perf(imports): break the github/client <-> db/repositories cycle taxing every test import

[JSONbored]

Summary

Follow-up to the TestD1Database template-clone fix: with schema replay gone, the suite's next-largest fixed cost is module import time (~505s aggregate per full run). Profiling found it concentrates in one place: the github/client ↔ db/repositories dependency cycle, which cost ~1.1s of cold import under vitest and taxed nearly every test file that touches the app — and it existed for trivially small reasons:

• db/repositories imported the whole @loopover/engine barrel (~420ms of the cycle) for exactly one tiny parser. → Engine gains a ./parse-pull-request-target-key subpath export (additive; the barrel still exports it), and repositories imports the subpath.
• gittensor/api (imported by db/repositories) imported github/client for ONE constant, PRODUCT_USER_AGENT. → The constant moves to a new leaf module src/github/user-agent.ts; client imports and re-exports it, so its ~8 other importers are unchanged.
• github/client statically imported db/repositories for two call sites (resolveRepoActionMode's freeze read; the suppressed-write audit hook). → Both become lazy await import(...) at the call site — the exact idiom processors.ts already uses — module-cached after first call.

Measured cold-import cost under vitest (same machine):

module	before	after
github/client	1,155ms	37ms
gittensor/api	1,119ms	11ms
db/repositories	1,106ms	771ms
queue/processors	2,677ms	1,978ms

db/repositories' remaining ~770ms is its own 8.3k lines + drizzle schema — reducible only by splitting the file itself (a #4013-style sequence of its own, deliberately out of scope here).

Behavior is unchanged: no cycle member had import-time side effects the other depended on, both lazy call sites are in already-async functions, and killing a real import cycle is a layering improvement in its own right.

Suite-level effect, measured with a back-to-back A/B on the identical tree (stash → baseline → pop → treatment, same machine conditions): aggregate import time 795s → 715s (−80s, −10%); wall time within noise on a loaded 10-core machine but the import saving concentrates in exactly the files CI's single validate-tests runner grinds through. The honest framing: this is a compounding follow-up to the template-clone fix (the big win), not a second 3× — plus it kills a real dependency cycle, which is worth having at any speed.

Validation

• npm run typecheck clean; npm run test:coverage (full, unsharded): 21,473 passed / 0 failed on the treatment run (the baseline run of the same A/B had 2 unrelated pre-existing flakes).
• Both lazy-import lines execute under existing tests (resolveRepoActionMode across the actuation suites; the audit hook via github-app's dry-run suppressed-write tests).

[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	c322a24	Commit Preview URL Branch Preview URL	Jul 24 2026, 07:29 PM

[superagent-security]

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

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.67%. Comparing base (61f6ec6) to head (28c7b39).
⚠️ Report is 2 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8548      +/-   ##
==========================================
+ Coverage   89.58%   93.67%   +4.08%     
==========================================
  Files          97      697     +600     
  Lines       22706    56845   +34139     
  Branches     3872    20164   +16292     
==========================================
+ Hits        20341    53248   +32907     
- Misses       2187     2613     +426     
- Partials      178      984     +806     

Flag	Coverage Δ
backend	93.67% <100.00%> (?)
control-plane	?
rees	?

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

Files with missing lines	Coverage Δ
src/db/repositories.ts	96.73% <ø> (ø)
src/github/client.ts	100.00% <100.00%> (ø)
src/github/user-agent.ts	100.00% <100.00%> (ø)
src/gittensor/api.ts	98.18% <ø> (ø)
src/review/rag-index.ts	93.86% <100.00%> (ø)
src/review/rag.ts	99.61% <100.00%> (ø)

... and 788 files with indirect coverage changes

[loopover-orb]

Important

🟪🟪🟪🟪🟪🟪🟪🟪🟪🟪🟪🟪

🔍 LoopOver is reviewing…

AI analysis is in progress. This comment will update when the review is complete.

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