feat(miner): add a Grafana dashboard for coding-agent-driver usage, cost, and attempt outcomes

[JSONbored]

Summary

Depends on #5637 (merged) for the provider/cost_usd/tokens_used columns this dashboard reads.

• grafana/dashboards/miner-usage.json: per-provider (claude-cli/codex-cli/agent-sdk) attempt success/fail counts, total cost, and total tokens, all in one dashboard behind a $provider template variable — not one dashboard per provider, per the issue's own reasoning (AMS's local ledger has far lower data volume than ORB's three cloud reviewers).
• Every panel queries the AMS Attempt Log SQLite datasource's attempt_outcome_summary rows only (event_type = 'attempt_outcome_summary') — never Prometheus, never the live ledger.
• Cross-references ai-usage.json (ORB) in both directions via a top-level links entry — the only existing dashboard-to-dashboard link convention already in this repo (resource-hub.json's own links array). ai-usage.json's other panels/variables/datasources are untouched.
• tokens_used reads 0/NULL honestly today (documented on the panel itself) — no coding-agent driver reports real token usage yet (Decide fate of attempt-metering.ts: wire real per-attempt budget enforcement, or remove it #5395); the dashboard needs no changes once one does.

Test plan

• npm run selfhost:validate-observability (repo's own dashboard-JSON structural validator)
• New test/unit/selfhost-grafana-miner-usage-dashboard.test.ts (9 cases), mirroring selfhost-grafana-ai-usage-dashboard.test.ts's real-sqlite3-execution rigor: structural checks (uid/title/tags, single $provider variable, SQLite-only datasource, cross-reference links both directions, additive-only ai-usage.json diff), an invariant test asserting no panel ever mixes AMS's attempt_log_events table with ORB's ai_usage_events/Prometheus metric names, and a seeded-database test that runs every panel's actual query and verifies $provider narrows correctly (including that a non-attempt_outcome_summary row on the same attempt_id, with provider IS NULL, never leaks into any aggregate or the recent-outcomes table)
• npm run typecheck
• Full observability + AMS test sweep (132 tests across 14 files) green
• git diff --check

Closes #5185

[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 95.03%. Comparing base (941c300) to head (9ed1ea3).
⚠️ Report is 2 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5646   +/-   ##
=======================================
  Coverage   95.03%   95.03%           
=======================================
  Files         577      577           
  Lines       45920    45920           
  Branches    14695    14695           
=======================================
  Hits        43640    43640           
  Misses       1528     1528           
  Partials      752      752           

Flag	Coverage Δ
shard-1	43.49% <ø> (-0.43%)	⬇️
shard-2	35.81% <ø> (+0.13%)	⬆️
shard-3	32.13% <ø> (-0.06%)	⬇️
shard-4	30.81% <ø> (-1.77%)	⬇️
shard-5	32.53% <ø> (+0.75%)	⬆️
shard-6	44.84% <ø> (+0.37%)	⬆️

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

🚀 New features to boost your workflow:

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

[loopover-orb]

Warning

🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨🟨

⏸️ LoopOver review result - manual review recommended

_{Review updated: 2026-07-13 19:28:55 UTC}

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

⏸️ Suggested Action - Manual Review

• AI review could not be completed: The dual-model AI review did not return a usable verdict for this change.

Review summary
This PR adds a new Grafana dashboard (miner-usage.json) for AMS coding-agent-driver usage/cost/outcome metrics, wires a bidirectional links cross-reference into the existing ai-usage.json (ORB) dashboard, and ships a thorough unit test suite that runs the panel queries against a real seeded sqlite3 database. The queries consistently scope by event_type='attempt_outcome_summary', the $provider template variable, and the selected time window, and the tests explicitly verify that a non-outcome-summary row with a NULL provider on the same attempt_id never leaks into any aggregate — a real edge case, not a fabricated one. The change is additive-only to ai-usage.json (only a links entry added) and depends on the already-merged #5637 for the provider/cost_usd/tokens_used columns it reads.

Nits — 6 non-blocking

• Panel 3's 'Succeeded' definition treats every action_class other than 'attempt_submitted' as failure, including 'attempt_governed', which may not always represent a true failure outcome — worth a sentence in the panel description clarifying that classification choice is intentional and not just a placeholder.
• grafana/dashboards/miner-usage.json panel 7's sortBy ('total', desc) doesn't provide a secondary tiebreaker, so provider rows with equal totals (e.g. codex-cli and agent-sdk in the test fixture) have no guaranteed stable order — low-impact for a dashboard table but worth noting for reproducible screenshots/docs.
• Consider adding a brief comment/description on the $provider template variable itself noting that its 'All' sentinel ('$__all') is compared literally against the query's ${provider:sqlstring} filter, mirroring whatever convention ai-usage.json already established, so future contributors don't accidentally break the All-selection path when editing either dashboard.
• If a future migration adds more action_class values (e.g. a new terminal state), the binary succeeded/failed split in panels 3 and 7 will need re-auditing — worth a short code comment pointing at the AMS ledger's action_class enum as the source of truth.
• 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.
• AI review could not be completed — The gate is held for a human reviewer rather than passed automatically; it re-evaluates on the next update.

Signal	Result	Evidence
Code review	✅ No blockers	1 reviewer
Linked issue	✅ Linked	#5185
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), 36 merged, 320 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	JSONbored; Gittensor profile; 44 PR(s), 320 issue(s).
Gate result	⚠️ Not blocking	Advisory; not blocking this PR.
Improvement	⚠️ ℹ️ Insufficient signal	risk: clean · value: insufficient-signal — Nothing measurable for the structural-improvement analyzers on this PR (e.g. no code files changed). LLM value judgment: moderate — The PR is a well-scoped, thoroughly tested observability addition that directly implements open issue #5185's dashboard requirement without touching any review/gate logic, so its value is real but confined to operator tooling rather than the core engine.

Linked issue satisfaction

Addressed
The PR adds grafana/dashboards/miner-usage.json with a single $provider template variable covering all three coding-agent providers, panels for success/fail counts, cost, and tokens queried directly from the AMS SQLite datasource, plus bidirectional cross-reference links matching ORB's convention and an additive-only change to ai-usage.json.

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), 320 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