Repository navigation
rollout(fleet): OpenCode-half pilot on the code queue + extension to opencode/openchamber/nix (OpenCode-half step 5) #1658
Description
Part of the OpenCode-half redesign (design: files/opencode-half-redesign-2026-10-28.md in the keep-the-four-repos-healthy goal workspace; tracking: #1425). Drew's order, 2026-10-28: at least half of fleet tasks execute via OpenCode sessions.
Scope
Rollout sequencing once the pieces land (lane assignment, executor mode, keeper coverage, lane-share measurement):
- Pilot: enable the split on the code queue only (lanes A–D), concurrency cap 4 queue-spawned sessions, for one week.
- Pilot review against the exit criteria: rolling 7-day executed share ≥50%; zero evaporated claims attributable to the split; no rise in packet-defect or verification-failure rates attributable to session execution; miss reasons understood (a chronic
saturatedorquotashare is a capacity finding, reported as itself). - Extension: enable for the opencode, openchamber, and nix queues (their lanes share
lane-shift-instruction.md, so the instruction change is written once) with the cap revisited against observed laptop load and pool headroom — noting the Go plan lapses (Personal Nov 2, Lab Nov 6) shrink workhorse seats during this window. - Record the pilot outcome and the extension decision as comments on Deployed laptop opencode config has drifted from canonical HEAD — claude-code pause (PR #1316) not in effect live #1425 and in the design doc's rollout section.
Acceptance criteria
- One full pilot week on the code queue with daily lane-share published.
- Pilot review comment on Deployed laptop opencode config has drifted from canonical HEAD — claude-code pause (PR #1316) not in effect live #1425 addressing every exit criterion with numbers.
- Extension enabled (or a documented decision not to extend yet, with the blocker named) for the other three queues.
Metadata
Metadata
Assignees
Labels
Infrastructure, deployment, and operationsLarge: multi-day or open-ended; expect staged check-insNew feature or capability requestIntake review done: acceptance criteria, links, and normalized labels addedFollow-up spun out of earlier workImportant and time-sensitiveFiled directly by a humanQueue packet kind: verify, reproduce, and route this item
Activity
developing-today commentedon Oct 28, 2026
CODE-672
dezren39 commentedon Oct 28, 2026
Intake review
files/opencode-half-redesign-2026-10-28.md, keep-the-four-repos-healthy goal workspace); tracking epic Deployed laptop opencode config has drifted from canonical HEAD — claude-code pause (PR #1316) not in effect live #1425; depends on the earlier steps feat(queue): claim-time execution-lane assignment — ledger + lane verb in repo-queue.py (OpenCode-half step 1) #1654 (lane assignment), feat(fleet): shift executor mode — launch + shepherd OpenCode sessions for opencode-lane tasks (OpenCode-half step 2) #1655 (executor mode), feat(keeper): binding-aware supervision for queue-spawned OpenCode sessions (OpenCode-half step 3) #1656 (keeper supervision), feat(fleet): daily lane-share measurement, publishing + below-floor alert (OpenCode-half step 4) #1657 (lane-share measurement).## Acceptance criteriasection, so no body edit was made.priority/high(Drew-ordered rollout gate),effort/l(pilot week + extension),source/human,status/needs-triage,area/infra,meta/enriched.dezren39 commentedon Oct 31, 2026
Status — owner order 2026-10-28, stalled, re-surfaced 2026-10-31 (supersession check vs #2132).
Verdict of the owed supersession check, 2026-10-31:
NOT superseded by #2132. #2132/#2153 reorganise two OpenChamber jobs; this step is the rollout sequencing for the OpenCode-half redesign itself (code-queue pilot → review against exit criteria → extension to the opencode/openchamber/nix queues) and depends on steps 2–4 (#1655 executor mode, #1656 binding-aware supervision, #1657 lane-share measurement) landing first. The conductor work neither sequences nor gates it.
Queue state, verified by direct read 2026-10-31: live triage task 2626 (repo-health-code-pilot, queued, priority 70, attempts 0/3). No duplicate packet enqueued. Triage should treat the #1655–#1657 landing state as this step's gate, exactly as the issue body scopes it.
5 remaining items
developing-today commentedon Nov 3, 2026
agent-note:v1
state: reviewed (coverage loop-back)
Coverage pass 2026-11-03, loop-back review: this issue was re-reviewed (drift / overlap / already-handled / stale / blockers / ordering) so every open issue in the four repos is reviewed and in the system. Verdict: still live; in the system as task 7246 (triage, queued, priority 50) in repo-health-code-pilot. Labels audited and synced. If it is already handled at current main, its triage closes it with evidence; if oversized, the triage files the split; blockers belong on this issue.
Found via: coverage loop-back 2026-11-03
dezren39 commentedon Nov 3, 2026
🤖 Agent note · picked up
dezren39 commentedon Nov 3, 2026
Triage verdict — NOT done as specified; still live, gated (task 7246)
Re-verified fresh on 2026-11-03 (~23:10Z), superseding the packet's "believed DONE" claim and updating the task 2626 triage (2026-11-03T00:45Z):
What is genuinely live (the packet's kernel of truth):
opencode(task 4876's ledger row showsexecuted_lane: opencodewith a bound session), and the laptop claimant launches the bound sessions.Why the acceptance criteria are NOT met — do not close:
queue/blocked+status/blocked; its latest triage (task 8602, 2026-11-03T22:16Z) confirms AC1 is not complete end-to-end — the ledger-backed lookup is not registered yet, so queue-launched sessions still reportqueue: null.lane-share.jsonwas last generated 2026-11-01T23:58Z, and its most recent rolling 7-day executed share is 0.3372 — below the 50% pilot exit floor. No clean pilot week exists on record.Routing: no follow-up packet enqueued from this triage — the rollout is effort/l and gated, exactly as task 2626 routed it. Resume condition unchanged: when #1656 and #1657 are both CLOSED and a live tracking home for the pilot review is confirmed, enqueue the rollout implementation task and start the one-week code-queue pilot clock. This issue stays open.
🤖 Agent note · done
triaged