feat(visual): automatically detect hover interactions from CSS diffs

[JSONbored]

Summary

• review.visual.interactions required a maintainer to hand-author CSS selectors ahead of time before any hover/click/drag GIF could be captured — the actual goal is for relevant screenshots/GIFs to be captured automatically for any frontend/visual change, with zero pre-selection.
• New review.visual.auto_detect_interactions (config-as-code, global-default + per-repo): scans a PR's own diff for a newly-added :hover/:focus-visible CSS rule in a plain .css/.scss/.sass/.less file and automatically captures a hover-interaction GIF for it, targeting the PR's own first captured route — no maintainer selector-authoring step.
• Composes with the existing manual interactions list: a hand-authored selector always wins over the same one auto-detected (it may carry a label/path the detector can't infer); both draw from the same 3-per-PR cap.
• Scoped to plain stylesheets on purpose — a Tailwind utility class or CSS-in-JS :hover state has no selector explicit in the diff text to extract this way. Click/drag auto-detection is out of scope (far higher false-positive rate; not every onClick is visually interesting).
• Default false, byte-identical to today.

Test plan

• npm run typecheck
• Full unit suite (npx vitest run test/unit) — 18,585 tests passing
• npm run docs:drift-check / manifest:drift-check / engine-parity:drift-check
• Contract parity suites (engine-parity, live-gate-parity, coding-agent-driver-parity)
• Scoped coverage confirms 100% patch coverage on every changed line/branch
• New unit tests for the detection module itself (added/removed/context diff lines, comma-separated selector lists, file-extension scoping, dedup, the 3-selector cap) and for buildCapture's wiring (manual-vs-auto dedup, shared cap, route targeting, config gating)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.13%. Comparing base (6c834ab) to head (5b8d150).
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #7382   +/-   ##
=======================================
  Coverage   91.13%   91.13%           
=======================================
  Files         715      716    +1     
  Lines       72579    72613   +34     
  Branches    20166    20177   +11     
=======================================
+ Hits        66143    66177   +34     
  Misses       5394     5394           
  Partials     1042     1042           

Flag	Coverage Δ
shard-1	41.59% <27.77%> (-0.01%)	⬇️
shard-2	35.50% <19.44%> (+0.04%)	⬆️
shard-3	31.72% <16.66%> (+0.09%)	⬆️
shard-4	31.62% <91.66%> (-0.09%)	⬇️
shard-5	31.05% <16.66%> (+0.68%)	⬆️
shard-6	43.17% <30.55%> (-0.29%)	⬇️

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

Files with missing lines	Coverage Δ
packages/loopover-engine/src/focus-manifest.ts	99.08% <100.00%> (+<0.01%)	⬆️
src/queue/processors.ts	95.50% <100.00%> (+<0.01%)	⬆️
src/review/visual/capture.ts	95.73% <100.00%> (+0.08%)	⬆️
src/review/visual/interaction-detection.ts	100.00% <100.00%> (ø)

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[cloudflare-workers-and-pages]

Deploying with    Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	loopover-ui	5b8d150	Commit Preview URL Branch Preview URL	Jul 19 2026, 11:43 PM

[loopover-orb]

Warning

⏸️ LoopOver review result - manual review recommended

_{Review updated: 2026-07-19 23:51:43 UTC}

10 files · 1 AI reviewer · 3 blockers · CI green · clean

⏸️ Suggested Action - Manual Review

Review summary
This PR adds an opt-in `review.visual.auto_detect_interactions` config that scans a PR's own CSS/SCSS/SASS/LESS diff for newly-added `:hover`/`:focus-visible` rules and auto-captures a hover-interaction GIF for the detected selector, composing with the existing hand-authored `interactions` list under the same 3-per-PR cap. The regex-based selector extraction (including the last-selector-in-comma-list heuristic) is correctly reasoned through and matches its extensive test coverage, the config plumbing (schema, overlay, parseVisualConfig, reviewConfigToJson) follows the existing boolean-flag pattern used by `bugAnalysis`/`actionsFallback`, and the feature is default-off/byte-identical when unused. The PR body does not link or close any specific open issue number, which this repo's contribution rules require of external PRs regardless of code quality.

