fix(miner): wire real dollar-cost tracking into the loop's budgetSpent

[JSONbored]

Summary

Follow-up to #5303/#5320: loop-cli.js's GovernorCapUsage.budgetSpent was never incremented (only turnsTaken/elapsedMs used real signals), so a capLimits.budget dollar-cost dimension could never trip from the loop's own spend.

Investigation found a real, already-flowing cost signal was being silently dropped: the Agent SDK's own result message declares total_cost_usd: number unconditionally (both SDKResultSuccess and SDKResultError carry it — a session is billed whether or not it succeeds), but agent-sdk-driver.ts only ever read num_turns.

Changes

• CodingAgentDriverResult gains an optional costUsd field.
• agent-sdk-driver.ts extracts the SDK's real total_cost_usd whenever a result message arrived, on both the success and the "did not complete successfully" paths (never fabricated on a thrown/no-result stream, where no billing signal exists at all).
• iterate-loop.ts sums it into a new totalCostUsd on IterateLoopResult, mirroring the existing totalTurnsUsed accumulator exactly (same 4 return points).
• attempt-cli.js surfaces it through options.onResult alongside totalTurnsUsed/iterationsUsed.
• loop-cli.js increments usage.budgetSpent with the real value before calling governor-state.js's saveCapUsage.

Documented, deliberate gap

The CLI-subprocess providers (claude-cli/codex-cli) report no cost signal today — their production argv never passes --output-format json, so there's nothing to parse cost from without a separate, larger change to how those providers invoke the real CLI binaries (a real scope decision, not a simple additive fix — I did not make that change here). budgetSpent is real for agent-sdk and an honest 0 (not fabricated) for the CLI-subprocess providers; a capLimits.budget dimension only ever meaningfully trips against agent-sdk spend today.

Test plan

• New regression: okDriverResult's real costUsd flows through runMinerAttempt → result.loopResult.totalCostUsd
• New regression: a mid-loop abandon (real iteration ran and was billed before the ceiling forced abandon) still reports real totalTurnsUsed/totalCostUsd
• New regression: loop-cli.js's budgetSpent reflects the real totalCostUsd from a submitted attempt
• New regression: a provider reporting no cost signal (mirrors claude-cli/codex-cli) leaves budgetSpent honestly at 0 across multiple real attempts
• coding-agent-driver-parity.test.ts: documented-divergence tests locking in "CLI driver never reports cost" vs. "Agent-SDK driver reports the real total_cost_usd"
• npx vitest run test/unit — 735 files / 14,582 tests pass
• npm run test:coverage — verified patch coverage of every new/touched line via lcov (including the mid-loop abandon path, previously untested at the root level)
• npx tsc --noEmit -p . --incremental false (fresh, no cache) — clean
• npm run test:engine-parity, npm run test:driver-parity, npm run test --workspace @jsonbored/gittensory-engine (520 node:test cases) — all pass
• npm run --prefix packages/gittensory-miner build — pass
• git diff --check — clean

[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	gittensory-ui	7c4221c	Commit Preview URL Branch Preview URL	Jul 12 2026, 02:51 PM

[codecov]

Codecov Report

❌ Patch coverage is 66.66667% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 94.42%. Comparing base (181cc0f) to head (7c4221c).
⚠️ Report is 5 commits behind head on main.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
...ckages/gittensory-engine/src/miner/iterate-loop.ts	60.00%	1 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #5356      +/-   ##
==========================================
+ Coverage   94.40%   94.42%   +0.01%     
==========================================
  Files         550      550              
  Lines       44117    44120       +3     
  Branches    14631    14633       +2     
==========================================
+ Hits        41650    41660      +10     
+ Misses       1792     1785       -7     
  Partials      675      675              

