fix(engine): report below-baseline calibration delta as a signed negative

[glorydavid03023]

Summary

computePhase7CalibrationLoop (packages/gittensory-engine/src/phase7-calibration-loop.ts) computed deltaFromBaseline — the signed deviation of the combined calibration accuracy from the documented 62% baseline — by rounding through roundScore, which clamps its input to [0, 1]:

const deltaFromBaseline =
  combinedAccuracy === null ? null : roundScore(combinedAccuracy - DOCUMENTED_CALIBRATION_BASELINE);

combinedAccuracy is in [0, 1] and the baseline is 0.62, so the true delta ranges over [-0.62, +0.38]. Every below-baseline run (a calibration regression) produced a negative delta that roundScore flattened to 0, so a regression was reported as "on baseline / no deviation."

deltaFromBaseline is a signed field (number | null), the module tracks accuracy "against the documented 62% baseline," and the audit renderer prints it as a signed magnitude (${(deltaFromBaseline * 100).toFixed(2)} percentage points) — so surfacing below-baseline regressions is exactly its purpose. The sibling combinedAccuracy is a genuine [0, 1] score and correctly uses roundScore; only the signed delta was mis-clamped.

The fix rounds the signed difference without the [0, 1] clamp via a dedicated roundDelta helper.

Example: two sources at 0.5 accuracy → combined 0.5 (0.12 below baseline) now yields deltaFromBaseline === -0.12 (rendered -12.00 percentage points) instead of 0.

No linked issue: a small, self-contained correctness fix in pure engine logic; per this repo's linkedIssuePolicy: preferred, the defect is described in full above.

Scope

• The PR title follows type(scope): short summary Conventional Commit format.
• This PR is focused and does not mix unrelated backend, UI, MCP, docs, dependency, and deploy changes.
• This follows CONTRIBUTING.md and does not reintroduce GitHub Pages, VitePress, site/, or CNAME.
• I linked an issue, or this is small enough that the summary explains why an issue is not needed.

Validation

• git diff --check
• npm run typecheck (green)
• npm run build:miner builds + typechecks the engine package with the change (green)
• New behavior has regression tests covering below / at / above baseline, added in both conventions: a root test/unit/phase7-calibration-delta.test.ts (vitest, run by test:coverage) and the engine packages/gittensory-engine/test/phase7-calibration-loop.test.ts (node:test).
• Verified empirically: below-baseline combined accuracy 0.5 now reports deltaFromBaseline === -0.12 (was 0).

If any required check was skipped, explain why:

• The full npm run test:ci was not run in my local dev environment (several self-host suites require Linux/bash/Docker/Postgres and fail only on Windows). This change is confined to packages/gittensory-engine (a pure function, no src/ app, API/OpenAPI, MCP, binding, DB, or migration surface), so no generated artifacts change; typecheck, the engine build, and the new regression tests are green, and the remaining CI runs on this PR.

Safety

• No secrets, wallets, hotkeys, coldkeys, PATs, private keys, trust scores, private rankings, or maintainer evidence are exposed.
• Public GitHub text stays sanitized and low-noise; no compensation/optimization implications.
• No auth/cookie/CORS/GitHub App/Cloudflare/session changes.
• No API/OpenAPI/MCP behavior changes.

UI Evidence

N/A - pure backend engine logic; no UI, frontend, docs, or extension change.

[superagent-security]

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

[loopover-orb]

Caution

🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥

🛑 Gittensory review result - reject/close recommended

_{Review updated: 2026-07-07 05:47:46 UTC}

3 files · 1 AI reviewer · 2 blockers · readiness 100/100 · CI pending · blocked

🛑 Suggested Action - Reject/Close

• 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.

Review summary
The diff fixes a real clamping bug: `deltaFromBaseline` in `phase7-calibration-loop.ts` previously ran through `roundScore`, which clamps to `[0,1]`, so any below-baseline (negative) delta was flattened to `0`, hiding calibration regressions. The new `roundDelta` helper (lines ~121-124) rounds without clamping and is correctly swapped in at the `deltaFromBaseline` computation (~line 473), leaving the unrelated `combinedAccuracy` clamp (which is a genuine `[0,1]` score) untouched. The added unit and integration tests exercise the real `computePhase7CalibrationLoop` path with actual `prOutcome`/`historicalReplay` inputs (not fabricated payloads) and assert the negative, positive, and exact-baseline-zero cases, matching the rendered Markdown output too.

Nits — 4 non-blocking

• The `roundDelta` doc comment (phase7-calibration-loop.ts ~118-121) is a bit long for a one-line helper; could be trimmed to a single sentence per repo comment conventions.
• The new top-level `test/unit/phase7-calibration-delta.test.ts` duplicates the same three scenarios already added to `packages/gittensory-engine/test/phase7-calibration-loop.test.ts` (only the negative case is duplicated exactly) — consider keeping the coverage in one location to avoid divergence over time.
• Consider consolidating the two new test files (unit/ and package test/) since they test the same function with overlapping scenarios.
• Since `roundDelta` and `roundScore` are now easy to confuse, a brief comment at `roundScore`'s definition noting 'do not use for signed deltas — see roundDelta' would help future readers.

Why this is blocked

• 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.

Signal	Result	Evidence
Code review	❌ 2 blockers	1 reviewer
Linked issue	✅ No-issue rationale	PR body explains why no issue is linked.
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: 308 registered-repo PR(s), 198 merged, 15 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	glorydavid03023; Gittensor profile; 308 PR(s), 15 issue(s).
Gate result	❌ Blocking	Repo-configured hard blocker found.

Review context

• Author: glorydavid03023
• 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: 308 PR(s), 15 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Link the issue being solved, or explicitly explain why this is a no-issue PR.

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.

_{🟩 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 Gittensory, a quiet PR intelligence layer for OSS maintainers.

• Re-run Gittensory review

[loopover-orb]

Gittensory is closing this pull request on the maintainer's behalf (No linked issue detected; Maintainer requires a linked issue). This is an automated maintenance action — to pursue this change, please open a new pull request with the issues resolved. Closed PRs may be analyzed later to improve review accuracy, but they are not automatically reopened or re-reviewed.

[codecov]

Codecov Report

❌ Patch coverage is 50.00000% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 93.77%. Comparing base (bfe55f6) to head (895a0dd).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
...s/gittensory-engine/src/phase7-calibration-loop.ts	50.00%	0 Missing and 1 partial ⚠️

❌ Your patch status has failed because the patch coverage (50.00%) is below the target coverage (99.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3890      +/-   ##
==========================================
+ Coverage   93.57%   93.77%   +0.19%     
==========================================
  Files         340      340              
  Lines       33516    33517       +1     
  Branches    12262    12262              
==========================================
+ Hits        31364    31431      +67     
+ Misses       1528     1389     -139     
- Partials      624      697      +73     

Files with missing lines	Coverage Δ
...s/gittensory-engine/src/phase7-calibration-loop.ts	36.70% <50.00%> (+35.63%)	⬆️

🚀 New features to boost your workflow:

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