fix(review): keep a fetched file distinguishable from an omitted one under thin budget

[shin-core]

What & why

Closes #8646.

review-grounding.ts's header guarantees (twice) that a file whose content is successfully fetched is "NEVER rendered as pure 'omitted' content again, no matter how large it grows" (#7465-class fix). fetchFullFileContents broke that: when cumulative used lands within MIN_SAMPLE_CHARS (400) of FILE_CONTENT_BUDGET (96,000), the computed share is too small, sampleHeadAndTail returns "" by design, and the fetched file gets pushed as { text: "", truncated: true } — the same empty placeholder a never-fetched file uses.

Consequence: the AI reviewer can't distinguish "fetched but budget ran out" from "never fetched" — both render identically, undermining the omission-tracking guarantee. test/unit/review-grounding.test.ts even asserted this broken behavior as correct.

The fix

Sample at the MIN_SAMPLE_CHARS floor for a genuinely-fetched file — sampleHeadAndTail(text, Math.max(share, MIN_SAMPLE_CHARS)) — so a fetched file always yields a minimal, non-empty head+tail with the omission marker. This is a small, bounded overrun of the overall budget only on the last thin file, not a per-file unbounded cost. A never-fetched file (text == null) still skips/omits, unchanged.

Tests

• Updated the existing 1MB-src/huge.ts fixture test (was asserting { text: "", truncated: true }) to assert the corrected behavior: the fetched file is truncated but non-empty, carries genuine content (z), and includes the omission marker — verifiably distinguishable from a never-fetched file's empty placeholder.
• The never-fetched → omitted path is still covered by the existing "nothing readable" test (requirement 2, unchanged).

Verified bug-catching: reverting the fix (back to the empty-placeholder degrade) fails the updated test.

Validation

• test/unit/review-grounding.test.ts: 50 tests pass; typecheck clean; 100% of changed lines/branches covered.
• Branched off current main, mergeable-clean.

[superagent-security]

🚨 Contributor flagged. Click here for more info: Superagent Dashboard

[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 90.59%. Comparing base (8f4a1f4) to head (0347dab).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8711      +/-   ##
==========================================
+ Coverage   90.56%   90.59%   +0.03%     
==========================================
  Files          96       97       +1     
  Lines       22490    22589      +99     
  Branches     3884     3921      +37     
==========================================
+ Hits        20367    20465      +98     
  Misses       1945     1945              
- Partials      178      179       +1     

Flag	Coverage Δ
backend	98.98% <100.00%> (?)

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

Files with missing lines	Coverage Δ
src/review/review-grounding.ts	98.98% <100.00%> (ø)

[loopover-orb]

Warning

⏸️ LoopOver review result - manual review recommended

_{Review updated: 2026-07-26 00:23:29 UTC}

2 files · 1 AI reviewer · no blockers · CI green · unstable

⏸️ Suggested Action - Manual Review

Review summary
The fix correctly moves the MIN_SAMPLE_CHARS floor into the sampleHeadAndTail call at fetchFullFileContents.ts (was: `sampleHeadAndTail(text, share)` degrading to `{text:"",truncated:true}` when share fell below MIN_SAMPLE_CHARS; now: `sampleHeadAndTail(text, Math.max(share, MIN_SAMPLE_CHARS))`), so a genuinely-fetched file always yields a non-empty head+tail sample instead of the same empty placeholder a never-fetched file produces. I traced the overrun bound claim: since `used` is only ever incremented past FILE_CONTENT_BUDGET on this one thin-share branch, the very next loop iteration's `used >= FILE_CONTENT_BUDGET` guard fires before any further fetch, so the overrun really is capped to a single file, as described. The updated test exercises the real call path (three files, last one thin) and asserts non-empty text + marker + real content — a faithful, non-fabricated regression test for the #8646 bug.

Nits — 3 non-blocking

• test/unit/review-grounding.test.ts: `expect(huge?.text).toContain("z")` is trivially true since the entire 1MB fixture is `z` repeated — it doesn't actually verify head/tail slicing correctness the way the earlier `sampleHeadAndTail` boundary tests do (e.g. asserting `startsWith`/`endsWith` on distinct head/tail markers) — consider a fixture with distinct head/tail sentinels for a stronger assertion.
• review-grounding.ts:307-311's comment references 'fix(orb): review-grounding.ts's 'never rendered as omitted' guarantee breaks under thin remaining budget #8646' inline in code — fine, but worth double-checking that issue number matches the one this PR is meant to close (per the description it does).
• Consider asserting in the test that `used` after this thin-file sample can exceed FILE_CONTENT_BUDGET by at most `MIN_SAMPLE_CHARS`, to lock in the 'small bounded overrun' invariant explicitly rather than only implicitly via the next-file skip behavior.

Flagged checks (non-blocking)

• Contributor trust — Contributor flagged for review

Decision drivers

• ✅ Code review — No blockers (1 reviewer)
• ✅ Gate result — Passing (No configured blocker found.)

Context & advisory signals — never blocks the verdict

Signal	Result	Evidence
Linked issue	✅ Linked	#8646
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 (1 linked issue).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 44 registered-repo PR(s), 24 merged, 0 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	shin-core; Gittensor profile; 44 PR(s), 0 issue(s).
Improvement	✅ Minor	risk: clean · value: minor · LLM: moderate

Linked issue satisfaction

Addressed
The diff changes fetchFullFileContents to sample at Math.max(share, MIN_SAMPLE_CHARS) so a genuinely-fetched file always gets a non-empty sample instead of the empty placeholder, and updates the exact test at lines 548-562 to assert the corrected, distinguishable behavior on the same 1MB fixture while leaving the never-fetched path untouched.

Review context

• Author: shin-core
• Role context: outside_contributor
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: not available
• Official Gittensor activity: 44 PR(s), 0 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Keep the PR focused and include validation evidence before maintainer review.

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

[loopover-orb]

LoopOver approves — the gate is satisfied and CI is green.