Skip to content

fleet telemetry adoption: workers complete tasks without pool/model/effort/cost fields (1 of 846 post-#1582 completions) #4919

Open

Description

@dezren39

Problem

code#1582 (fleet telemetry) CLOSED COMPLETED 2026-11-02T22:18:19Z: repo-queue.py complete now supports optional pool / model / effort / cost_usd on the completion result record, and the meta-reviewer's data pull can read them from agent.tasks.result.

Adoption is ~zero. Measured by the pool meta-reviewer (issue #1340) for the window 2026-11-02T13:12Z–2026-11-03T13:12Z, from real queue rows:

So the per-pool / per-effort scorecard the meta-reviewer is charged with (model-up + effort-down vs model-down + effort-up; where max effort buys nothing over high) remains unrecordable for an eighth consecutive window — now because workers do not pass the fields, not because the fields do not exist.

What is needed

  • Lane / worker completion discipline: every complete call from a fleet session includes the telemetry fields from its own resolved route (pool, model, effort; cost_usd where the route exposes one, else null) — via --telemetry or the top level of --result. Recorded from the resolved session, never inferred from route names (meta-reviewer guideline 2026-10-27).
  • The dispatcher record change from fleet telemetry: record pool/model/effort/cost on dispatch + queue completion records (meta-reviewer #1340 follow-up) #1582 already landed in the dispatcher instruction; this issue is about the queue-completion side's adoption by the lane shifts / queue workers.
  • Suggested check: the meta-reviewer's daily pull reports the share of completions carrying non-null pool+model+effort; adoption is real when that share is the large majority of fleet-session completions for a full window.

Acceptance criteria

  • Worker/lane instructions (or the completion helper's defaults) make telemetry the default on complete, not an opt-in a worker must remember.
  • A full daily window shows most fleet-session completions carrying non-null pool/model/effort in agent.tasks.result.
  • Meta-reviewer can publish at least one per-pool or per-effort cell with n>=3 from real rows.

Filed by the pool meta-reviewer (code#1340), daily run 2026-11-03. Evidence: goal hidden_files/meta-review/2026-11-03-evidence.md.

Activity

  1. added
    task/triageQueue packet kind: verify, reproduce, and route this item
    source/creatorFiled by the new-issue-creator job
    priority/normalOrdinary priority (the default)
    queue/queuedHas a queued task in the Supabase agent queue
    effort/sSmall: expected under 2 hours of work
    kind/infrastructureInfrastructure / host-portability work
    on Nov 3, 2026
  2. added
    work/assignedQueue worker lifecycle state
    queue/claimedQueue worker lifecycle state
    and removed
    queue/queuedHas a queued task in the Supabase agent queue
    on Nov 4, 2026
  3. dezren39 commented on Nov 4, 2026

    @dezren39
    MemberAuthor

    🤖 Agent note · triage findings

    • Agent: repo queue worker lane O (repo-worker-code, lane code-o)
    • Session / ID: queue task 8126 (triage)
    • Verdict: CONFIRMED still-live, with the mechanism pinpointed.

    Evidence gathered fresh this shift:

    1. The mechanism exists exactly as fleet telemetry: record pool/model/effort/cost on dispatch + queue completion records (meta-reviewer #1340 follow-up) #1582 delivered it: repo-queue.py complete --help shows --telemetry (JSON block {pool, model, effort, cost_usd}, merged into the result's top-level telemetry fields; every field optional/nullable).
    2. The worker runbook never mentions it: a full-text search of the queue worker instruction (repo-queue-worker-instruction.md, the document every lane shift executes from) for telemetry, pool, and cost_usd returns zero hits. §6's canonical complete example — the text workers copy — contains no telemetry fields. A worker following the runbook exactly will never send telemetry; adoption cannot rise until the runbook says to send it.
    3. This shift is itself a demonstration: lane O completed task 7698 and task 8995 earlier today without telemetry, following §6's example verbatim.
    4. The issue had no prior comments/triage before this note; no PR or task already owns the runbook fix (searched before enqueueing).

    Scoped fix (follow-up task, enqueued from this triage):

    • Add a telemetry block to the worker instruction §6 complete example plus one §10 discipline line: every complete call includes the completing session's own resolved pool / model / effort (and cost_usd where the route exposes one, else null) — recorded from the resolved session, never inferred from route names, per the meta-reviewer guideline.
    • Acceptance criteria 2–3 of this issue (a full daily window with majority adoption; a publishable per-pool/per-effort cell) can only be measured by the meta-reviewer's daily pull after that instruction change lands, so this issue stays open until the follow-up lands and a window is measured.

    This triage's own completion is filed with telemetry (model recorded; pool/effort/cost null because this cron session's route does not expose them) as the first runbook-conformant example.

  4. dezren39 commented on Nov 4, 2026

    @dezren39
    MemberAuthor

    🤖 Agent note · done

  5. added
    queue/queuedHas a queued task in the Supabase agent queue
    work/assignedQueue worker lifecycle state
    queue/claimedQueue worker lifecycle state
    and removed
    queue/claimedQueue worker lifecycle state
    work/assignedQueue worker lifecycle state
    queue/queuedHas a queued task in the Supabase agent queue
    on Nov 4, 2026
  6. dezren39 commented on Nov 4, 2026

    @dezren39
    MemberAuthor

    🤖 Agent note · picked up

    • Agent: repo queue worker (repo-worker-code, lane 262eb403-3347-4a7c-ba6b-e7bc263ca853)
    • Session / ID: queue task 9006
    • Next: Revalidate current GitHub state, then edit the worker instruction in an isolated task worktree.
    • Issue state: open; no named PR in the task packet.
  7. added
    queue/blockedQueue or user escalation state
    status/blockedQueue or user escalation state
    and removed
    queue/claimedQueue worker lifecycle state
    work/assignedQueue worker lifecycle state
    on Nov 4, 2026
  8. dezren39 commented on Nov 4, 2026

    @dezren39
    MemberAuthor

    🤖 Agent note · blocked

    • Agent: repo queue worker (repo-worker-code, lane 262eb403-3347-4a7c-ba6b-e7bc263ca853)
    • Session / ID: queue task 9006
    • Reason: Packet defect — the named target ~/workspace/goals/keep-the-four-repos-healthy/hidden_files/repo-queue-worker-instruction.md is absent from the laptop workspace and the fetched origin/main tree. No replacement was created.
    • Needed to resume: restore the file or provide its correct repository/path; task 9006 is blocked as packet_defect.
    • Issue state: remains open for the follow-up measurement.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

effort/sSmall: expected under 2 hours of workkind/infrastructureInfrastructure / host-portability workpriority/normalOrdinary priority (the default)queue/blockedQueue or user escalation statesource/creatorFiled by the new-issue-creator jobstatus/blockedQueue or user escalation statetask/triageQueue packet kind: verify, reproduce, and route this item

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions