R0: module-migration rails (src/ serving, live-edit cache, scriptType loading, governance) - #812
Merged
Merged
Conversation
…ype loading (R0)
Host rails for the module-migration refactor: a plugin can move off a single
global-scope screen.js IIFE onto native ES modules with no build step.
- New GET /api/plugins/{id}/src/{path:path} route serving a plugin's src/
source subtree, containment-checked with the same safe_join guard as assets/.
- Live-edit cache contract (Cache-Control: no-cache + weak mtime/size ETag +
If-None-Match -> 304) on src/, screen.js, and assets/ so edited modules reload
on refresh while unchanged ones 304 (screen.js sent no cache headers before;
assets/ emitted an ETag but never revalidated).
- scriptType/minHost passthrough from plugin.json to /api/plugins; the loader
injects a "scriptType":"module" plugin as <script type="module">, which fires
onload only after its static-import graph evaluates -> preserves the
completion-by-onload + _loadingPluginId contract. minHost is passthrough-only.
Classic plugins are unaffected. Verified end-to-end: full pytest (2334) + JS
suite (1019) green; real-server curl (serve/304/traversal/live-edit) and a
headless-browser check (module plugin injected type=module, src/ graph loaded
and executed) both pass.
Tests: tests/test_plugin_src_route.py, tests/js/plugin_loader_script_type.test.js
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…r, eslint gate (R0) The governance rails for the module-migration refactor, gating the later code-move phases. - Constitution -> v1.2.0: Principle II names native ES modules as a first-class, build-free extension mechanism (scriptType:"module" load path, plugins and eventually core static/js/), keeping the no-bundler/source-served rule; new Operating Constraints "Module load contract" clause captures the two findings that make it correct (type=module preserves completion-by-onload; per-visit re-init comes from screen:changed, not screen.js re-execution). Mirrored in CLAUDE.md. - docs/plugin-modules.md: the migration playbook (layering, import-time purity, import.meta.url assets, worklet-separate-graph, the ETag live-edit loop). - docs/size-exemptions.md: the signed 1,500-line size-norm register (Byron signs core/bundled rows, Christian the authored virtuoso row). - Maintainer/CI-only ESLint gate: eslint.config.js (max-lines warn ratchet with register-mirrored ceilings; import-x/no-unresolved + no-cycle hard-error on ES-module graphs, dormant until module code lands) + a `lint` CI job + `npm run lint`. Never on the serve/Docker path (Principle I). eslint . -> 0 errors. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
scripts/perf-baseline.mjs (maintainer/CI-only, Playwright-driven) captures server p50/p95/p99 latency, cold boot-to-interactive, JS-heap after an idle soak, and the injected plugin-script shape (count of type=module) so every refactor phase (R0 -> R3c) can be checked for "screen-entry and frame-time no worse." docs/perf-baseline.md holds the methodology + the R0 baseline; the playback frame-time and chart-loaded screen-entry rows need a seeded library and are re-taken per environment. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (3)
✅ Files skipped from review due to trivial changes (2)
🚧 Files skipped from review as they are similar to previous changes (1)
📝 WalkthroughWalkthroughThis PR adds ES-module plugin support with sandboxed ChangesPlugin module loading and live-edit caching
CI lint gate and size exemptions
Performance baseline harness
Estimated code review effort: 3 (Moderate) | ~30 minutes Sequence Diagram(s)sequenceDiagram
participant Browser
participant PluginAPI
participant FileSystem
Browser->>PluginAPI: GET /api/plugins/{id}/src/{path} with If-None-Match
PluginAPI->>FileSystem: read file mtime+size
PluginAPI->>PluginAPI: compute weak ETag
alt ETag matches
PluginAPI-->>Browser: 304 no body
else changed
PluginAPI-->>Browser: 200 with content, new ETag, no-cache
end
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
There was a problem hiding this comment.
