[hyoban]

Important

1. Make sure you have read our contribution guidelines
2. Ensure there is an associated issue
3. Use the correct syntax to link this PR: Fixes #<issue number>.

Summary

Related to #38351.

New Knowledge (web/features/new-rag) and the KnowledgeFS console bridge are developed and deployed from deploy/konwledge. That branch has already rewritten both sides: compared with main, it deletes 38 of main's 57 new-rag files, renames 14, and replaces api/services/knowledge_fs_proxy.py / knowledge_fs_operations.py / controllers/console/knowledge_fs_proxy.py with controllers/console/knowledge_fs/. The copies on main are stale. They are off by default (KNOWLEDGE_FS_ENABLED=false), but they keep receiving repo-wide refactors (46 commits so far), which then conflict when main is merged into the branch.

This PR removes them from main, so deploy/konwledge becomes the only source of truth for this feature.

Removed

• Web: web/features/new-rag, the /datasets/new/** routes, the Legacy/New view switcher on the datasets list, the /datasets/new special cases in the datasets layout and main-nav routes, and the newKnowledge.* translations. newKnowledge.documentsSelected is kept because the legacy document list still uses it.
• Contracts: generated/knowledge-fs, its generator script and config, the contract test, the ./knowledge-fs/* export, and the knowledgeFs segment in the console router and contract loader.
• API: the KnowledgeFS console proxy and operations service, KnowledgeFSConfig, the knowledge_fs_enabled system feature (and its generated contract fields and OpenAPI docs), generate_knowledge_fs_contract.py and its lock file, the KNOWLEDGE_FS_* env examples, and the related unit tests.

Kept

• .github/workflows/deploy-knowledge.yml. It is a workflow_run trigger, and GitHub only reads those from the default branch, so removing it would stop deploy/konwledge deploys.
• Shared helpers that deploy/konwledge still imports: the workspace-permission atoms in web/context/permission-state.ts, registerPageLeaveGuard, and getSystemFeatures. Main no longer uses them, so they are marked /** @public */ for knip. Deleting them here would merge cleanly into the branch and break it.

Follow-up for deploy/konwledge

The next merge of main into the branch will have modify/delete conflicts on these files. Resolve them in favour of the branch (git checkout --ours on the branch side for web/features/new-rag, web/app/(commonLayout)/datasets/new, packages/contracts/generated/knowledge-fs, KnowledgeFS API files, and web/i18n/locales/*/dataset.json).

One change will merge without a conflict and needs a manual fix on the branch: common:operation.deleteConfirmTitle is removed from web/i18n/locales/*/common.json (the build's unused-i18n check requires it, and it has no allowlist). The branch still uses this key in four New Knowledge components and never edited common.json, so after merging main, restore the key in each locale and regenerate the i18n types.

Screenshots

N/A. With the flag off (the default), the datasets list renders exactly as before.

Checklist

• This change requires a documentation update, included: Dify Document
• I understand that this PR may be closed in case there was no previous discussion or issues. (This doesn't apply to typos!)
• I've verified the change and added or updated tests where meaningful regression risk justifies coverage.
• I've updated the documentation accordingly.
• I ran make lint && make type-check (backend) and vp staged (frontend) to appease the lint gods

Verification run locally:

• tsc --noEmit for web and packages/contracts
• vp lint / vp fmt on changed files, ruff check on changed API files
• vp run knip, knip:production, and knip:production-unused-check
• Vitest: datasets list, datasets layout, document titles, main-nav layout, console client, legacy document list
• Pytest: configs/test_env_consistency.py, configs/test_dify_config.py, and two system_feature_service suites

🤖 Generated with Claude Code

[github-actions]

Pyrefly Type Coverage

Metric	Base	PR	Delta
Type coverage	74.88%	74.80%	-0.08%
Strict coverage	74.54%	74.46%	-0.08%
Typed symbols	61,247	60,984	-263
Untyped symbols	20,642	20,639	-3
Modules	3771	3761	-10

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 88.92%. Comparing base (cd8301a) to head (52fd47a).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #44163      +/-   ##
==========================================
- Coverage   89.01%   88.92%   -0.10%     
==========================================
  Files        5336     5288      -48     
  Lines      330154   324120    -6034     
  Branches    65739    63727    -2012     
==========================================
- Hits       293889   288221    -5668     
+ Misses      31064    30720     -344     
+ Partials     5201     5179      -22     

Flag	Coverage Δ
api	88.57% <ø> (-0.01%)	⬇️
cli	90.56% <ø> (ø)
dify-ui	92.25% <ø> (ø)
web	89.21% <100.00%> (-0.22%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[github-actions]

Vinext bundle analysis

JS gzip: 🟢 −211.84 KiB (−2.09%) · CSS gzip: 🟢 −2.78 KiB (−0.83%)

All client output	Base gzip → Merge gzip	Gzip change
JS	10149.80 KiB → 9937.95 KiB	🟢 −211.84 KiB (−2.09%)
CSS	335.16 KiB → 332.39 KiB	🟢 −2.78 KiB (−0.83%)

Entry changes: 🟠 6 increased · 🟢 168 decreased (gzip).

Largest entry changes

Top 5 by absolute gzip delta. Static dependencies overlap; do not sum entries.

