Build: NX resource-class eval — linux-browsers-js = extra_large+#34572
Closed
kasperpeulen wants to merge 30 commits into
Closed
Build: NX resource-class eval — linux-browsers-js = extra_large+#34572kasperpeulen wants to merge 30 commits into
kasperpeulen wants to merge 30 commits into
Conversation
|
View your CI Pipeline Execution ↗ for commit 4065d56
☁️ Nx Cloud last updated this comment at |
This was referenced Apr 17, 2026
Contributor
📝 WalkthroughWalkthroughThis PR transitions CI infrastructure by removing CircleCI's dynamic configuration entrypoint and trigger workflow, introducing an NX experiment flag to conditionally exclude certain jobs, and adding a comprehensive CI evaluation system with SQLite-based metrics database and analytics scripts for comparing CircleCI vs NX Cloud performance across runs. Changes
Estimated code review effort🎯 4 (Complex) | ⏱️ ~60 minutes Possibly related PRs
✨ Finishing Touches📝 Generate docstrings
Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 4
🧹 Nitpick comments (7)
scripts/ci/sandboxes.ts (1)
142-143: Module-level mutable state creates implicit coupling.Using a module-level
nxExperimentvariable that's mutated bygetSandboxesand read bydefineSandboxFlowcreates hidden dependencies. IfgetSandboxesis called multiple times with different options, the state persists incorrectly.Consider threading
nxExperimentthrough the function parameters instead:♻️ Proposed refactor to pass flag through parameters
-let nxExperiment = false; - -export function defineSandboxFlow<Key extends string>(key: Key) { +export function defineSandboxFlow<Key extends string>(key: Key, nxExperiment = false) {Then update the call in
getSandboxes:- const sandboxes = getListOfSandboxes(workflow).map(defineSandboxFlow); + const sandboxes = getListOfSandboxes(workflow).map((key) => defineSandboxFlow(key, nxExperiment));🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@scripts/ci/sandboxes.ts` around lines 142 - 143, The module-level mutable variable nxExperiment introduces hidden coupling between getSandboxes and defineSandboxFlow; remove nxExperiment from module scope and thread it as an explicit boolean parameter instead: compute the flag locally inside getSandboxes (or its callers) and pass it into defineSandboxFlow (and any other functions that currently read nxExperiment), update the defineSandboxFlow signature to accept the nxExperiment flag, and update all call sites of defineSandboxFlow/getSandboxes to supply the flag so state is not shared implicitly..github/workflows/nx.yml (1)
13-16: Considercancel-in-progress: truefor normal development.Setting
cancel-in-progress: falsemakes sense for the evaluation sweep (collecting 30 runs per branch), but for typical development workflows, canceling superseded runs saves CI resources. Consider reverting this totrueafter the evaluation or making it conditional.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In @.github/workflows/nx.yml around lines 13 - 16, The concurrency block currently sets cancel-in-progress: false which keeps superseded workflow runs running; update the concurrency configuration in the .github/workflows/nx.yml file by setting cancel-in-progress: true for normal development (or make it conditional using an environment variable or workflow input/evaluation flag so it remains false only during the evaluation sweep). Locate the concurrency block (symbols: concurrency, group, cancel-in-progress) and either change cancel-in-progress to true or add a conditional that flips it based on a flag (e.g., EVALUATION_SWEEP or a workflow input) so CI runs are canceled for typical PR updates but retained during the evaluation period.scripts/run-backfill-only.ts (2)
101-104: DuplicatedinferAgentTemplatefunction.This function and
LINUX_JS_TARGETSset are duplicated fromscripts/evaluate-ci.ts(lines 1128-1132). This should be extracted to the same shared module suggested forgetNxCloudHeaders.🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@scripts/run-backfill-only.ts` around lines 101 - 104, The inferAgentTemplate function and LINUX_JS_TARGETS set are duplicated; extract them into the shared module you created/reused for getNxCloudHeaders so both scripts import the same implementation. Move the const LINUX_JS_TARGETS and function inferAgentTemplate to that shared module, export them, then replace the local definitions in scripts/run-backfill-only.ts and scripts/evaluate-ci.ts with imports of the shared inferAgentTemplate and LINUX_JS_TARGETS to eliminate duplication and keep behavior identical.
155-157: Empty catch blocks discard error details.When API calls fail, the empty catch blocks at lines 155-157 and 256-258 only increment a counter. Logging the error message would help diagnose issues during backfill runs.