Pull request overview
This PR introduces the “R0 module-migration rails” needed to migrate plugins from a single global-scope screen.js IIFE to a source-served native ES-module graph (no bundler/build step). It adds a sandboxed src/ serving route, establishes a consistent live-edit cache revalidation contract (ETag + 304) across plugin file routes, and updates the loader to inject module plugins via <script type="module">, alongside governance/docs and a CI-only ESLint guardrail + perf baseline harness.
Changes:
- Add
/api/plugins/{id}/src/{path}source-serving route plus sharedno-cache+ weak ETag +If-None-Match→304behavior forsrc/,screen.js, andassets/. - Update frontend plugin loader to set
script.type = 'module'whenplugin.script_type === 'module'and add tests for both the route + loader contract. - Add CI-only ESLint flat config (size norm + module hygiene), a perf-baseline script/docs, and governance/documentation updates.
Reviewed changes
Copilot reviewed 14 out of 15 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
tests/test_plugin_src_route.py |
New FastAPI-level tests covering src/ serving, containment, and ETag/304 live-edit behavior. |
tests/js/plugin_loader_script_type.test.js |
New Node tests asserting the loader’s guarded <script type="module"> injection structure. |
static/app.js |
Loader change: set script.type = 'module' for module plugins while preserving onload completion semantics. |
scripts/perf-baseline.mjs |
New Playwright-driven perf harness to capture server latency + client boot/memory metrics. |
plugins/__init__.py |
Adds src/ route, shared file-response helper with ETag/304, and scriptType/minHost passthrough to /api/plugins. |
package.json |
Adds eslint + eslint-plugin-import-x dev deps and npm run lint. |
package-lock.json |
Lockfile updates for new lint dependencies. |
eslint.config.js |
Flat ESLint config enforcing size warning ratchet and ESM graph hygiene (CI-only). |
docs/size-exemptions.md |
New signed register of line-count exemptions for the size norm. |
docs/plugin-modules.md |
New ES-module migration playbook and non-negotiable rules for plugin module graphs. |
docs/perf-baseline.md |
New methodology doc + initial baseline results section. |
CLAUDE.md |
Adds guidance for ES-module plugins and the module load contract. |
CHANGELOG.md |
Documents the new module-migration rails, governance/lint gate, and perf baseline harness. |
.specify/memory/constitution.md |
Constitution bump to v1.2.0 incorporating native ES modules and the module load contract. |
.github/workflows/ci.yml |
Adds a lint job running npm run lint on Node 20. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
….path inserts - scripts/perf-baseline.mjs: createRequire(import.meta.url) instead of a directory URL (still resolves @playwright/test via upward traversal). - tests/test_plugin_src_route.py: drop the sys.path inserts — pyproject's pytest pythonpath=[".","lib"] already makes `plugins` importable, and the global inserts persisted for the whole run. (Kept _if_none_match as a strict compare — we serve one weak ETag the browser echoes verbatim; weak/strong/wildcard matching is dead for these GETs in a single-user self-hosted app.) Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Contributor
Author
|
@coderabbitai review |
Contributor
✅ Action performedReview finished.
|
Contributor
Author
|
@coderabbitai review |
Contributor
✅ Action performedReview finished.
|
Contributor
Author
|
@coderabbitai review |
Contributor
✅ Action performedReview finished.
|
Contributor
There was a problem hiding this comment.
Actionable comments posted: 3
🧹 Nitpick comments (7)
eslint.config.js (1)
22-25: 📐 Maintainability & Code Quality | 🔵 TrivialMagic-number "unlimited" ceiling.