Source entry	Base gzip → Merge gzip	Gzip change
app/components/datasets/list/index.tsx	563.59 KiB → 550.64 KiB	🟢 −12.96 KiB (−2.30%)
app/components/workflow-app/index.tsx	1449.19 KiB → 1446.44 KiB	🟢 −2.76 KiB (−0.19%)
app/components/workflow/nodes/agent/panel.tsx	1494.86 KiB → 1492.15 KiB	🟢 −2.71 KiB (−0.18%)
app/components/workflow/nodes/agent-v2/panel.tsx	1639.68 KiB → 1636.98 KiB	🟢 −2.70 KiB (−0.16%)
app/(commonLayout)/datasets/(datasetDetailLayout)/[datasetId]/pipeline/page.tsx	1398.43 KiB → 1395.76 KiB	🟢 −2.67 KiB (−0.19%)

Top 20 of 174 changed entries (full paths and raw sizes)

Source entry	Base gzip → Merge gzip	Gzip change	Raw change
dynamic-entry: app/components/datasets/list/index.tsx	563.59 KiB → 550.64 KiB	🟢 −12.96 KiB (−2.30%)	🟢 −50.87 KiB (−2.81%)
dynamic-entry: app/components/workflow-app/index.tsx	1449.19 KiB → 1446.44 KiB	🟢 −2.76 KiB (−0.19%)	🟢 −17.05 KiB (−0.36%)
dynamic-entry: app/components/workflow/nodes/agent/panel.tsx	1494.86 KiB → 1492.15 KiB	🟢 −2.71 KiB (−0.18%)	🟢 −16.95 KiB (−0.35%)
dynamic-entry: app/components/workflow/nodes/agent-v2/panel.tsx	1639.68 KiB → 1636.98 KiB	🟢 −2.70 KiB (−0.16%)	🟢 −16.95 KiB (−0.32%)
dynamic-entry: app/(commonLayout)/datasets/(datasetDetailLayout)/[datasetId]/pipeline/page.tsx	1398.43 KiB → 1395.76 KiB	🟢 −2.67 KiB (−0.19%)	🟢 −16.85 KiB (−0.37%)
dynamic-entry: app/components/snippets/index.tsx	1360.13 KiB → 1357.47 KiB	🟢 −2.66 KiB (−0.20%)	🟢 −16.80 KiB (−0.38%)
dynamic-entry: app/components/app/configuration/index.tsx	1389.32 KiB → 1386.68 KiB	🟢 −2.64 KiB (−0.19%)	🟢 −16.43 KiB (−0.36%)
dynamic-entry: app/components/explore/try-app/index.tsx	1342.68 KiB → 1340.05 KiB	🟢 −2.62 KiB (−0.20%)	🟢 −16.41 KiB (−0.38%)
dynamic-entry: app/components/workflow/panel/workflow-preview.tsx	1353.47 KiB → 1350.85 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/rag-pipeline/components/panel/test-run/index.tsx	1348.30 KiB → 1345.69 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/panel/record.tsx	1346.74 KiB → 1344.14 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/snippets/components/snippet-run-panel.tsx	1348.62 KiB → 1346.02 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/nodes/trigger-webhook/panel.tsx	1351.57 KiB → 1348.97 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/nodes/code/panel.tsx	1349.54 KiB → 1346.94 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/nodes/iteration/panel.tsx	1346.26 KiB → 1343.66 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/nodes/loop/panel.tsx	1353.08 KiB → 1350.47 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/nodes/variable-assigner/panel.tsx	1348.45 KiB → 1345.85 KiB	🟢 −2.61 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/rag-pipeline/components/panel/input-field/index.tsx	1347.56 KiB → 1344.95 KiB	🟢 −2.60 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/nodes/document-extractor/panel.tsx	1346.16 KiB → 1343.56 KiB	🟢 −2.60 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)
dynamic-entry: app/components/workflow/nodes/start/panel.tsx	1354.10 KiB → 1351.50 KiB	🟢 −2.60 KiB (−0.19%)	🟢 −16.64 KiB (−0.38%)

📦 0 package attribution changes (uncompressed modules)

Showing 0 of 0, ordered by absolute module-size delta.

Package	Base	Merge	Δ
No package attribution changes	—	—	—

Measurement details and build revisions

Base: 5e09cfe050b64d0d435afa750cc6deaf3170d3c8 · Merge: 295a864bd23faace6a66c84536ea4d2bded4f412
Vinext: 1.1.0 → 1.1.0; Rolldown: 1.2.12 → 1.2.12

Raw JS change: 🟢 −903.30 KiB (−2.54%) · Raw CSS change: 🟢 −17.79 KiB (−1.05%)

🟠 Increase · 🟢 Decrease · ⚪ Unchanged. Colors indicate direction, not a performance budget verdict. Percentages use the base size.

Entries follow only static edges in the official Rolldown chunk graph, deduplicating shared files. Dynamic entries are measured when loaded; they are not page first-load metrics. Entry totals overlap and must not be summed.
Package attribution sums module sizes reported by Rolldown, not compressed emitted bytes. Gzip totals are measured per emitted file. This is Vinext output, not Next.js/Turbopack output.
Report only: increases do not fail this check. Missing or incompatible analysis data does.

Download base/merge snapshots and official Rolldown analysis reports.

Comparison: event base SHA vs the tested merge commit at GITHUB_SHA (PR or merge queue).