🔧 Proposed improvement
- } catch { + } catch (e: any) { + if (process.env.DEBUG) console.error(` API error for ${id}: ${e.message}`); apiFailures++; }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@scripts/run-backfill-only.ts` around lines 155 - 157, The catch blocks in scripts/run-backfill-only.ts currently only do "apiFailures++" which discards error details; change both catches to catch the error (e.g., "catch (err)") and log the error before incrementing apiFailures — include err.message and err.stack (or JSON.stringify(err) if needed) and a short contextual message identifying the API call or loop iteration so failures can be diagnosed; keep the existing apiFailures++ behavior after logging.scripts/investigate-nx-cache.ts (1)
19-42: DuplicatedgetNxCloudHeadersfunction across multiple scripts.This function is duplicated in
scripts/run-backfill-only.ts(lines 22-45) andscripts/evaluate-ci.ts(lines 389-413). Consider extracting to a shared utility module.♻️ Proposed: Extract to shared module
Create
scripts/lib/nx-cloud-auth.ts:import { readFileSync } from 'node:fs'; import { join } from 'node:path'; import { parse as parseIni } from 'ini'; const NX_CLOUD_URL = 'https://cloud.nx.app'; const NX_CLOUD_ID = '6929fbef73e98d8094d2a343'; export function getNxCloudHeaders(): Record<string, string> { // ... shared implementation }Then import in each script:
import { getNxCloudHeaders } from './lib/nx-cloud-auth.ts';🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@scripts/investigate-nx-cache.ts` around lines 19 - 42, Extract the duplicated getNxCloudHeaders function into a single shared module (e.g., create scripts/lib/nx-cloud-auth.ts) that exports getNxCloudHeaders and the constants NX_CLOUD_URL and NX_CLOUD_ID, copy the existing implementation (including reading env vars, fallback ini parsing, and header building) into that module, then replace the duplicated definitions in scripts/investigate-nx-cache.ts, scripts/run-backfill-only.ts, and scripts/evaluate-ci.ts by importing getNxCloudHeaders from the new module and removing the local duplicate functions.scripts/evaluate-ci.ts (1)
168-182: Silent migration failures could mask schema issues.The migration try/catch blocks silently swallow all errors. While this is intentional for "column already exists" errors, it also hides actual failures (e.g., disk full, permissions). Consider logging in debug mode.
🔧 Optional: Log migration errors in debug mode
// Migration: add commit_sha column to existing databases try { db.exec(`ALTER TABLE runs ADD COLUMN commit_sha TEXT`); - } catch {} + } catch (e: any) { + // Expected to fail if column exists; log only if DEBUG for other errors + if (process.env.DEBUG && !e.message?.includes('duplicate column')) { + console.warn(`Migration warning: ${e.message}`); + } + }🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed. In `@scripts/evaluate-ci.ts` around lines 168 - 182, The empty catch blocks around the migration db.exec calls swallow all errors; update the two migration blocks (the ALTER TABLE runs ADD COLUMN commit_sha and ALTER TABLE nx_cipe_retry_stats ADD COLUMN hypothetical_no_cache_ms) to handle errors more safely by inspecting the thrown error: if it indicates the column already exists, ignore it, otherwise log the full error and SQL (using the project's logger or console.debug) or rethrow; include the SQL text and error details in the log so real migration failures (disk/permission/etc.) are visible.scripts/generate-canvas-data.ts (1)
563-577: Template literals with embedded expressions may cause incorrect parsing.The brace-matching logic tracks backtick strings but doesn't handle
${...}expressions inside template literals. If a replaced constant contains a template literal like`value: ${obj.prop}`, the{inside${...}would incorrectly incrementdepth.This is likely acceptable for the current use case if the canvas constants don't contain such patterns, but worth noting for future robustness.
🤖 Prompt for all review comments with AI agents
Verify each finding against the current code and only fix it if needed.
Inline comments:
In `@scripts/ci/sandboxes.ts`:
- Around line 521-522: The getSandboxes(workflow, options) function declares the
nxExperiment flag but the caller that invokes getSandboxes(...) never passes
options so nxExperiment is always false; fix by either (A) updating the caller
that invokes getSandboxes(...) to pass { nxExperiment: true } or a boolean
derived from CI config when you want experimental NX jobs enabled, ensuring the
call supplies the options object, or (B) if the option is unused project-wide,
remove the options parameter and the nxExperiment references inside getSandboxes
to eliminate dead code; locate the flag usage in getSandboxes and the single
caller site that invokes getSandboxes(...) to apply one of these two fixes.
