docs: Completely remove deprecated docs

[ovr]

/docs-mintlify has been the active documentation site for a while, so this removes the frozen Next.js/Nextra site at /docs (458 files, 5.7 MB) along with the four docs-mintlify/scripts/*.py migration scripts that only existed to read from or write into it. The README hero images lived under docs/content/ and were hotlinked via raw.githubusercontent, so they move to .github/assets/ with the two URLs updated. Dangling references are cleaned up: the /docs/ CODEOWNERS rule, the docs/ oxlint ignore, the CONTRIBUTING link pointing contributors at the wrong folder, and the "legacy site" notes in both CLAUDE.md files — and the docs PR label glob is repointed from docs/* to docs-mintlify/**/*, which previously had no labeler rule at all. docs was never a yarn workspace or a lerna package and no workflow referenced it, so nothing in the build/test/publish pipeline is affected.

⚠️ Before merging: the legacy site also served the cross-domain 301s from cube.dev/docs/* to docs.cube.dev (~1900 entries in redirects-new-docs.json), so the Vercel project building from docs/ needs to be disconnected and those redirects re-homed outside this repo — otherwise the legacy URLs start 404ing.

Check List

• Tests have been run in packages where changes have been made if available
• Linter has been run for changed code
• Tests for the changes have been added if not covered yet
• Docs have been added / updated if required

Note on the linter: this worktree has no node_modules, so yarn lint could not execute (oxlint fails to load @stylistic/eslint-plugin). The .oxlintrc.json change is the removal of a single ignorePatterns entry and the file was verified to still parse; CI will run the real lint.

🤖 Generated with Claude Code

[claude]

Claude finished @ovr's task in 2m 11s —— View job

---

Review: deletion is clean and fully de-referenced — 0 high, 1 medium, 2 low. The ⚠️ redirect caveat in the PR body is the only merge-order blocker.

Full review — verification performed, findings, and notes

Todo

• Gather context (diff, changed files, commit message)
• Check for dangling references to docs/ across the repo
• Verify README / CONTRIBUTING / CLAUDE.md edits and asset moves
• Verify labeler / oxlint / CODEOWNERS changes
• Check deleted docs-mintlify/scripts/*.py for lost functionality
• Post review

What I verified

The non-docs surface is small (13 files, +9/−1179), so I checked each claim in the PR body against the tree rather than trusting the summary.

Claim	Verified
docs was never a yarn workspace or lerna package	✅ root package.json workspaces are rust/* + packages/* only; lerna.json has no explicit package list
No workflow referenced docs/	✅ only hit under .github/ is .claude/docs/review-threads.md in claude-code-review.yml — unrelated path
No dangling references remain	✅ repo-wide grep for docs/content, cube/master/docs/, tree/master/docs, redirects-new-docs returns only the updated CONTRIBUTING line
No Vercel/Netlify config in-repo	✅ no vercel.json / netlify.toml / .vercelignore at root — the Vercel wiring is project-side, matching the PR body's warning
Legacy site was a redirect shim	✅ docs/next.config.mjs was / → docs.cube.dev/docs/introduction plus the spread of redirects-new-docs.json; content was frozen
Both hero images landed	✅ present at .github/assets/, 6.7 KB + 437 KB, detected by git as pure renames (Bin 0 → 0)
.oxlintrc.json still parses	✅ single ignorePatterns entry removed, docs-mintlify/ retained

The docs labeler glob is a genuine fix, not just a repoint: docs/* matched only top-level files in a folder whose content all lived under content/, so the label effectively never fired. docs-mintlify/**/* is correct.

Findings

Medium — the redirect table becomes unregenerable after this commit (inline)

build_old_site_redirects.py derives destinations by text-matching old page bodies against the Mintlify tree (--redirects ../docs/redirects.json --old-content ../docs/content/product). Both inputs and the generated docs/redirects-new-docs.json are deleted in this same commit. So the person re-homing the ~1900 301s can no longer regenerate them from the tree — only via git show 34e7b2b^:docs/redirects-new-docs.json. The script's own header documents that a naive prefix rewrite got ~76% of destinations wrong, so hand-rolling /docs/:path* → docs.cube.dev/:path* is a trap, not a fallback. Cleanest fix is merge-order: land the re-homing first. Otherwise keep just the generated JSON in-repo until the new host consumes it.

Low — three sibling migration scripts survive the cull (docs-mintlify/scripts/)

transform_components.py (Nextra→Mintlify component rewrites), rewrite_links.py (/product/* → /docs/* path table), and update_frontmatter.py are the same one-off migration vintage as the four deleted. They take a path argument so they don't break — nothing references ../docs — but by the PR's own rationale ("only existed to read from or write into it") they're equally vestigial, and rewrite_links.py's path table is the stale intermediate layout that build_old_site_redirects.py's header explicitly warns produced 404s. Not a blocker; either delete them in this sweep or leave them and let a follow-up handle it. Flagging only so the inconsistency is a decision rather than an oversight.

