feat(ops): home account-spend tiles (anthropic-cost-integration Phase 2)

[silversurfer562]

Summary

Wires Phase 1's anthropic_cost.fetch_summary into the home page route. Adds three account-spend KPI tiles ("Today (account)", "7d (account)", "MTD (account)") gated on cost_summary availability, with a "Connect billing" CTA when no admin key is configured and a friendly "Key rejected" notice on auth-failed.

Implements spec docs/specs/anthropic-cost-integration/tasks.md Phase 2.

What's in the diff

• src/attune/ops/routes/dashboard.py: home() now calls fetch_summary inside a defensive try/except, passes cost_summary and cost_error to the template; ?refresh=1 forces a cache bypass
• src/attune/ops/templates/home.html: new <section class="kpi-grid kpi-cluster-account"> block after the existing telemetry cluster, with three render paths (valid summary → tiles, no_key → CTA, auth_failed → notice, other errors → silent fall-through)
• src/attune/ops/static/css/main.css: .kpi-cluster-account styles, including the thin left-edge accent border and color-coded variants for .kpi-cta and .kpi-notice
• tests/unit/ops/test_anthropic_cost_home_integration.py: six new tests covering all four spec-2.4 scenarios plus the default-refresh-false and exception-swallow paths

Trade-offs / non-obvious choices

• CTA copy: shipped option 1 from the open-question list ("Set up cost reporting" linking to docs/reference/) rather than option 2 ("Coming in Phase 4 …"). Cleaner, doesn't reference a future state. Easily revisited when Phase 4 lands.
• Visual distinction: left-edge accent border (3px solid var(--accent)) rather than the spec's "right-edge" suggestion. Reads more naturally with the existing tile flex direction; functionally equivalent for the "tell at a glance which data source" goal.
• Defensive try/except: catches everything (except Exception with # noqa: BLE001) because a billing-fetch surprise must not block the dashboard render. Surfaces in DEBUG logs only — verified by the silent-degradation test which mocks fetch_summary to raise RuntimeError.
• Patch target in tests: attune.ops.anthropic_cost.fetch_summary (source module) rather than attune.ops.routes.dashboard.anthropic_cost.fetch_summary (consumer site). The latter triggers mock.patch's _importer to attempt module-walk on a function attribute. Source-module patching works because dashboard.anthropic_cost IS the same module object as the source; replacing the attribute on the source flows through.

Caught in flight

The formatter stripped the from attune.ops import anthropic_cost import after my initial import-only edit (before the usage line was added), per the existing CLAUDE.md lesson about formatters stripping unused-at-the-moment imports. Re-added with the usage line in place; runtime confirmed dashboard.anthropic_cost resolves to the module. Worth a session lesson: even brief edit-formatter cycles trigger this, not just multi-minute pauses.

Test plan

• 6 new tests in test_anthropic_cost_home_integration.py pass
• 50 tests in test_home.py + test_anthropic_cost.py + new file pass (no regressions)
• Ruff clean on Python files
• Pre-commit hooks pass (black, ruff, bandit, detect-secrets, markdown checks, help-template freshness)
• CI green on all platforms
• Manual verification: dashboard launches with no admin key → CTA tile renders; with valid CostSummary mock → three tiles populated

Related

• #432 — Phase 1 backend client this PR consumes
• #431 — original spec
• Next: Phase 3 (/telemetry per-day panel), Phase 4 (attune ops setup-billing CLI), Phase 5 (polish + docs)

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
website	Ready	Preview, Comment	May 19, 2026 5:29pm

[github-actions]

🔒 Security Scan Results

✅ Status: PASSED - No blocking issues

Summary

Severity	Count	Action
🔴 CRITICAL	0	❌ BLOCKS PR
🟡 MEDIUM	0	⚠️ Review recommended
🔵 LOW	0	ℹ️ Informational

Total Findings: 0

---

---

🛠️ Need Help?

If findings are false positives:

1. Add clarifying comments in code (e.g., # Security Note: Test data only)
2. Request security review: Add security-review label
3. Security team will evaluate and add security-approved label if safe

For emergency hotfixes:

1. Add hotfix label to bypass blocking
2. Create follow-up ticket to address findings
3. Security team will review post-deployment

Scanner Accuracy: ~82% (Industry-leading!)

_{Powered by Attune AI Security Scanner | Documentation}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[silversurfer562]

Rebased onto current main (which now includes #439 telemetry fix and #440 docs). The earlier diff against origin/main showed -300 lines because the branch was behind main — what looked like a bundled removal of _track_sdk_run_telemetry and friends was actually just absence of #439's additions on this branch.

After rebase, the diff is exactly 4 files (+255 / -1):

• src/attune/ops/routes/dashboard.py — home route wires anthropic_cost.fetch_summary
• src/attune/ops/templates/home.html — 3 KPI tiles + no_key CTA + auth_failed notice
• src/attune/ops/static/css/main.css — .kpi-cluster-account styling
• tests/unit/ops/test_anthropic_cost_home_integration.py — 6 integration tests

40/40 tests pass post-rebase. PR is now cleanly aligned with what the title advertises.