100000is a de facto "no ceiling" for the twoscreen.jsexemptions, but a named constant (e.g.const NO_CEILING = Number.MAX_SAFE_INTEGERorInfinity) would make the intent explicit rather than relying on readers inferring "huge number = unlimited."🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@eslint.config.js` around lines 22 - 25, The two `screen.js` ESLint exemptions in `eslint.config.js` use a magic number to represent an effectively unlimited max, so replace the hardcoded `100000` values with a named constant such as `NO_CEILING` defined near the config and reused in both entries. Update the `files` rules for `plugins/capability_inspector/screen.js` and `plugins/folder_library/screen.js` to reference that constant so the intent is explicit and easy to maintain.package.json (1)
15-16: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winTrack the ESLint v10 bump soon.
eslint-plugin-import-xsupports ESLint v10, so the upgrade path isn’t blocked here.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@package.json` around lines 15 - 16, The dependency set in package.json is still pinned to ESLint v9, so plan and apply the ESLint v10 upgrade soon. Update the eslint entry in the package manifest and verify any related lint tooling, including eslint-plugin-import-x integration, still works with the new major version; use the package.json dependency list as the main place to locate the change.scripts/perf-baseline.mjs (3)
53-70: 🩺 Stability & Availability | 🔵 Trivial | ⚡ Quick winWrap browser lifecycle in try/finally to avoid orphaned Chromium processes.
If
page.goto(60s timeout) or anypage.evaluatecall throws,browser.close()at Line 68 never runs, leaking a Chromium process on every failed run.♻️ Proposed fix
async function clientMetrics() { const browser = await chromium.launch(); - const page = await browser.newPage(); - const t0 = Date.now(); - await page.goto(BASE, { waitUntil: 'networkidle', timeout: 60000 }); - const bootMs = Date.now() - t0; + try { + const page = await browser.newPage(); + const t0 = Date.now(); + await page.goto(BASE, { waitUntil: 'networkidle', timeout: 60000 }); + const bootMs = Date.now() - t0; - // performance.memory is Chromium-only; JS heap after settle. - const mem0 = await page.evaluate(() => (performance.memory ? performance.memory.usedJSHeapSize : null)); - await page.waitForTimeout(SOAK_S * 1000); - const mem1 = await page.evaluate(() => (performance.memory ? performance.memory.usedJSHeapSize : null)); + // performance.memory is Chromium-only; JS heap after settle. + const mem0 = await page.evaluate(() => (performance.memory ? performance.memory.usedJSHeapSize : null)); + await page.waitForTimeout(SOAK_S * 1000); + const mem1 = await page.evaluate(() => (performance.memory ? performance.memory.usedJSHeapSize : null)); - const scripts = await page.evaluate(() => - document.querySelectorAll('script[data-plugin-id]').length); + const scripts = await page.evaluate(() => + document.querySelectorAll('script[data-plugin-id]').length); - await browser.close(); - return { bootMs, memStartMB: mem0 && mem0 / 1048576, memSoakMB: mem1 && mem1 / 1048576, scripts }; + return { bootMs, memStartMB: mem0 && mem0 / 1048576, memSoakMB: mem1 && mem1 / 1048576, scripts }; + } finally { + await browser.close(); + } }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@scripts/perf-baseline.mjs` around lines 53 - 70, The clientMetrics() flow in the perf-baseline script should guarantee browser cleanup even when page.goto or either page.evaluate call throws. Wrap the Chromium launch/newPage work in a try/finally block and move browser.close() into the finally path so the browser is always closed regardless of failures; keep the existing clientMetrics() return shape and timing/memory collection logic unchanged.
33-50: 🩺 Stability & Availability | 🔵 Trivial | ⚡ Quick winNo timeout on server-latency fetches.
If
BASEis unreachable or hangs,fetch(BASE + path)at Line 41 can block the loop indefinitely (Node's default fetch has no timeout), stalling the whole harness with no output. Consider anAbortControllerwith a bounded timeout so a hung endpoint degrades to a recorded failure rather than a hang.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@scripts/perf-baseline.mjs` around lines 33 - 50, The serverLatency() helper currently calls fetch(BASE + path) without any timeout, so a hung or unreachable endpoint can stall the perf harness indefinitely. Update serverLatency() to use an AbortController (or equivalent bounded timeout) around the fetch call, and treat timeout aborts as a recorded failure by setting status appropriately and continuing the loop, so the existing per-path timing rows still get produced.
