feat(tasks): add runtime_adapter label to run token metrics

[tatoalo]

Problem

The terminal-run token metrics (tasks_run_total_tokens, tasks_run_tokens_total) carry an rtk_enabled cohort label but no runtime adapter, so the rtk cohort panels on the cloud-background-agents-runs dashboard can't separate Claude runs (where rtk is applied deterministically via a PreToolUse rewrite hook) from Codex runs (where rtk adoption is instruction-level, added in PostHog/code#3491). Without the split there's no way to see whether the Codex integration actually moves tokens per run.

Why: rtk command-output compression is being extended to cloud Codex runs, and the per-adapter cohort is the graph that validates (or falsifies) that change.

Changes

• record_run_token_usage now takes the run's runtime_adapter (read from TaskRun.state, where run creation already persists it) and labels both token metrics with it, bounded to claude/codex/other/unknown to cap cardinality.
• Corrected the sandbox Dockerfile.sandbox-base comment that claimed "the agent" auto-detects rtk on PATH — that is true only for Claude; Codex has no command-rewrite channel and is instructed instead.

Companion dashboard change (splitting the rtk cohort panels by this label) goes to the grafana-dashboards repo.

[trunk-io]

     

View Full Report ↗︎ ⋅ Docs

[greptile-apps]

_{Reviews (1): Last reviewed commit: "feat(tasks): add runtime_adapter label t..." | Re-trigger Greptile}

[stamphog]

Note

🤖 stamphog reviewed 41b1c55bb3055ff30360f262564c5aafab354c46 — verdict: REFUSED

Gate verdict is DENIED (deny-list hit on toolchain/infra plus T2-never tier), which forces routing to a human regardless of how contained the diff looks; the diff itself (metrics label + Dockerfile comment fix) appears low-risk and Greptile left a 👍 with no concerns.

• 👍 on the PR from greptile-apps[bot], hex-security-app[bot].
• Gates DENIED: deny-list matched deps_toolchain and infra_cicd (Dockerfile.sandbox-base change), and the PR was classified as T2-never (never auto-approved regardless of size).

Gate mechanics and policy version

Gate		Result
prerequisites	✓	all clear
deny-list	✗	matches: deps_toolchain, infra_cicd
size	✓	22L, 3F substantive, 24L/4F incl. docs/generated/snapshots — within ceiling
tier	✗	classified as T2-never: T2-never (24L, 4F, single-area, feat)
stamphog 2.0.0b3		.stamphog/policy.yml @ 9267ac7 · reviewed head 41b1c55

[github-actions]

🤖 CI report

⚠️ Backend coverage — 28.0% of changed backend lines covered — 5 uncovered

🧪 Backend test coverage

Patch coverage — changed backend lines (products + core): ██████░░░░░░░░░░░░░░ 28.0% (2 / 7)

File	Patch	Uncovered changed lines
products/tasks/backend/temporal/process_task/activities/tests/test_update_task_run_status.py	0.0%	120
products/tasks/backend/temporal/process_task/activities/update_task_run_status.py	0.0%	107
products/tasks/backend/temporal/metrics.py	40.0%	79–81

🤖 Agents: add a test covering the lines above, or note why under "How did you test this code?". Machine-readable gap list: the patch-coverage artifact on this run (gh run download 29490050673 -n patch-coverage), or the coverage-data block at the end of this comment.

Per-product line coverage (touched products)

Product	Coverage	Lines
tasks	█████████████░░░░░░░ 67.4%	25,787 / 38,257

Report-only. Patch coverage = changed backend lines covered vs origin/master. Sorted lowest first.
Known gaps: lines covered only by Temporal tests show as uncovered; core line numbers may drift if master changed the same file.

[deployment-status-posthog]

Deploy status

Environment	Status	Deployed At	Workflow
dev	✅ Deployed	2026-07-16 15:52 UTC	Run
prod-us	✅ Deployed	2026-07-16 16:08 UTC	Run
prod-eu	✅ Deployed	2026-07-16 16:07 UTC	Run