aiwg/agentic/code/addons/ralph/schemas/multi-loop-registry.yaml

Cognitive architecture for AI-augmented software development with structured memory, ensemble validation, and closed-loop correction. FAIR-aligned artifacts, 84% cost reduction via human-in-the-loop, standards adopted by 100+ organizations.

# Multi-Loop Registry Schema
# Based on REF-086 Coordination Tax, REF-088 Multi-Agent Guide
# Issue: #266
#
# Tracks all active Ralph loops in a repository to enable:
# - Concurrent loop execution without state conflicts
# - MAX_CONCURRENT_LOOPS enforcement (research-backed limit)
# - Cross-loop coordination and learning
# - Recovery from crashed loops

$schema: "https://json-schema.org/draft/2020-12/schema"
$id: "https://aiwg.io/schemas/multi-loop-registry/v1"
title: "Multi-Loop Registry Schema"
description: |
  Central registry for tracking multiple concurrent Ralph loops.
  Enforces MAX_CONCURRENT_LOOPS=4 based on coordination tax research.

type: object
required:
  - version
  - max_concurrent_loops
  - updated_at
  - active_loops
  - total_active

properties:
  version:
    type: string
    const: "2.0.0"
    description: "Registry schema version"

  max_concurrent_loops:
    type: integer
    const: 4
    description: |
      Maximum concurrent loops (REF-086, REF-088).
      Research shows 3-4 loops optimal, >4 creates excessive
      coordination overhead (n*(n-1)/2 communication paths).

  updated_at:
    type: string
    format: date-time
    description: "Last registry update timestamp"

  active_loops:
    type: array
    maxItems: 5  # Allow 1 over for --force override
    items:
      $ref: "#/$defs/LoopEntry"
    description: "Currently active loops"

  total_active:
    type: integer
    minimum: 0
    description: "Count of active loops (auto-calculated)"

  total_completed:
    type: integer
    minimum: 0
    default: 0
    description: "Lifetime completed loops count"

  total_aborted:
    type: integer
    minimum: 0
    default: 0
    description: "Lifetime aborted loops count"

$defs:
  LoopEntry:
    type: object
    required:
      - loop_id
      - task_summary
      - status
      - started_at
    properties:
      loop_id:
        type: string
        pattern: "^ralph-[a-z0-9-]+-[a-f0-9]{8}$"
        description: "Unique loop identifier (ralph-{slug}-{short-uuid})"
        examples:
          - "ralph-fix-all-tests-a1b2c3d4"
          - "ralph-typescript-migration-e5f6g7h8"

      task_summary:
        type: string
        maxLength: 200
        description: "Brief description of the task"

      status:
        type: string
        enum:
          - running      # Currently executing
          - paused       # Temporarily suspended
          - waiting      # Waiting for resources/dependencies
          - completing   # Finishing up
        description: "Current loop status"

      started_at:
        type: string
        format: date-time
        description: "When loop was started"

      owner:
        type: string
        description: "Agent or user that owns this loop"
        examples:
          - "agent-1"
          - "software-implementer"
          - "user:jmagly"

      pid:
        type: ["integer", "null"]
        description: "Process ID if running (null if paused)"

      iteration:
        type: integer
        minimum: 0
        description: "Current iteration number"

      max_iterations:
        type: integer
        minimum: 1
        description: "Maximum iterations for this loop"

      completion_criteria:
        type: string
        description: "Verification command for completion"

      priority:
        type: string
        enum: [low, medium, high, critical]
        default: medium
        description: "Loop priority for resource allocation"

      tags:
        type: array
        items:
          type: string
        description: "Tags for categorization and filtering"