15-17: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winDeclare the Node version floor for this script.
package.jsonhas noenginesfield, while this.mjsentrypoint uses top-levelawaitand CI is pinned to Node 20. Add an explicit engine requirement so local and CI expectations stay aligned.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@scripts/perf-baseline.mjs` around lines 15 - 17, This script relies on a minimum Node runtime, so add an explicit engine floor to keep local and CI expectations aligned. Update the package metadata to declare the required Node version for the perf-baseline entrypoint, and make sure it matches the top-level await usage in the script and the Node 20 CI pin. Reference the perf-baseline.mjs entrypoint and the package.json engines configuration when making the change.docs/perf-baseline.md (1)
9-16: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winAdd a language to the fenced code block (MD040).
Static analysis flags this block as missing a language tag.
📝 Proposed fix
-``` +```sh # 1. start core against a library with real charts (see caveat below) CONFIG_DIR=… DLC_DIR=/path/to/songs PYTHONPATH=lib \ python3 -m uvicorn server:app --host 127.0.0.1 --port 8000🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@docs/perf-baseline.md` around lines 9 - 16, The fenced code block in the perf baseline docs is missing a language tag, triggering MD040. Update the markdown fence around the example commands in the perf baseline section to include an appropriate shell language identifier, using the existing fenced block in docs/perf-baseline.md as the target.Source: Linters/SAST tools
plugins/__init__.py (1)
2444-2497: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winConsider extracting the shared "serve file from plugin subtree" logic.
plugin_assetandplugin_srcnow duplicate the samesafe_join→is_file()→_plugin_file_response(_plugin_media_type(...))sequence almost verbatim, differing only in subdirectory name and log label. Since both were touched together in this PR, extracting a small shared helper would remove the duplication and prevent the two routes from silently drifting apart in future edits.♻️ Proposed helper extraction
+def _serve_plugin_subtree_file( + request: Request, plugin_dir: Path, subdir: str, rel_path: str, plugin_id: str, kind: str +) -> Response | None: + """Shared safe_join + conditional-cache serving for a plugin's assets/ or + src/ subtree. Returns None (caller should 404) if the path is rejected or + missing.""" + target = safe_join(plugin_dir / subdir, rel_path) + if target is None: + log.warning("Plugin %r: %s path rejected: %r", plugin_id, kind, rel_path) + return None + if target.is_file(): + return _plugin_file_response(request, target, _plugin_media_type(target)) + return None + + `@app.get`("/api/plugins/{plugin_id}/assets/{asset_path:path}") def plugin_asset(request: Request, plugin_id: str, asset_path: str): ... for p in snapshot: if p["id"] == plugin_id: if p.get("status", "ready") != "ready": break - target = safe_join(p["_dir"] / "assets", asset_path) - if target is None: - log.warning("Plugin %r: asset path rejected: %r", plugin_id, asset_path) - break - if target.is_file(): - return _plugin_file_response(request, target, _plugin_media_type(target)) - break + resp = _serve_plugin_subtree_file(request, p["_dir"], "assets", asset_path, plugin_id, "asset") + if resp is not None: + return resp + break return Response("", status_code=404)Same pattern applies to
plugin_src.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@plugins/__init__.py` around lines 2444 - 2497, Extract the duplicated plugin file-serving flow in plugin_asset and plugin_src into a shared helper so both routes use the same safe_join, is_file, and _plugin_file_response/_plugin_media_type logic. Keep the helper parameterized by the base subtree and log label, then have both plugin_asset and plugin_src call it so their behavior stays identical and future changes only happen in one place.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In @.github/workflows/ci.yml:
- Line 136: The new lint job’s checkout currently uses the default
actions/checkout@v4 behavior, which persists the GitHub token in local git
config; update the checkout step to disable credential persistence by setting
persist-credentials to false. Make this change on the lint job’s checkout in the
workflow so subsequent steps like npm ci cannot access the token, and keep the
rest of the job unchanged.