In `@scripts/evaluate-ci.ts`:
- Around line 729-743: resolvePRNumber currently calls GitHub API even when no
token is set (it passes an empty Authorization header), risking unauthenticated
rate limits; update the function to first check process.env.GITHUB_TOKEN and
skip or short-circuit the network request when it's missing (e.g., return branch
or cached value immediately) or explicitly fail with a clear error, and only
call fetchJSON with the Authorization header when a non-empty token exists;
reference symbols: resolvePRNumber, prNumberCache, fetchJSON.
In `@scripts/generate-canvas-data.ts`:
- Around line 26-29: CANVAS_PATH is hardcoded to a user-specific absolute path
causing failures for other devs; update the scripts/generate-canvas-data.ts
logic that defines CANVAS_PATH to use a portable default (e.g., resolve relative
to the script using import.meta.dirname or __dirname) and allow override via a
CLI flag or environment variable (check process.env.CANVAS_PATH or parse an
--canvas argument and use its value if present), keeping the symbol name
CANVAS_PATH and falling back to the relative path when no override is provided.
In `@scripts/investigate-nx-cache.ts`:
- Around line 75-78: The code calls fetch(...).then(r => r.json()) without
checking HTTP status, which can throw on error responses; change the fetch usage
in investigate-nx-cache.ts (the call that uses NX_CLOUD_URL, getNxCloudHeaders,
and cipeId to produce details and rg) to first await the Response, verify
response.ok (or throw a new Error including response.status and statusText or
response body), and only then parse response.json() into details so any HTTP
errors produce a clear, descriptive error instead of a JSON parse failure.
---
Nitpick comments:
In @.github/workflows/nx.yml:
- Around line 13-16: The concurrency block currently sets cancel-in-progress:
false which keeps superseded workflow runs running; update the concurrency
configuration in the .github/workflows/nx.yml file by setting
cancel-in-progress: true for normal development (or make it conditional using an
environment variable or workflow input/evaluation flag so it remains false only
during the evaluation sweep). Locate the concurrency block (symbols:
concurrency, group, cancel-in-progress) and either change cancel-in-progress to
true or add a conditional that flips it based on a flag (e.g., EVALUATION_SWEEP
or a workflow input) so CI runs are canceled for typical PR updates but retained
during the evaluation period.
In `@scripts/ci/sandboxes.ts`:
- Around line 142-143: The module-level mutable variable nxExperiment introduces
hidden coupling between getSandboxes and defineSandboxFlow; remove nxExperiment
from module scope and thread it as an explicit boolean parameter instead:
compute the flag locally inside getSandboxes (or its callers) and pass it into
defineSandboxFlow (and any other functions that currently read nxExperiment),
update the defineSandboxFlow signature to accept the nxExperiment flag, and
update all call sites of defineSandboxFlow/getSandboxes to supply the flag so
state is not shared implicitly.
In `@scripts/evaluate-ci.ts`:
- Around line 168-182: The empty catch blocks around the migration db.exec calls
swallow all errors; update the two migration blocks (the ALTER TABLE runs ADD
COLUMN commit_sha and ALTER TABLE nx_cipe_retry_stats ADD COLUMN
hypothetical_no_cache_ms) to handle errors more safely by inspecting the thrown
error: if it indicates the column already exists, ignore it, otherwise log the
full error and SQL (using the project's logger or console.debug) or rethrow;
include the SQL text and error details in the log so real migration failures
(disk/permission/etc.) are visible.
In `@scripts/investigate-nx-cache.ts`:
- Around line 19-42: Extract the duplicated getNxCloudHeaders function into a
single shared module (e.g., create scripts/lib/nx-cloud-auth.ts) that exports
getNxCloudHeaders and the constants NX_CLOUD_URL and NX_CLOUD_ID, copy the
existing implementation (including reading env vars, fallback ini parsing, and
header building) into that module, then replace the duplicated definitions in
scripts/investigate-nx-cache.ts, scripts/run-backfill-only.ts, and
scripts/evaluate-ci.ts by importing getNxCloudHeaders from the new module and
removing the local duplicate functions.
In `@scripts/run-backfill-only.ts`:
- Around line 101-104: The inferAgentTemplate function and LINUX_JS_TARGETS set
are duplicated; extract them into the shared module you created/reused for
getNxCloudHeaders so both scripts import the same implementation. Move the const
LINUX_JS_TARGETS and function inferAgentTemplate to that shared module, export
them, then replace the local definitions in scripts/run-backfill-only.ts and
scripts/evaluate-ci.ts with imports of the shared inferAgentTemplate and
LINUX_JS_TARGETS to eliminate duplication and keep behavior identical.