# Coordination metrics (REF-086, REF-088)
coordination:
  communication_paths:
    formula: "n * (n - 1) / 2"
    description: "Number of potential inter-loop communication paths"
    examples:
      2_loops: 1
      3_loops: 3
      4_loops: 6
      5_loops: 10
      7_loops: 21

  complexity_levels:
    trivial: 1-2     # 1 path, simple parallel
    manageable: 3-6  # 3-4 loops, RECOMMENDED
    significant: 10-21  # 5-7 loops, complex workflows only
    excessive: 28+   # 8+ loops, AVOID

  overhead_factors:
    independent_error_amplification: 17.2  # REF-086
    centralized_error_amplification: 4.4   # REF-086
    recommendation: "Use centralized coordination (registry pattern)"

# Loop lifecycle
lifecycle:
  creation:
    - validate_max_concurrent_loops
    - generate_unique_loop_id
    - create_loop_directory
    - register_in_registry
    - initialize_state

  execution:
    - update_iteration_count
    - update_status
    - checkpoint_periodically
    - coordinate_with_registry

  completion:
    - mark_completed
    - unregister_from_active
    - archive_loop_directory
    - extract_learnings
    - increment_total_completed

  abortion:
    - mark_aborted
    - unregister_from_active
    - preserve_state_for_analysis
    - increment_total_aborted

# File locking strategy (REF-082, Kleppmann)
locking:
  mechanism: lease_based
  lease_duration_ms: 30000
  retry_interval_ms: 100
  max_attempts: 300
  stale_detection:
    - check_lease_expiry
    - check_process_exists
  fencing_tokens: true

# Storage configuration
storage:
  registry_path: ".aiwg/ralph/registry.json"
  loops_dir: ".aiwg/ralph/loops/"
  archive_dir: ".aiwg/ralph/archive/"
  shared_dir: ".aiwg/ralph/shared/"

  loop_structure: |
    .aiwg/ralph/loops/{loop-id}/
    ├── state.json              # Loop state
    ├── iterations/             # Per-iteration artifacts
    │   ├── iteration-1.json
    │   └── ...
    ├── checkpoints/            # Recovery checkpoints
    │   ├── checkpoint-1.json.gz
    │   └── ...
    └── debug-memory/           # Debug/learning data
        └── ...

# Example registry
examples:
  - version: "2.0.0"
    max_concurrent_loops: 4
    updated_at: "2026-02-02T15:00:00Z"
    active_loops:
      - loop_id: "ralph-fix-tests-a1b2c3d4"
        task_summary: "Fix all failing tests"
        status: "running"
        started_at: "2026-02-02T14:00:00Z"
        owner: "software-implementer"
        pid: 12345
        iteration: 3
        max_iterations: 10
        completion_criteria: "npm test passes"
        priority: "high"

      - loop_id: "ralph-typescript-e5f6g7h8"
        task_summary: "Migrate to TypeScript"
        status: "paused"
        started_at: "2026-02-02T13:00:00Z"
        owner: "software-implementer"
        pid: null
        iteration: 7
        max_iterations: 20
        completion_criteria: "npx tsc --noEmit exits 0"
        priority: "medium"
        tags: ["migration", "typescript"]

    total_active: 2
    total_completed: 15
    total_aborted: 3

# CLI integration
cli_commands:
  create:
    command: "aiwg ralph <task> --completion <criteria>"
    output: "Loop started: {loop-id}"

  status:
    single: "aiwg ralph-status --loop-id <id>"
    all: "aiwg ralph-status --all"

  abort:
    command: "aiwg ralph-abort --loop-id <id>"

  resume:
    command: "aiwg ralph-resume --loop-id <id>"

  list:
    command: "aiwg ralph list"
    description: "Show all active loops"

# References
references:
  research:
    - "@.aiwg/research/findings/REF-086-multi-agent-coordination-tax.md"
    - "@.aiwg/research/findings/REF-088-dev-multi-agent-guide-2026.md"
    - "@.aiwg/research/findings/REF-082-multi-agent-orchestration.md"
  planning:
    - "@.aiwg/working/multi-loop-ralph-plan.md"
  implementation:
    - "#266"
  related:
    - "@agentic/code/addons/ralph/schemas/checkpoint.yaml"
    - "@agentic/code/addons/ralph/schemas/iteration-analytics.yaml"