In `@docs/perf-baseline.md`:
- Line 15: The perf baseline document has a mismatch between the recorded run
header and the example command: the `Running it` invocation uses `--n 60`, while
the R0 baseline summary says `n=50`. Verify the actual harness value used for
the recorded run, then update the baseline header in the `R0 baseline` section
(and the example command if needed) so the documented `n` matches the real run.
In `@docs/plugin-modules.md`:
- Around line 47-50: The asset-path example in the plugin module docs uses a
module-relative URL that points to the wrong location from files like
src/main.js. Update the example in the assets/import.meta.url guidance to show a
path relative to the importing module, such as ../assets/x.js from src/main.js,
or use the absolute /api/plugins/<id>/assets/x.js route. Keep the references to
import.meta.url and assets/ clear so readers can locate the correct pattern.
---
Nitpick comments:
In `@docs/perf-baseline.md`:
- Around line 9-16: The fenced code block in the perf baseline docs is missing a
language tag, triggering MD040. Update the markdown fence around the example
commands in the perf baseline section to include an appropriate shell language
identifier, using the existing fenced block in docs/perf-baseline.md as the
target.
In `@eslint.config.js`:
- Around line 22-25: The two `screen.js` ESLint exemptions in `eslint.config.js`
use a magic number to represent an effectively unlimited max, so replace the
hardcoded `100000` values with a named constant such as `NO_CEILING` defined
near the config and reused in both entries. Update the `files` rules for
`plugins/capability_inspector/screen.js` and `plugins/folder_library/screen.js`
to reference that constant so the intent is explicit and easy to maintain.
In `@package.json`:
- Around line 15-16: The dependency set in package.json is still pinned to
ESLint v9, so plan and apply the ESLint v10 upgrade soon. Update the eslint
entry in the package manifest and verify any related lint tooling, including
eslint-plugin-import-x integration, still works with the new major version; use
the package.json dependency list as the main place to locate the change.
In `@plugins/__init__.py`:
- Around line 2444-2497: Extract the duplicated plugin file-serving flow in
plugin_asset and plugin_src into a shared helper so both routes use the same
safe_join, is_file, and _plugin_file_response/_plugin_media_type logic. Keep the
helper parameterized by the base subtree and log label, then have both
plugin_asset and plugin_src call it so their behavior stays identical and future
changes only happen in one place.
In `@scripts/perf-baseline.mjs`:
- Around line 53-70: The clientMetrics() flow in the perf-baseline script should
guarantee browser cleanup even when page.goto or either page.evaluate call
throws. Wrap the Chromium launch/newPage work in a try/finally block and move
browser.close() into the finally path so the browser is always closed regardless
of failures; keep the existing clientMetrics() return shape and timing/memory
collection logic unchanged.
- Around line 33-50: The serverLatency() helper currently calls fetch(BASE +
path) without any timeout, so a hung or unreachable endpoint can stall the perf
harness indefinitely. Update serverLatency() to use an AbortController (or
equivalent bounded timeout) around the fetch call, and treat timeout aborts as a
recorded failure by setting status appropriately and continuing the loop, so the
existing per-path timing rows still get produced.
