable review findings remain.

Verification

Check	Result	Evidence
Real behavior	Verified	Sufficient (live_output): The PR body reports an after-fix no-flag Poe fetch under the QuickJS default, supplemented by QuickJS and JavaScriptCore A/B suites.
Evidence reviewed	3 items	Unresolved merge content: The submitted changelog hunk contains conflict markers, leaving competing release-note text and matching GitHub's dirty merge state. Stack-limit history: The preceding QuickJS safety commit describes a 4 MiB native stack and 1 MiB JavaScript limit, while the changed documentation says 2 MiB. Area provenance: QuickJS portability and the JavaScript provider cutover are established main-branch work in this runtime area.
Findings	2 actionable findings	[P2] Resolve the committed changelog conflict [P3] Document the 1 MiB QuickJS stack limit
Security	None	None.

How this fits together

Provider plugins execute JavaScript to collect provider usage data. Engine selection chooses QuickJS or the Apple rollback engine before limits and watchdogs govern execution and results reach CodexBar.

flowchart LR
  A[Provider plugin source] --> B[Engine selection]
  B --> C[QuickJS default]
  B --> D[JavaScriptCore rollback]
  C --> E[Watchdog and limits]
  D --> E
  E --> F[Usage fetch result]
  F --> G[CodexBar display]

Loading

Decision needed

Question	Recommendation
Should the QuickJS default land after the conflict and documentation mismatch are repaired, with JavaScriptCore retained as the compatibility rollback?	Land with rollback safeguards: Repair the two defects, then merge with the Apple rollback and A/B coverage retained.

Why: The runtime-default change is an intentional compatibility tradeoff that automated coverage cannot fully settle for third-party plugins.

Before merge

• Resolve the committed changelog conflict (P2) - CHANGELOG.md contains <<<<<<<, =======, and >>>>>>> in the PR diff. This leaves competing release-note entries and prevents a clean merge; retain one accurate entry before rebasing.
• Document the 1 MiB QuickJS stack limit (P3) - The implementation sets a 1 MiB JavaScript stack limit, but the changed plugin documentation says 2 MiB. Correct the public limit so plugin authors can reason about recursion failures.
• Resolve merge risk (P1) - Existing provider plugins may behave differently after the engine default changes; the Apple-only JavaScriptCore rollback mitigates but does not eliminate that upgrade risk.

Findings

• [P2] Resolve the committed changelog conflict — CHANGELOG.md:6-12
• [P3] Document the 1 MiB QuickJS stack limit — docs/plugin-prototype.md:22-23

Agent review details

Security

None.

Review metrics

Metric	Value	Why it matters
Changed surface	35 files, +787/-155 lines	The runtime default changes across execution, settings, CI, tests, documentation, and localization.

Merge-risk options

Maintainer options:

1. Repair and revalidate before merge (recommended)
   Resolve the conflicted changelog hunk, correct the stack-limit documentation, rebase, and rerun both engine paths.
2. Defer the default switch
   Keep JavaScriptCore as the Apple default if maintainers want additional compatibility observation.

Copy recommended automerge instruction

@clawsweeper automerge

Special instructions:
Resolve the CHANGELOG.md conflict to one accurate entry, document the 1 MiB QuickJS JavaScript stack limit, rebase onto main, and rerun both engine paths.

Technical review

Best possible solution:

Resolve the exceptional release-note conflict, correct the documented 1 MiB QuickJS limit, rebase, and retain the JavaScriptCore rollback plus A/B coverage when deciding whether to land the default switch.

Do we have a high-confidence way to reproduce the issue?

Not applicable as a bug report; the PR intentionally changes a runtime default and reports a live no-flag Poe fetch plus A/B engine coverage.

Is this the best way to solve the issue?

Unclear until the merge conflict and documentation mismatch are corrected; retaining the rollback is the narrowest compatibility safeguard for the proposed default.

Full review comments:

• [P2] Resolve the committed changelog conflict — CHANGELOG.md:6-12
CHANGELOG.md contains <<<<<<<, =======, and >>>>>>> in the PR diff. This leaves competing release-note entries and prevents a clean merge; retain one accurate entry before rebasing.
  Confidence: 0.99
• [P3] Document the 1 MiB QuickJS stack limit — docs/plugin-prototype.md:22-23
  The implementation sets a 1 MiB JavaScript stack limit, but the changed plugin documentation says 2 MiB. Correct the public limit so plugin authors can reason about recursion failures.
  Confidence: 0.98

Overall correctness: patch is incorrect
Overall confidence: 0.97

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against af9bc9382f4c.