- Around line 155-157: The catch blocks in scripts/run-backfill-only.ts
currently only do "apiFailures++" which discards error details; change both
catches to catch the error (e.g., "catch (err)") and log the error before
incrementing apiFailures — include err.message and err.stack (or
JSON.stringify(err) if needed) and a short contextual message identifying the
API call or loop iteration so failures can be diagnosed; keep the existing
apiFailures++ behavior after logging.
🪄 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: Organization UI
Review profile: CHILL
Plan: Pro
Run ID: c9c2b57f-9c5f-4e91-9104-2c5fdbac33ed
⛔ Files ignored due to path filters (1)
yarn.lockis excluded by!**/yarn.lock,!**/*.lock
📒 Files selected for processing (16)
.circleci/config.yml.cursor/mcp.json.github/workflows/nx.yml.github/workflows/trigger-circle-ci-workflow.yml.gitignorenx.jsonpackage.jsonscripts/ci-eval.db.README.mdscripts/ci/init-empty.tsscripts/ci/sandboxes.tsscripts/evaluate-ci.tsscripts/generate-canvas-data.tsscripts/investigate-nx-cache.tsscripts/run-backfill-only.tstest-storybooks/portable-stories-kitchen-sink/react-vitest-3/project.jsontest-storybooks/yarn-pnp/project.json
💤 Files with no reviewable changes (2)
- .circleci/config.yml
- .github/workflows/trigger-circle-ci-workflow.yml
Comment on lines
+521
to
+522
| export function getSandboxes(workflow: Workflow, options: { nxExperiment?: boolean } = {}) { | ||
| nxExperiment = options.nxExperiment ?? false; |
Contributor
There was a problem hiding this comment.
🧩 Analysis chain
🏁 Script executed:
#!/bin/bash
# Find all call sites of getSandboxes to verify they're updated appropriately
rg -n "getSandboxes\(" --type ts -A2Repository: storybookjs/storybook
Length of output: 579
🏁 Script executed:
#!/bin/bash
# Check for all call sites of getSandboxes in the entire codebase
echo "=== All getSandboxes call sites ==="
rg -n "getSandboxes\(" --type ts
echo -e "\n=== Check if nxExperiment is used in the function body ==="
# Get the function body from sandboxes.ts starting at line 521
sed -n '521,600p' scripts/ci/sandboxes.ts | head -100Repository: storybookjs/storybook
Length of output: 1006
The nxExperiment option is not being utilized at the call site.
The getSandboxes() function defines an options parameter to control whether experimental NX jobs are added (via the !nxExperiment condition on line 530), but the call in scripts/ci/main.ts:48 does not pass this option. This means nxExperiment will always be false at this call site, rendering the parameter ineffective. Either pass the option from the call site or remove the unused parameter.
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@scripts/ci/sandboxes.ts` around lines 521 - 522, The getSandboxes(workflow,
options) function declares the nxExperiment flag but the caller that invokes
getSandboxes(...) never passes options so nxExperiment is always false; fix by
either (A) updating the caller that invokes getSandboxes(...) to pass {
nxExperiment: true } or a boolean derived from CI config when you want
experimental NX jobs enabled, ensuring the call supplies the options object, or
(B) if the option is unused project-wide, remove the options parameter and the
nxExperiment references inside getSandboxes to eliminate dead code; locate the
flag usage in getSandboxes and the single caller site that invokes
getSandboxes(...) to apply one of these two fixes.
Comment on lines
+729
to
+743
| async function resolvePRNumber(branch: string): Promise<string> { | ||
| if (prNumberCache.has(branch)) return prNumberCache.get(branch)!; | ||
|
|
||
| try { | ||
| const data = await fetchJSON<{ number: number }[]>( | ||
| `https://api.github.com/repos/storybookjs/storybook/pulls?head=storybookjs:${branch}&state=open&per_page=1`, | ||
| { headers: { Authorization: `token ${process.env.GITHUB_TOKEN ?? ''}` } } | ||
| ); | ||
| const num = data[0]?.number?.toString() ?? branch; | ||
| prNumberCache.set(branch, num); | ||
| return num; | ||
| } catch { | ||
| return branch; | ||
| } | ||
| } |
Contributor
There was a problem hiding this comment.