Low — README image URLs pin /master/ (inline)

Both hero images render broken on this branch until the commit reaches master. Self-resolving on merge; relative paths would sidestep it.

Notes, no action needed

• docs-mintlify/CLAUDE.md:3 still reads "All documentation work should happen here" after the "legacy site" note above it was dropped. Slightly orphaned phrasing now that there's no other site to contrast against, but it's still true and still useful.
• Commit message is unusually good — the removed-and-why table means the deleted scripts' provenance is recoverable from git log alone, which is most of what mitigates the medium finding.
• Nothing to review on testing: no test touches docs/, and the deletion can't affect build/test/publish given the workspace check above. The PR body's note that yarn lint couldn't run locally (no node_modules) is fine — CI covers it and the .oxlintrc.json delta is one removed line.

· branch [`remove-legacy-docs-folder`](https://github.com/cube-js/cube/tree/remove-legacy-docs-folder)

[claude]

The ~1900-entry redirect table stops being regenerable after this commit.

Per this script's own usage block, it derives destinations by text-matching old page bodies against the Mintlify tree:

python build_old_site_redirects.py \
    --redirects ../docs/redirects.json \
    --old-content ../docs/content/product \
    --new-root ..

Both inputs — docs/redirects.json and docs/content/product — are deleted in this same commit, and so is the generated output docs/redirects-new-docs.json. The PR body correctly flags that the cube.dev/docs/* 301s need re-homing outside the repo, but once this lands, whoever does that re-homing cannot regenerate the table from the tree; they have to git show 34e7b2b^:docs/redirects-new-docs.json (and the script's header comment explains why a naive prefix rewrite gets ~76% of destinations wrong, so hand-rolling is not an option either).

Concretely: someone re-homes the redirects six months from now, doesn't know the generator ever existed, writes a /docs/:path* -> docs.cube.dev/:path* prefix rule, and most legacy URLs 301 into 404s.

Two ways out, either is fine:

• Land the re-homing first, then merge this. Cleanest — the ⚠️ in the PR body becomes a no-op.
• Or keep docs/redirects-new-docs.json (the generated output only — the 3862-line JSON, not the site) in-repo until the new host consumes it, so the data survives in the working tree rather than only in history.

If neither, at minimum record the recovery commit SHA in the PR/issue that tracks the Vercel disconnect, so the pointer isn't just this commit message.

[claude]

Minor: these URLs hardcode /master/, so both hero images render broken on this branch (and in the PR's own README preview) until the commit reaches master — the old path is gone and the new one isn't on master yet. Self-resolving on merge, so not a blocker.

Since the assets now live in this repo rather than being cross-referenced, relative paths would sidestep the branch pin entirely and survive future moves:

Suggested change

	<a href="https://cube.dev?ref=github-readme"><img src="https://raw.githubusercontent.com/cube-js/cube/master/.github/assets/cube-core-logo.png" alt="Cube Core — Open-Source Semantic Layer" width="300px"></a>
	</p>
	<p align="center">
	<a href="https://cube.dev?ref=github-readme"><img src=".github/assets/cube-core-logo.png" width="300px"></a>
	</p>

One caveat if you take this: relative image paths only resolve where the README is rendered alongside the repo (GitHub, and it stays correct across branches/forks). This root README.md isn't published to npm — the client packages ship their own — so there's no registry-rendering regression, but it's the reason the absolute form was probably chosen originally. Your call; the absolute form is correct as written.

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated
cubed3	🟢 Ready	View Preview	Sep 10, 2026, 4:34 PM

💡 Tip: Enable Automations to automatically generate PRs for you.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 60.16%. Comparing base (98eb17b) to head (34e7b2b).
⚠️ Report is 1 commits behind head on master.

❗ There is a different number of reports uploaded between BASE (98eb17b) and HEAD (34e7b2b). Click for more details.

HEAD has 1 upload less than BASE

Flag	BASE (98eb17b)	HEAD (34e7b2b)
cubesql	1	0

Additional details and impacted files

@@             Coverage Diff             @@
##           master   #11839       +/-   ##
===========================================
- Coverage   84.74%   60.16%   -24.58%     
===========================================
  Files         261      239       -22     
  Lines       87151    19204    -67947     
  Branches        0     3886     +3886     
===========================================
- Hits        73855    11555    -62300     
+ Misses      13296     7099     -6197     
- Partials        0      550      +550     

Flag	Coverage Δ
cube-backend	60.16% <ø> (?)
cubesql	?

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:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.