Home follows the runtime switcher: no OpenClaw gateway, no other runtimes' run health, activity, autonomy, anomalies or eval tile

[vivekchand]

Problem

On app.clawmetry.com with Codex selected, Home still showed things that belong to other runtimes:

• System Health listed OpenClaw's Gateway :18789. The hosted snapshot names it a bare Gateway, so the /openclaw/i filter from System Health follows the runtime switcher for all 32 runtimes: no OpenClaw cards elsewhere, no invented sub-agent success rate #5982 missed it.
• Run Health drew a claude_code row. The snapshot's healthTimeline had only claude_code: the newest 60 sessions on that node were all Claude Code, so Codex never got a row.
• 30-day activity, How independent is your agent?, Anomaly Detection, Reliability and the session quality tile counted every runtime on the machine.

Verified live before the fix, by decrypting the snapshot in the page: healthTimeline.runtimes = ["claude_code:30"], system carries ["Gateway","Running"], and the sh-services pill rendered Gateway :18789 under ?runtime=codex.

Fix

Card	Now
System Health services	drops OpenClaw*, a bare Gateway, and port 18789 off OpenClaw/NemoClaw
Run Health	?runtime= on /api/health-timeline (reads deeper, keeps that runtime's newest); page renders only that runtime's row. Daemon slice adds each runtime's own recent sessions via query_recent_sessions_by_runtime, memoised 120 s
30-day activity	/api/activity-heatmap?runtime= filters by session-id prefix and echoes runtime; page hides the card if the answer is not scoped
Independence	/api/autonomy?runtime= reads only that runtime's user turns; empty (never node-wide) when it has none. New snapshot slice autonomyByRuntime, memoised 300 s
Session quality tile	/api/evals/summary?runtime= + LocalStore.query_eval_summary(runtime=); echoes runtime
Anomaly Detection	keeps only the runtime's sessions; node-wide aggregate rows and baselines left out
Reliability	has no per-runtime form (daemon heartbeats + all errors), hidden under a runtime
Runtime switch	resets the loadAll 2 s coalesce so Home re-scopes immediately

The cloud-only leaks (the cm-cloud-runtimes chip row listing every runtime, the /api/system-health interceptor's Gateway, the autonomy/health-timeline/live-activity/diagnostics interceptors) are fixed in a companion clawmetry-cloud PR, which reads autonomyByRuntime from this release.

Verification

• tests/test_overview_runtime_scope.py: 16 tests (Flask routes, real DuckDB for autonomy and eval summary, fake store for the daemon slice, node harness running the real loadHealthTimeline). All 16 fail with the source changes reverse-applied, all pass with them.
• tests/test_system_health_runtime_scope.py, test_autonomy_v3_shape.py, test_health_timeline_route_gate.py, test_evals_route_gates.py, and a 209-test runtime/snapshot/overview/evals subset pass. Two pre-existing local failures reproduce on unchanged source (test_eval_runner.py x5: "no judge API key configured"; test_device_snapshot.py::test_empty_store_returns_valid_zero_payload reads a live alert) and are not touched here.

Product record: https://factory.8090.ai/project/b415065f-ab2f-4f53-8864-0c009fd098cb/requirements/8e389016-a9c8-4352-9121-72f0e361fdf6 (Runtime and Session Observability; FLYWHEEL §0a gate 2 and §1c runtime-filter rule).

🤖 Generated with Claude Code

[8090-software-factory]

✅ Drift Bot (ClawMetry): no drift detected

Drift Bot analyzed the changed files against this project's blueprints and requirements and found no drift.

[vivekchand]

visual-diff failing by design — this PR intentionally changes which tiles are visible under runtime filtering (tiles that belong to other runtimes are hidden when a runtime is selected). The baseline screenshots pre-date that behaviour change and no longer match.

To clear the check: regenerate the Playwright visual-diff baselines locally with npx playwright test --update-snapshots (or whatever the repo's update command is), commit the updated .png snapshots, and push. The diff itself is not a regression.

---

Generated by Claude Code

[vivekchand]

blocked on author decision — skipping (auto-mergeability sweep): PR is UNSTABLE because visual-diff (non-required, continue-on-error: true) shows pixel diffs from the intentional runtime-scoped UI changes. All 52 required checks pass including E2E Gate. The visual-diff failure is expected for this PR and cannot be resolved without reverting the UI change; author should review the diff screenshots and approve.

---

Generated by Claude Code

[8090-software-factory]

✅ Drift Bot (ClawMetry): no drift detected

Drift Bot analyzed the changed files against this project's blueprints and requirements and found no drift.

[8090-software-factory]

✅ Drift Bot (ClawMetry): no drift detected

Drift Bot analyzed the changed files against this project's blueprints and requirements and found no drift.

[github-actions]

Visual diff

Bot run failed before producing screenshots. Check the workflow logs.

This check is non-blocking.

[8090-software-factory]

✅ Drift Bot (ClawMetry): no drift detected

Drift Bot analyzed the changed files against this project's blueprints and requirements and found no drift.

[8090-software-factory]

✅ Drift Bot (ClawMetry): no drift detected

Drift Bot analyzed the changed files against this project's blueprints and requirements and found no drift.