Pipeline Registry - AGI-Corporation/frontier-os-app-builder GitHub Wiki

Pipeline Registry

Navigation: [Home]] ](/AGI-Corporation/frontier-os-app-builder/wiki/[Core-Concepts) | [Evolution Bridge Service]] | Task Dispatch and Routing Logs

Overview

The Pipeline Registry is the catalog of all self-evolving agent structures available in agent.bay. Every entry in the marketplace is an EvolutionPipeline record managed through the registry.

The registry supports:

  • Registration — adding new pipelines with roles, endpoints, and payment configuration.
  • Retrieval — listing and fetching individual pipelines for discovery and dispatch.
  • Sync — heartbeat updates that reflect the pipeline's evolving state.
  • Activation — toggling pipelines active/inactive without deletion.

Built-In Pipelines

Three pipelines ship with the mock store out of the box:

Ralph Hive — Bug Repair Loop

{
  id: 'ralph-hive-bug-repair',
  name: 'Ralph Hive — Bug Repair Loop',
  description: 'Automated log monitoring, patch generation, and safety auditing cycle',
  roles: ['observer', 'architect', 'auditor'],
  endpoint: 'https://ralph-hive.frontier.ai/bug-repair',
  paymentAddress: '0xRalphHiveBugRepair...',
  pricePerTask: 10.0,
  isActive: true
}

Evolution behavior: Learns which log patterns consistently precede bugs, refines patch strategies by success rate, and hardens auditor checks based on rejected patches.

Ralph Hive — Feature Planner

{
  id: 'ralph-hive-feature-planner',
  name: 'Ralph Hive — Feature Planner',
  description: 'GitManager-integrated feature planning with safe branching and rollback',
  roles: ['planner'],
  endpoint: 'https://ralph-hive.frontier.ai/feature-planner',
  paymentAddress: '0xRalphHiveFeaturePlanner...',
  pricePerTask: 25.0,
  isActive: true
}

Evolution behavior: Learns team velocity and delivery patterns, improves sprint planning heuristics, integrates GitManager history to avoid repeating failed strategies.

NANDA Cross-Agent Bridge

{
  id: 'nanda-cross-agent',
  name: 'NANDA Cross-Agent Bridge',
  description: 'Routes tasks to external specialist agents across agent networks via NANDA',
  roles: ['observer', 'architect', 'auditor', 'planner'],
  endpoint: 'https://nanda.frontier.ai/bridge',
  paymentAddress: '0xNANDABridge...',
  pricePerTask: 5.0,
  isActive: true
}

Evolution behavior: Routing heuristics improve with each delegation; learns which external networks handle which problem types best.

Registering a New Pipeline

Via the Bridge Service

const bridge = createEvolutionBridgeService()

const newPipeline = await bridge.registerPipeline({
  name: 'My Security Scanner',
  description: 'Continuously scans for CVEs and auto-patches dependencies',
  roles: ['observer', 'architect'],
  endpoint: 'https://my-scanner.example.com/dispatch',
  paymentAddress: '0xMyPaymentAddress...',
  pricePerTask: 8.0
})

console.log(newPipeline.id)  // auto-generated
console.log(newPipeline.isActive)  // true
console.log(newPipeline.taskCount) // 0

Registration Checklist

Before registering a production pipeline, ensure:

  • endpoint is publicly reachable and returns valid JSON responses
  • paymentAddress is a valid x402-compatible address you control
  • pricePerTask is set appropriately for the compute cost of your pipeline
  • roles accurately reflects what the pipeline can execute
  • The endpoint accepts the DispatchTaskParams payload schema

Fetching Pipelines

// List all active pipelines
const all = await bridge.listPipelines()

// Fetch a specific pipeline
const pipeline = await bridge.getPipeline('ralph-hive-bug-repair')
if (!pipeline) {
  console.error('Pipeline not found')
}

Filtering by Role

The listPipelines() result can be filtered client-side by role:

const observerPipelines = (await bridge.listPipelines())
  .filter(p => p.roles.includes('observer'))

Syncing a Pipeline

Sync is the mechanism by which a pipeline's live state is reflected in agent.bay.

const updated = await bridge.syncPipeline('ralph-hive-bug-repair')
console.log(updated.syncedAt)  // fresh ISO timestamp

When to Sync

Trigger Frequency
After significant evolution policy change On-demand
Health check / heartbeat Every 5–15 minutes (production)
After a batch of tasks completes Post-batch
On pipeline owner dashboard refresh On-demand

In production, sync also pulls:

  • Current Bitrefill-style prepaid balance remaining.
  • Updated evolution policy version.
  • Health check status from the pipeline endpoint.

Pipeline States

State isActive Description
Active true Accepting task dispatches
Inactive false Not accepting dispatches; hidden from default discovery
Archived false + evolution lock Evolution frozen; history preserved for audit

To deactivate a pipeline:

// Production bridge would expose a deactivate method
// In mock: mutate the store directly during dev
const pipeline = await bridge.getPipeline('my-pipeline')
pipeline.isActive = false

Pipeline Evolution Metadata

Each pipeline accumulates evolution signals over time:

{
  taskCount: 142,       // total dispatches, used as evolution maturity signal
  syncedAt: '2026-03-29T16:00:00Z',  // last heartbeat
  // future: evolutionVersion, policyHash, childPipelineIds
}

High taskCount pipelines are considered more evolved and may receive discovery boosts, governance weight, or automatic child pipeline spawning in future versions.

Related Pages