Blockers

• The PR description does not reference or close a specific open, maintainer-authorized issue number (only internal `#auto-interaction-detection` tags and historical PR references) — per this repo's contribution policy every external PR must close/link an eligible open issue, so this cannot be marked merge-ready without that link.

Nits — 5 non-blocking

• src/queue/processors.ts: `changedCssFiles` is now built by mapping over every `unifiedFiles` entry on every visual capture, even when `review.visual.autoDetectInteractions` is off or the repo has no CSS changes — cheap, but could be gated behind the config flag to avoid the unconditional allocation.
• src/review/visual/interaction-detection.ts: `HOVER_SELECTOR_LINE_PATTERN` only matches when the selector and opening `{` are on the same diff line, so a multi-line selector list (`.a:hover,\n.b:hover {`) silently produces no match — worth a one-line doc caveat since it's a real, if narrow, gap in an otherwise well-scoped heuristic.
• overlayVisualConfig's OR-style merge (`override.autoDetectInteractions ? override : base`) means a per-repo config can never turn the feature back off once a global default sets it true — consistent with the existing pattern for other visual booleans in this file, but worth a comment noting it's intentional rather than an oversight.
• Confirm/attach the issue this PR is meant to close before merge, per repo policy.
• Consider gating the `changedCssFiles` mapping in src/queue/processors.ts:10387 behind `reviewVisualConfig.autoDetectInteractions` to skip the extra work on the common (feature-off) path.

Concerns raised — review before merging

• The PR description does not reference or close a specific open, maintainer-authorized issue number (only internal `#auto-interaction-detection` tags and historical PR references) — per this repo's contribution policy every external PR must close/link an eligible open issue, so this cannot be marked merge-ready without that link.
• No linked issue detected — If this PR is intended to solve an issue, link it explicitly in the PR body.
• Maintainer requires a linked issue — Link the relevant issue (for example Closes #123) before opening the PR.

📋 Copy for AI agents — paste into your coding agent

Fix the following blocker(s) from this PR review:

1. The PR description does not reference or close a specific open, maintainer-authorized issue number \(only internal \`\#auto-interaction-detection\` tags and historical PR references\) — per this repo's contribution policy every external PR must close/link an eligible open issue, so this cannot be marked merge-ready without that link.

2. No linked issue detected — If this PR is intended to solve an issue, link it explicitly in the PR body.

3. Maintainer requires a linked issue — Link the relevant issue (for example `Closes #123`) before opening the PR.

Decision drivers

• ❌ Code review — 3 blockers (1 reviewer)
• ❌ Gate result — Blocking (Repo-configured hard blocker found.)

Context & advisory signals — never blocks the verdict

Signal	Result	Evidence
Linked issue	⚠️ Missing	No linked issue or no-issue rationale found.
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (no linked issue context).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 28 registered-repo PR(s), 21 merged, 263 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 28 PR(s), 263 issue(s).
Improvement	✅ Minor	risk: clean · value: minor · LLM: moderate

Review context

• Author: JSONbored
• Role context: owner (maintainer lane)
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: Python, TypeScript, Ruby, Go, JavaScript, MDX, Shell, Solidity
• Official Gittensor activity: 28 PR(s), 263 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Start here: Treat this as maintainer-lane context rather than normal contributor-lane activity.
• Then work through the remaining 2 steps in the Signals table above.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

🧪 Chat with LoopOver

Ask LoopOver a question about this PR directly in a comment — grounded only in the same cached, public-safe facts shown above, never a new claim.

• @loopover ask <question> answers contribution-quality Q&A with source citations and freshness.
• @loopover chat <question> answers in natural prose from cached decision-pack facts via local inference (maintainer/collaborator; read-only).
• A plain-language @loopover mention with a real question is routed to the closest matching read-only command automatically — no exact syntax required.

Full command reference: https://loopover.ai/docs/loopover-commands

🧪 Experimental — new and may change.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by LoopOver, a quiet PR intelligence layer for OSS maintainers.

• Re-run LoopOver review