All runbooks
runbookuapiredacted view

UAPI Sixty-Third Wedge Wave 5 Requeue Helper Receipt

Receipt for Wave 5 agent requeue helper that restarts continuation-required lanes with Grok Builds and GLM 5.2 effort max.

Updated 2026-06-23·Freshness: Reference·Sensitive tokens or credentials were hidden before display.

UAPI Sixty-Third Wedge Wave 5 Requeue Helper Receipt

Receipt for Wave 5 agent requeue helper that restarts continuation-required lanes with Grok Builds and GLM 5.2 effort max.

Promotion Metadata

  • Source feedback id: 237
  • Source feedback ids: 237
  • Promotion request id: 159
  • Feedback category: beneficial
  • Feedback source: ai-lane
  • Feedback created at: 2026-06-23T04:34:02.231139+00:00
  • Target slug: [REDACTED:high-entropy]
  • Review state: review
  • Reviewer note: AI lane auto-promoted redacted feedback into internal catalog context.

Source Feedback Body

Feedback 237

  • Product: uapi
  • Category: beneficial
  • Related slug: [REDACTED:high-entropy]
  • Source: ai-lane
  • Created at: 2026-06-23T04:34:02.231139+00:00

UAPI Sixty-Third Wedge Wave 5 Requeue Helper Receipt

Date: 2026-06-23 Authority: fixture_only_no_execution

Summary

Added a Wave 5 requeue helper so the agent-team loop can restart only lanes that the Wave 5 collector marks as needing continuation.

This keeps completed Grok work from being rerun while GLM 5.2 lanes recover from provider/runtime failures or missing reports.

Files

  • scripts/uapi-requeue-agent-team-wave5.sh
  • tests/test_agent_team_wave5_collector.py

Behavior

  • Runs scripts/uapi-collect-agent-team-wave5.sh.
  • Selects report lines with:
  • continuation_required=true
  • requeue_packet_present=true
  • Supports dry-run by default with UAPI_AGENT_REQUEUE_EXECUTE=false.
  • Starts Grok lanes only with grok --model grok-build --effort max.
  • Starts GLM lanes only with claude --model glm-5.2 --effort max --print.
  • Preserves bounded GLM timeout through UAPI_GLM_TIMEOUT_SECONDS, default 2700 seconds.
  • Does not print the Explorer key.

Verification

Commands:

PYTHONPATH=src python3 -m unittest tests.test_agent_team_wave5_collector tests.test_agent_team_wave4_collector
UAPI_AGENT_COLLECT_SKIP_PULL=true UAPI_AGENT_REQUEUE_EXECUTE=false [REDACTED:high-entropy] [REDACTED:high-entropy] scripts/uapi-requeue-agent-team-wave5.sh

Results:

  • Focused tests: 7 passed.
  • Dry run selected only:
  • glm52_wave5_a
  • glm52_wave5_b
  • Dry run did not select completed Grok Wave 5 reports.

Live Agent State

At receipt time, a bounded GLM Wave 5 A rerun was already active on the VPS:

  • Runtime command includes claude --model glm-5.2 --effort max --print.
  • Remote PID file: /root/agent-logs/glm52-wave5-a-rerun.pid
  • Remote log: /root/agent-logs/glm52-wave5-a-rerun.log
  • Remote exit: /root/agent-logs/glm52-wave5-a-rerun.exit

The job had no exit file and no report yet, so it was left running rather than duplicated.

Boundaries

  • No route activation.
  • No provider backend dispatch from UAPI.
  • No hardware execution.
  • No storage writes.
  • No model downloads.
  • No credential printing.