GitHub API calls may hit rate limits without authentication.
resolvePRNumber makes GitHub API calls with an optional token (falling back to empty string). Unauthenticated GitHub API requests are limited to 60/hour, which could be exhausted quickly when syncing many branches.
🛡️ Suggested improvement
async function resolvePRNumber(branch: string): Promise<string> {
if (prNumberCache.has(branch)) return prNumberCache.get(branch)!;
+ const token = process.env.GITHUB_TOKEN;
+ if (!token) {
+ // Skip API call if no token - just use branch name
+ prNumberCache.set(branch, branch);
+ return branch;
+ }
+
try {
const data = await fetchJSON<{ number: number }[]>(
`https://api.github.com/repos/storybookjs/storybook/pulls?head=storybookjs:${branch}&state=open&per_page=1`,
- { headers: { Authorization: `token ${process.env.GITHUB_TOKEN ?? ''}` } }
+ { headers: { Authorization: `token ${token}` } }
);🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@scripts/evaluate-ci.ts` around lines 729 - 743, resolvePRNumber currently
calls GitHub API even when no token is set (it passes an empty Authorization
header), risking unauthenticated rate limits; update the function to first check
process.env.GITHUB_TOKEN and skip or short-circuit the network request when it's
missing (e.g., return branch or cached value immediately) or explicitly fail
with a clear error, and only call fetchJSON with the Authorization header when a
non-empty token exists; reference symbols: resolvePRNumber, prNumberCache,
fetchJSON.
Comment on lines
+26
to
+29
| const CANVAS_PATH = join( | ||
| process.env.HOME!, | ||
| '.cursor/projects/Users-kasperpeulen-cursor-worktrees-storybook-jafl/canvases/nx-vs-circleci-findings.canvas.tsx' | ||
| ); |
Contributor
There was a problem hiding this comment.
Hardcoded user-specific path will break for other developers.
CANVAS_PATH is hardcoded to a specific user's local path (/Users-kasperpeulen-cursor-worktrees-storybook-jafl/...). This script will fail for anyone else trying to run it.
🔧 Proposed fix: Use a relative or configurable path
const CANVAS_PATH = join(
- process.env.HOME!,
- '.cursor/projects/Users-kasperpeulen-cursor-worktrees-storybook-jafl/canvases/nx-vs-circleci-findings.canvas.tsx'
+ import.meta.dirname,
+ '../canvases/nx-vs-circleci-findings.canvas.tsx'
);Alternatively, accept it as a CLI argument:
const CANVAS_PATH = args.includes('--canvas')
? args[args.indexOf('--canvas') + 1]
: join(import.meta.dirname, '../canvases/nx-vs-circleci-findings.canvas.tsx');📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| const CANVAS_PATH = join( | |
| process.env.HOME!, | |
| '.cursor/projects/Users-kasperpeulen-cursor-worktrees-storybook-jafl/canvases/nx-vs-circleci-findings.canvas.tsx' | |
| ); | |
| const CANVAS_PATH = process.env.CANVAS_PATH || join( | |
| process.cwd(), | |
| 'canvases/nx-vs-circleci-findings.canvas.tsx' | |
| ); |
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@scripts/generate-canvas-data.ts` around lines 26 - 29, CANVAS_PATH is
hardcoded to a user-specific absolute path causing failures for other devs;
update the scripts/generate-canvas-data.ts logic that defines CANVAS_PATH to use
a portable default (e.g., resolve relative to the script using
import.meta.dirname or __dirname) and allow override via a CLI flag or
environment variable (check process.env.CANVAS_PATH or parse an --canvas
argument and use its value if present), keeping the symbol name CANVAS_PATH and
falling back to the relative path when no override is provided.
Comment on lines
+75
to
+78
| const details = await fetch(`${NX_CLOUD_URL}/nx-cloud/mcp-context/pipeline-executions/${cipeId}`, { | ||
| headers: getNxCloudHeaders(), | ||
| }).then((r) => r.json()); | ||
| const rg = details?.runGroups?.[0]?.runGroupName; |
Contributor
There was a problem hiding this comment.
Missing response status check before parsing JSON.
The fetch at line 75-77 calls .json() without first checking if the response was successful. If the API returns an error status, this will likely throw a confusing JSON parse error instead of a meaningful HTTP error.