Labels

Label changes:

• add proof: sufficient: Contributor real behavior proof is sufficient. The PR body reports an after-fix no-flag Poe fetch under the QuickJS default, supplemented by QuickJS and JavaScriptCore A/B suites.

Label justifications:

• P2: This broad runtime-default change has normal-priority compatibility impact without an emergency regression.
• merge-risk: 🚨 compatibility: Existing provider plugins change JavaScript engines on upgrade, with JavaScriptCore available only as an Apple rollback.
• rating: 🦐 gold shrimp: Overall readiness is 🦐 gold shrimp; proof is 🐚 platinum hermit and patch quality is 🦐 gold shrimp.
• status: ⏳ waiting on author: ClawSweeper has contributor-facing work open and is waiting for author action. Sufficient (live_output): The PR body reports an after-fix no-flag Poe fetch under the QuickJS default, supplemented by QuickJS and JavaScriptCore A/B suites.
• proof: sufficient: Contributor real behavior proof is sufficient. The PR body reports an after-fix no-flag Poe fetch under the QuickJS default, supplemented by QuickJS and JavaScriptCore A/B suites.

Evidence

What I checked:

• Unresolved merge content: The submitted changelog hunk contains conflict markers, leaving competing release-note text and matching GitHub's dirty merge state. (CHANGELOG.md:6, 0f26b232a8b7)
• Stack-limit history: The preceding QuickJS safety commit describes a 4 MiB native stack and 1 MiB JavaScript limit, while the changed documentation says 2 MiB. (Sources/CodexBarCore/Plugins/QuickJSProviderPluginEngine.swift, 064b22d93725)
• Area provenance: QuickJS portability and the JavaScript provider cutover are established main-branch work in this runtime area. (Sources/CodexBarCore/Plugins/ProviderPluginRuntime.swift, cee0fd07)

Likely related people:

• steipete: Authored the QuickJS portability and provider cutover work as well as all commits in this PR. (role: recent area contributor; confidence: high; commits: cee0fd07, 8fc67d7f, 8bf4041fbebc; files: Sources/CodexBarCore/Plugins/ProviderPluginRuntime.swift, Sources/CodexBarCore/Plugins/QuickJSProviderPluginEngine.swift)

Rank-up moves

Optional improvements that raise the rating; they are not merge blockers.

• Resolve the committed changelog conflict.
• Correct the public QuickJS stack-limit value and refresh validation after rebasing.

Rating scale

Score	Internal tier	Crab rank	Meaning
6/6	S	🦀 challenger crab	Exceptional readiness
5/6	A	🦞 diamond lobster	Very strong readiness
4/6	B	🐚 platinum hermit	Good normal PR; ordinary maintainer review
3/6	C	🦐 gold shrimp	Useful, but confidence is limited
2/6	D	🦪 silver shellfish	Proof or implementation needs work
1/6	F	🧂 unranked krab	Not merge-ready
N/A	NA	🌊 off-meta tidepool	Rating does not apply

Overall follows the weaker of proof and patch quality.
Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

Workflow

• ClawSweeper keeps one durable marker-backed review comment per issue or PR.
• Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
• A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
• PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
• Maintainers can also comment @clawsweeper review to request a fresh review only.
• Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
• Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
• Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

History

Review history (6 earlier review cycles)

• reviewed 2026-08-08T17:50:34.145Z sha fc23e69 :: needs changes before merge. :: [P1] Expose the result box to the extracted transpiler
• reviewed 2026-08-08T18:22:38.138Z sha fc23e69 :: needs changes before merge. :: [P1] Expose the result box to the extracted transpiler
• reviewed 2026-08-08T18:51:48.048Z sha 919a309 :: needs maintainer review before merge. :: none
• reviewed 2026-08-08T19:41:01.441Z sha 2b0d93d :: found issues before merge. :: [P3] Document the reduced QuickJS stack limit
• reviewed 2026-08-08T19:50:46.795Z sha 2b0d93d :: needs changes before merge. :: [P3] Document the reduced QuickJS stack limit
• reviewed 2026-08-08T20:56:23.344Z sha 064b22d :: found issues before merge. :: [P2] Remove the unresolved changelog conflict markers | [P3] Document the 1 MiB QuickJS stack limit