- Around line 15-17: This script relies on a minimum Node runtime, so add an
explicit engine floor to keep local and CI expectations aligned. Update the
package metadata to declare the required Node version for the perf-baseline
entrypoint, and make sure it matches the top-level await usage in the script and
the Node 20 CI pin. Reference the perf-baseline.mjs entrypoint and the
package.json engines configuration when making the change.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: b691d7c9-7c05-462a-9cde-e3dfe9e28c7f
⛔ Files ignored due to path filters (1)
package-lock.jsonis excluded by!**/package-lock.json
📒 Files selected for processing (14)
.github/workflows/ci.yml.specify/memory/constitution.mdCHANGELOG.mdCLAUDE.mddocs/perf-baseline.mddocs/plugin-modules.mddocs/size-exemptions.mdeslint.config.jspackage.jsonplugins/__init__.pyscripts/perf-baseline.mjsstatic/app.jstests/js/plugin_loader_script_type.test.jstests/test_plugin_src_route.py
- .github/workflows/ci.yml: persist-credentials: false on the lint job's checkout (it runs npm ci with third-party postinstall scripts and never pushes) — zizmor artipacked. - docs/plugin-modules.md: correct the import.meta.url asset example — assets/ is at the plugin root, so a src/ module needs `../assets/x.js` (or the absolute /api/plugins/<id>/assets/ route), not `assets/x.js` (which resolves under src/). - docs/perf-baseline.md: note the recorded R0 numbers were a quick --n 50 --soak 8 pass; the recommended run stays --n 60 --soak 30. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
R0 — module-migration rails
The first phase of the monolith-killing refactor: the one-time host enablement that lets a plugin move off a single global-scope
screen.jsIIFE onto a native ES-module graph — no build step, no framework, source-served. No plugin can migrate until core can serve and load asrc/module graph; today a migrated plugin would 404. This unblocks the R1 pilots (stems, then studio) and gates the later code-move phases.Three commits, each its own concern:
A1 — host code (
feat(plugins))GET /api/plugins/{id}/src/{path}route serving a plugin'ssrc/source subtree, containment-checked by the samesafe_joinguard asassets/(traversal/absolute/NUL → 404).Cache-Control: no-cache+ a weak mtime/sizeETag+If-None-Match→304onsrc/,screen.js, andassets/(previouslyscreen.jssent no cache headers andassets/emitted an ETag but never revalidated). An edited module reloads on refresh; unchanged ones304.scriptType/minHostpassthrough fromplugin.jsonto/api/plugins; the loader injects a"scriptType":"module"plugin as<script type="module">. A module script's load event awaits its whole static-import graph, so the loader's completion-by-onload+_loadingPluginIdcontract is preserved. Classic plugins are untouched;minHostis passthrough-only (enforcement deferred to a later phase).A2 — governance & rails (
docs(governance))scriptType:"module"load path), keeping the no-bundler/source-served rule; a new Operating Constraints "Module load contract" clause. Mirrored inCLAUDE.md.docs/plugin-modules.md(migration playbook) anddocs/size-exemptions.md(the signed 1,500-line size-norm register).max-lineswarns at 1,500 as a non-blocking ratchet (exempt-file ceilings mirror the register), andimport-x/no-unresolved+no-cyclehard-error on ES-module graphs. NewlintCI job; never on the serve/Docker path.A0 — perf baseline (
chore(perf))scripts/perf-baseline.mjs(maintainer-only) +docs/perf-baseline.md— a rerunnable harness (server p50/p95/p99, cold boot, JS-heap soak) so every phase can be checked for "screen-entry and frame-time no worse."Deliberately scoped out (validated/enforced in later phases)
boot()re-mount hook — research showed core never re-injectsscreen.js; plugins re-init via thescreen:changedevent, so a module plugin behaves identically. Confirmed by the R1 pilot's leave/re-enter test instead of built speculatively.minHostenforcement (refuse-with-message) — deferred; passthrough only for now.Verification
pytest(2334 passed) +node --testJS suite (1019 passed);eslint .0 errors.type=module, itssrc/graph loaded and executed).main— no regression to classic-plugin loading.🤖 Generated with Claude Code
Summary by CodeRabbit
src/andscreen.jsviaplugin.json(scriptType,minHost).screen.js,src/, andassetsusingETag-based conditional responses.type="module"for correct load timing.