🛡️ Proposed fix
- const details = await fetch(`${NX_CLOUD_URL}/nx-cloud/mcp-context/pipeline-executions/${cipeId}`, {
- headers: getNxCloudHeaders(),
- }).then((r) => r.json());
+ const detailsRes = await fetch(`${NX_CLOUD_URL}/nx-cloud/mcp-context/pipeline-executions/${cipeId}`, {
+ headers: getNxCloudHeaders(),
+ });
+ if (!detailsRes.ok) {
+ console.error(`Failed to fetch pipeline execution: HTTP ${detailsRes.status}`);
+ process.exit(1);
+ }
+ const details = await detailsRes.json();🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.
In `@scripts/investigate-nx-cache.ts` around lines 75 - 78, The code calls
fetch(...).then(r => r.json()) without checking HTTP status, which can throw on
error responses; change the fetch usage in investigate-nx-cache.ts (the call
that uses NX_CLOUD_URL, getNxCloudHeaders, and cipeId to produce details and rg)
to first await the Response, verify response.ok (or throw a new Error including
response.status and statusText or response body), and only then parse
response.json() into details so any HTTP errors produce a clear, descriptive
error instead of a JSON parse failure.
kasperpeulen
changed the title
kasperpeulen
force-pushed
the
kasper/nx-rc-xlarge-plus
branch
2 times, most recently
from
b934e89 to
e774fab
Compare
kasperpeulen
force-pushed
the
kasper/nx-rc-xlarge-plus
branch
from
b4763fc to
9ba80b8
Compare
This was referenced Apr 17, 2026
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
2 participants
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.
Closes #
What I did
NX Cloud resource-class sweep — branch 1 of 6. Do not merge.
Changes exactly one line:
.nx/workflows/agents.yaml→linux-browsers-js.resource-class = docker_linux_amd64/extra_large+(60 credits/min).Part of measuring the real cost / duration / flake curve across
linux-browsers-jsresource classes on realci:normaltraffic. Replaces the single linearly-extrapolated "medium+ projected" column in the NX-vs-CircleCI findings canvas with a 6-point data-backed curve.linux-jsstays atextra_large+across all branches. Parent branchkasper/nx-ai(#34282) strips CircleCI +NX_EXPERIMENTso these 6 PRs only exercise NX Cloud.Sweep matrix:
kasper/nx-rc-xlarge-plusextra_large+kasper/nx-rc-xlargeextra_largekasper/nx-rc-large-pluslarge+kasper/nx-rc-largelargekasper/nx-rc-medium-plusmedium+kasper/nx-rc-mediummediumTraffic plan: cache-bust every ~10 min per branch until 30 runs/branch or Enterprise trial expiry (2026-04-19). Data →
scripts/ci-eval.db→ new "Resource class sweep" section in the findings canvas with cost + duration bar charts and a measured recommendation.Checklist for Contributors
Testing
The changes in this PR are covered in the following automated tests:
Manual testing
Caution
This section is mandatory for all contributions. If you believe no manual test is necessary, please state so explicitly. Thanks!
No manual test necessary — this PR exists only to exercise CI traffic. Verify by opening the PR's NX Cloud CIPE (GH check
nx: normal) and confirming thelinux-browsers-jsagents are running ondocker_linux_amd64/extra_large+(visible in the agent metadata on cloud.nx.app). The sync scriptscripts/evaluate-ci.tsrecordscredit_multiplier = 60per CIPE inscripts/ci-eval.db.Documentation
MIGRATION.MD
Checklist for Maintainers
When this PR is ready for testing, make sure to add
ci:normal,ci:mergedorci:dailyGH label to it to run a specific set of sandboxes. The particular set of sandboxes can be found incode/lib/cli-storybook/src/sandbox-templates.tsMake sure this PR contains one of the labels below:
Available labels
bug: Internal changes that fixes incorrect behavior.maintenance: User-facing maintenance tasks.dependencies: Upgrading (sometimes downgrading) dependencies.build: Internal-facing build tooling & test updates. Will not show up in release changelog.cleanup: Minor cleanup style change. Will not show up in release changelog.documentation: Documentation only changes. Will not show up in release changelog.feature request: Introducing a new feature.BREAKING CHANGE: Changes that break compatibility in some way with current major version.other: Changes that don't fit in the above categories.🦋 Canary release
This PR does not have a canary release associated. You can request a canary release of this pull request by mentioning the
@storybookjs/coreteam here.core team members can create a canary release here or locally with
gh workflow run --repo storybookjs/storybook publish.yml --field pr=<PR_NUMBER>