ENG-12682 Add opt-in React Compiler support

[FarhanAliRaza]

Add experimental, opt-in React Compiler support for generated components. Existing apps retain the current build and memoization behavior by default.

config = rx.Config(app_name="my_app", react_compiler=True)

REFLEX_REACT_COMPILER=true provides the equivalent environment setting. The compiler runs before React Router in development and production, targeting React 19 with normal inference and correctness bailouts. Scope is limited to generated app/, app_components/, and utils/components/ JavaScript/JSX; runtime state providers, event utilities, dependencies, and custom public assets are excluded.

The flag installs pinned Babel/compiler development dependencies through the existing dependency cache. Disabling it removes the transform and optional dependencies when no longer needed. The Vite plugin prebundles react/compiler-runtime, which is injected after dependency scanning, to prevent first-load HTTP 504 errors.

Validation:

• Compiler off/on × development/production browser coverage: 40 passed, 2 skipped. Covers handler arguments, controlled inputs, recursive components, keyed row identity, filtered arrays, and hot reload of an optimized component without losing state. The two skips inspect development-only source modules.
• Unit suite: 9,722 passed, 18 skipped, 1 deselected; 74.02% coverage. The deselected test is test_run_granian_backend_refuses_requests_while_the_app_is_broken: it also fails on clean upstream 7cac2faf9 (recovery assertion there, refusal assertion on this run). The backend implementation and that test are unchanged by this PR.
• Repository Ruff checks/formatting, Pyright, and every applicable pre-commit hook passed, including stub generation, ty, and Biome formatting.
• Disabled Vite output matches the previous template byte-for-byte across 32 existing flag combinations.

Runtime gains depend on the workload; list-heavy updates can regress and compilation adds build time/memory. Documentation makes these tradeoffs explicit. This PR preserves existing Reflex memoization and does not include the fixture-specific filter-callback rewrite from the performance experiment.

<source media="(prefers-color-scheme: dark)" srcset="https://www.cubic.dev/buttons/review-in-cubic-dark.svg"><source media="(prefers-color-scheme: light)" srcset="https://www.cubic.dev/buttons/review-in-cubic-light.svg"><img alt="Review in cubic" src="https://www.cubic.dev/buttons/review-in-cubic-dark.svg">

[greptile-apps]

Confidence Score: 5/5

[Medium impact] The PR appears safe to merge, with no outstanding previous findings or accepted new issues.

Summary

Adds experimental, opt-in React Compiler support for generated frontend components while preserving existing behavior by default.

• Adds the react_compiler configuration option and environment-variable support.
• Installs pinned compiler dependencies and injects a pre-router Vite transform for generated component modules.
• Documents performance tradeoffs and expands unit, development, production, and HMR coverage.
• The changes made since the previous review introduce no additional code changes, and both previous findings are resolved.

_{Reviews (7) · Last reviewed commit: "Merge branch 'main' into codex/react-com..." · Reviewed by Greptile}

[codspeed]

Merging this PR will not alter performance

⚠️ 15 benchmarks spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• test_modify_state[disk]
• test_cli_startup[help]
• test_cli_startup[version]
• test_emit_update
• test_set_state[disk]
• test_cli_startup[deploy_help]
• test_cli_startup[cloud_help]
• test_import_cli
• test_cli_startup[run_help]
• test_yield_intermediate_update
• …and 5 more

✅ 148 untouched benchmarks

---

_{Comparing codex/react-compiler-opt-in (269734d) with main (dd41524)}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-22T20:06:20.624073Z	258d66c	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to 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" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[cubic-dev-ai]

All reported issues were addressed across 14 files

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[linear-code]

ENG-12682

[masenf]

this would be the perfect thing to implement as a config Plugin... except the framework provides no "nice" mechanism for updating the vite config from a plugin.

implementation is fine for now, but we probably would want to move it into a plugin later, once the necessary interfaces are in place.

[FarhanAliRaza]

The integration failures on 269734d come from a timing problem on npm, not from this PR. In run 37386208888, every failing integration-app-harness and -playwright job broke while installing frontend packages:

error: No version matching "1.3.8" found for specifier "@radix-ui/react-popper" (but package exists)

radix-ui@1.7.0 was published at 23:09:03 UTC, but some of the exact @radix-ui/* versions it requires were published up to 23:10:51. For example, @radix-ui/react-popper@1.3.8 appeared at 23:10:34. The installs ran at 23:09:17–20, so bun resolved 1.7.0 before those packages existed. That's why all the jobs failed in the same second. The AssertionErrors in the summaries come from pytest re-running the fixtures that had already broken.

None of the failing fixtures turn on react_compiler. The compiler-on test_memo tests passed in both dev and prod with main's refreshed dependencies.

All of those packages are published now, so "Re-run failed jobs" on run 37386208888 should pass with no code change. I can't trigger re-runs from here (403).

CodSpeed's two test_state_access regressions aren't from this PR either: its diff against main changes no Python state code.

---

Generated by Claude Code

[FarhanAliRaza]

@masenf agreed. I briefly tried moving this into a plugin with a new Vite-config hook, but the framework doesn't have a settled interface for plugins to update the Vite config yet, so I dropped that attempt. The PR keeps the react_compiler config flag for now, and it can move into a plugin once those interfaces exist.

---

Generated by Claude Code

[cubic-dev-ai]

All reported issues were addressed across 21 files (changes from recent commits).

<sub>Tip: Review your code locally with the cubic CLI to iterate faster.

View guided diff | Turn on auto-fix | Re-trigger cubic</sub>