heady-pipeline-ops

name	heady-pipeline-ops
description	How to operate, debug, and monitor the HCFullPipeline continuous execution engine

HCFullPipeline Operations Skill

Overview

The HCFullPipeline is a 21-stage (fib(8)) DAG execution engine that runs continuously. It is the core autonomous execution backbone of the Heady™ platform.

Key Files

• src/hc_pipeline.js — DAG engine, topological sort, circuit breakers, worker pool
• src/hc_pipeline_handlers.js — 100+ registered task handlers (fail-closed)
• configs/hcfullpipeline.yaml — Stage definitions, tasks, dependencies, gate configs
• heady-manager.js — Boot sequence, continuous mode, API endpoints

Starting / Stopping

Auto-Start (default)

The pipeline auto-starts on server boot with a φ² delay (~2.6s). Check status:

GET /api/buddy/pipeline/continuous → { running, cycleCount, gates }

Manual Control

POST /api/buddy/pipeline/continuous { "action": "start" }
POST /api/buddy/pipeline/continuous { "action": "stop" }

Single Run

POST /api/pipeline/run → { runId, status, metrics }

Reading Pipeline State

GET /api/pipeline/state → { runId, status, metrics }
GET /api/pipeline/config → { name, version, stages, taskCount }

Debugging Fail-Closed Errors

If a task throws [FAIL-CLOSED] No handler registered for task 'task_name', you need to register a handler:

// In src/hc_pipeline_handlers.js
registerTaskHandler('new_task_name', async (ctx) => {
  // ctx has: runId, stageId, configs, supervisorConfig
  return { status: 'completed', /* results */ };
});

Gate System

The continuous pipeline checks 4 gates before each cycle:

• quality — always true (reserved for future code quality gate)
• resource — heap usage < 90% of total
• stability — fewer than 5 accumulated errors
• user — user hasn't requested stop

If any gate fails, the pipeline halts with exitReason: "gate_failed".

Sacred Geometry Constants

• Cycle interval: φ⁷ × 1000 = 29,034ms
• Max concurrent tasks: 8 (fib(6))
• Max retries: 3 (fib(4))
• Cache TTL: 2,618,000ms (φ² × 1M)
• Cache max entries: 233 (fib(13))

Monitoring

Check the pipeline log at .heady/hc_pipeline.log and each stage's state in the run result object.