Flag	Coverage Δ
shard-1	44.00% <0.00%> (-0.01%)	⬇️
shard-2	34.66% <66.66%> (+0.02%)	⬆️
shard-3	31.60% <0.00%> (-0.01%)	⬇️
shard-4	31.41% <0.00%> (-0.01%)	⬇️
shard-5	33.15% <0.00%> (-0.01%)	⬇️
shard-6	43.35% <0.00%> (-0.01%)	⬇️

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

Files with missing lines	Coverage Δ
...es/gittensory-engine/src/miner/agent-sdk-driver.ts	100.00% <100.00%> (ø)
...gittensory-engine/src/miner/coding-agent-driver.ts	100.00% <ø> (ø)
packages/gittensory-miner/lib/attempt-cli.js	96.55% <ø> (ø)
packages/gittensory-miner/lib/loop-cli.js	93.23% <ø> (ø)
...ckages/gittensory-engine/src/miner/iterate-loop.ts	64.28% <60.00%> (+10.58%)	⬆️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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

[gittensory-orb]

Caution

🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥🟥

🛑 Gittensory review result - fixes required

_{Review updated: 2026-07-12 15:24:12 UTC}

10 files · 1 AI reviewer · 2 blockers · readiness 93/100 · CI failing · unstable

🛑 Suggested Action - Manual Review

• 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
This diff wires the Agent SDK's real total_cost_usd through CodingAgentDriverResult -> IterateLoopResult.totalCostUsd -> attempt-cli's onResult -> loop-cli's usage.budgetSpent accumulator, mirroring the existing totalTurnsUsed pattern exactly at all 4 iterate-loop return points. The change is correct and honest about scope: costUsd is genuinely absent (undefined, not fabricated) for CLI-subprocess providers and for the thrown/no-result driver path, and present on both success and failure result-message branches per the SDK's actual type contract. Tests cover the accumulation across handoff/abandon paths and lock in the CLI-driver-reports-no-cost divergence.

Nits — 5 non-blocking

• packages/gittensory-miner/lib/loop-cli.js:353 — verify governor-state.js's saveCapUsage/capLimits.budget comparison logic actually treats a fractional budgetSpent correctly (e.g. no integer/rounding assumption), since this is the first place a non-integer usage field flows through that path.
• codecov/patch flags 66.66% patch coverage — confirm the uncovered lines aren't a real gap in loop-cli.js's new budgetSpent line or agent-sdk-driver.ts's costUsd branches, since only 2 of the ~7 changed source files have visible new tests in this diff excerpt.
• Add a direct unit test for agent-sdk-driver.ts asserting costUsd is read from total_cost_usd on the `is_error: true` / non-success subtype path too (currently only the success-subtype SDK test at test/contract/coding-agent-driver-parity.test.ts:242 exercises this field).
• nit: `packages/gittensory-engine/src/miner/agent-sdk-driver.ts:174` now promises cost is preserved on the non-success result path, but the visible tests only assert the success result frame; add a regression where `subtype !== "success"` or `is_error === true` still returns `costUsd`.
• nit: `packages/gittensory-engine/src/miner/iterate-loop.ts:277` sums raw provider numbers without rejecting `NaN`, `Infinity`, or negative values; if the SDK contract is the reason this is safe, leave that explicit or clamp to finite non-negative values before adding.

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.

CI checks failing

• codecov/patch — 66.66% of diff hit (target 99.00%)

Signal	Result	Evidence
Code review	❌ 2 blockers	1 reviewer
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: 44 registered-repo PR(s), 36 merged, 476 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 44 PR(s), 476 issue(s).
Gate result	❌ Blocking	Repo-configured hard blocker found.
Improvement	✅ Minor	risk: clean · value: minor — Code changes are accompanied by test evidence.

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: not available
• Official Gittensor activity: 44 PR(s), 476 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Treat this as maintainer-lane context rather than normal contributor-lane activity.
• Explain no-issue PR.
• 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.

[BETA] Chat with Gittensory

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

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

Full command reference: https://gittensory.aethereal.dev/docs/gittensory-commands

_{🟩 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