feat(miner): extract and persist real coding-agent token usage

[JSONbored]

Summary

• Agent SDK driver now reads real input_tokens/output_tokens off the SDK's own result message (usage field) instead of never reporting tokens
• CLI-subprocess driver ports src/selfhost/ai.ts's extractCliUsage to scan CLI JSON/JSONL stdout for input/output/total token keys (top-level and nested usage/token_usage/etc.), same convention already used there for cost
• iterate-loop.ts sums each iteration's real tokensUsed into finalMeterTotals.tokens (previously hardcoded to 0)
• attempt-cli.js surfaces the accumulated total as totalTokensUsed on the attempt result and persists it on the attempt_outcome_summary ledger event, alongside the existing costUsd/provider fields feeding the Grafana miner-usage dashboard

Absent (not zero) is preserved as the "no signal reported" case throughout, mirroring costUsd's existing honest-absence convention.

Closes #5653

Test plan

• packages/gittensory-engine node:test suite (537/537 pass)
• Root vitest mirrors for both drivers + miner-attempt-runner/miner-attempt-cli (136/136 pass)
• npm run test:ci green
• npm run test:coverage (unsharded) — 100% branch coverage on every new/changed line, verified directly against lcov.info
• npm run typecheck, npm run build:miner, npm run test:miner-pack, npm audit --audit-level=moderate all 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	loopover-ui	1fbf677	Commit Preview URL Branch Preview URL	Jul 13 2026, 09:40 PM

[loopover-orb]

Warning

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏸️ LoopOver review result - manual review recommended

_{Review updated: 2026-07-13 22:17:23 UTC}

11 files · 1 AI reviewer · no blockers · readiness 100/100 · CI green · clean

⏸️ Suggested Action - Manual Review

• Touches a guarded path — held for manual review: This PR changes guardrail-protected path(s): packages/gittensory-engine/src/miner/agent-sdk-driver.ts (matched packages/gittensory-engine/**), packages/gittensory-engine/src/miner/cli-subprocess-driver.ts (matched packages/gittensory-engine/**), packages/gittensory-engine/src/miner/coding-agent-driver.ts (matched packages/gittensory-engine/**), packages/gittensory-engine/src/miner/iterate-loop.ts (matched packages/gittensory-engine/**), packages/gittensory-engine/test/agent-sdk-driver.test.ts (matched packages/gittensory-engine/**).

Review summary
This PR wires real token-usage extraction through both coding-agent drivers (SDK result-message usage field, CLI stdout JSON/JSONL scan) and threads the totals through iterate-loop's meter accumulation into the attempt_outcome_summary ledger event, following the exact same honest-absence convention already established for costUsd. The implementation is careful: totalTokensFromUsage/tokensFromResultMessage both return undefined rather than fabricating 0, the CLI extractor's key-list/nested-object approach mirrors src/selfhost/ai.ts's existing extractCliUsage as documented, and iterate-loop.ts's accumulation now correctly sums driverResult.tokensUsed instead of hardcoding 0. Tests are updated across every call site to supply finalMeterTotals so the previously-passing mocked loopResult objects don't silently regress, and new coverage is added for both drivers' undefined/malformed/partial-usage branches. Closes #5653 as claimed and stays narrowly scoped to that issue.

Nits — 7 non-blocking

• packages/gittensory-engine/src/miner/cli-subprocess-driver.ts: the CLI JSONL scan takes `Math.max` per key across every line/nested-object seen (mergeCliUsage's maxNumber), which is fine for a monotonically-increasing cumulative counter but could silently pick a stale high-water value if a CLI ever resets/re-reports a lower running total mid-stream — worth a one-line comment confirming this is intentional (cumulative, not last-write-wins) since it's not obvious from the diff alone.
• totalTokensFromUsage's 'prefer total_tokens over input+output sum' rule (packages/gittensory-engine/src/miner/cli-subprocess-driver.ts) is asymmetric with mergeCliUsage's own max-across-keys behavior for cost — consider a one-line note on why tokens dedupe against double-counting but cost does not need the same treatment.
• packages/gittensory-miner/lib/attempt-cli.js is now ~504 lines per the size-smell flag; not a defect, but worth watching if more usage fields get bolted on here.
• Consider extracting the shared CLI-usage key-list/merge logic between src/selfhost/ai.ts and cli-subprocess-driver.ts into one module now that both duplicate the same convention twice, even if the no-src-import rule is intentional — a follow-up issue reference would help future readers understand why the duplication is permanent.
• In iterate-loop.ts, a short inline comment at the `tokens: driverResult.tokensUsed ?? 0` line confirming this differs from costUsd's undefined-preserving contract (tokens is zeroed per-iteration for accumulation, not left undefined) would preempt confusion since the two fields now have subtly different absence semantics at this layer.
• PR author also opened the linked issue — Link an issue that was opened by a different contributor, or provide a rationale for why this self-authored issue represents genuine discovery work.
• Touches a guarded path — held for manual review — A maintainer must review and merge this change.

Signal	Result	Evidence
Code review	✅ No blockers	1 reviewer
Linked issue	✅ Linked	#5653
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: 45 registered-repo PR(s), 36 merged, 333 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 45 PR(s), 333 issue(s).
Gate result	⚠️ Not blocking	Advisory; not blocking this PR.
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: Python, TypeScript, Ruby, Go, JavaScript, MDX, Shell, Solidity
• Official Gittensor activity: 45 PR(s), 333 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Treat this as maintainer-lane context rather than normal contributor-lane activity.

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 LoopOver, a quiet PR intelligence layer for OSS maintainers.

• Re-run LoopOver review

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.03%. Comparing base (0dfc40a) to head (1fbf677).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5658   +/-   ##
=======================================
  Coverage   95.03%   95.03%           
=======================================
  Files         577      577           
  Lines       45920    45946   +26     
  Branches    14695    14708   +13     
=======================================
+ Hits        43639    43665   +26     
  Misses       1528     1528           
  Partials      753      753           

Flag	Coverage Δ
shard-1	43.98% <81.08%> (+0.02%)	⬆️
shard-2	35.67% <29.72%> (-0.01%)	⬇️
shard-3	32.18% <8.10%> (-0.02%)	⬇️
shard-4	32.57% <27.02%> (-0.01%)	⬇️
shard-5	31.78% <37.83%> (+<0.01%)	⬆️
shard-6	44.44% <8.10%> (-0.02%)	⬇️

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)
...ttensory-engine/src/miner/cli-subprocess-driver.ts	100.00% <100.00%> (ø)
...gittensory-engine/src/miner/coding-agent-driver.ts	100.00% <ø> (ø)
...ckages/gittensory-engine/src/miner/iterate-loop.ts	74.32% <100.00%> (+0.35%)	⬆️
packages/gittensory-miner/lib/attempt-cli.js	96.96% <ø> (ø)

🚀 New features to boost your workflow:

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