aiwg/agentic/code/frameworks/sdlc-complete/schemas/flows/error-handling.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.

# Flow Error Handling Configuration Schema
# Based on REF-001 Production Agentic (error handling patterns)
# Issue: #110

$schema: "https://json-schema.org/draft/2020-12/schema"
$id: "https://aiwg.io/schemas/error-handling/v1"
title: "Flow Error Handling Configuration Schema"
description: |
  Structured error handling configuration for flow commands, enabling retry patterns,
  fallback agents, and graceful degradation. Based on production agentic best practices.

type: object
required:
  - strategy
  - classification

properties:
  strategy:
    type: string
    enum:
      - retry_then_escalate   # Retry, then escalate to human
      - retry_then_fallback   # Retry, then use fallback agent
      - fail_fast             # Fail immediately, no retry
      - graceful_degrade      # Continue with reduced functionality
      - checkpoint_recover    # Restore from last checkpoint
    description: "Primary error handling strategy"

  classification:
    type: object
    required: [categories]
    description: "How to classify errors"
    properties:
      categories:
        type: array
        items:
          $ref: "#/$defs/ErrorCategory"
      default_category:
        type: string
        default: "unknown"

  retry:
    $ref: "#/$defs/RetryConfig"
    description: "Retry configuration"

  fallback:
    $ref: "#/$defs/FallbackConfig"
    description: "Fallback agent configuration"

  escalation:
    $ref: "#/$defs/EscalationConfig"
    description: "Human escalation configuration"

  checkpoint:
    $ref: "#/$defs/CheckpointConfig"
    description: "Checkpoint and recovery configuration"

  logging:
    type: object
    properties:
      level:
        type: string
        enum: [error, warn, info, debug]
        default: error
      include_stack:
        type: boolean
        default: true
      include_context:
        type: boolean
        default: true
      destination:
        type: string
        default: ".aiwg/logs/errors.log"
    description: "Error logging configuration"

$defs:
  ErrorCategory:
    type: object
    required:
      - id
      - pattern
      - type
    properties:
      id:
        type: string
        description: "Category identifier"
      pattern:
        type: string
        description: "Regex pattern to match error messages"
      type:
        type: string
        enum:
          - transient      # Temporary failures (network, rate limit)
          - permanent      # Unrecoverable (invalid input, missing resource)
          - user           # User action needed
          - system         # System/infrastructure error
          - timeout        # Operation timed out
          - validation     # Validation failure
          - permission     # Permission denied
        description: "Error type for handling decision"
      severity:
        type: string
        enum: [critical, high, medium, low]
        default: medium
      action:
        type: string
        enum: [retry, fallback, escalate, abort, ignore]
        description: "Default action for this category"
      message_template:
        type: string
        description: "User-friendly message template"

  RetryConfig:
    type: object
    properties:
      enabled:
        type: boolean
        default: true
      max_attempts:
        type: integer
        minimum: 1
        maximum: 10
        default: 3
        description: "Maximum retry attempts"
      initial_delay_ms:
        type: integer
        minimum: 0
        default: 1000
        description: "Initial delay before first retry"
      backoff_strategy:
        type: string
        enum:
          - none          # No delay increase
          - linear        # Add fixed delay each retry
          - exponential   # Double delay each retry
          - fibonacci     # Fibonacci sequence delays
        default: exponential
      max_delay_ms:
        type: integer
        default: 30000
        description: "Maximum delay between retries"
      jitter:
        type: boolean
        default: true
        description: "Add random jitter to delays"
      retry_on:
        type: array
        items:
          type: string
        default: ["transient", "timeout"]
        description: "Error types to retry on"

  FallbackConfig:
    type: object
    properties:
      enabled:
        type: boolean
        default: false
      agents:
        type: array
        items:
          type: object
          properties:
            primary:
              type: string
              description: "Primary agent"
            fallback:
              type: string
              description: "Fallback agent"
            conditions:
              type: array
              items:
                type: string
              description: "When to trigger fallback"
        description: "Agent fallback mappings"
      model_fallback:
        type: object
        properties:
          enabled:
            type: boolean
            default: false
          primary:
            type: string
            enum: [opus, sonnet, haiku]
          fallback:
            type: string
            enum: [opus, sonnet, haiku]
        description: "Model tier fallback"

  EscalationConfig:
    type: object
    properties:
      enabled:
        type: boolean
        default: true
      trigger_after:
        type: integer
        default: 3
        description: "Escalate after N failures"
      channels:
        type: array
        items:
          type: string
          enum: [cli, issue_comment, slack, email]
        default: [cli, issue_comment]
      include_context:
        type: boolean
        default: true
        description: "Include error context in escalation"
      template:
        type: string
        description: "Escalation message template"
      auto_create_issue:
        type: boolean
        default: false
        description: "Auto-create tracking issue"

  CheckpointConfig:
    type: object
    properties:
      enabled:
        type: boolean
        default: true
      auto_checkpoint:
        type: boolean
        default: true
        description: "Automatically checkpoint before risky operations"
      checkpoint_dir:
        type: string
        default: ".aiwg/checkpoints/"
      retention_count:
        type: integer
        default: 5
        description: "Number of checkpoints to retain"
      checkpoint_on:
        type: array
        items:
          type: string
          enum:
            - phase_start
            - artifact_complete
            - before_external_call
            - iteration_boundary
        default: [phase_start, artifact_complete]
      recovery_strategy:
        type: string
        enum:
          - last_checkpoint     # Restore most recent
          - select_checkpoint   # Let user choose
          - smart_rollback      # Analyze and select best
        default: last_checkpoint

# Predefined error patterns
common_patterns:
  network_errors:
    id: "network"
    pattern: "(ECONNREFUSED|ETIMEDOUT|ENOTFOUND|network|socket)"
    type: transient
    severity: medium
    action: retry

  rate_limit:
    id: "rate_limit"
    pattern: "(rate.?limit|429|too.?many.?requests)"
    type: transient
    severity: low
    action: retry

  auth_errors:
    id: "auth"
    pattern: "(unauthorized|403|401|permission.?denied|access.?denied)"
    type: permission
    severity: high
    action: escalate

  validation_errors:
    id: "validation"
    pattern: "(invalid|validation.?failed|schema.?error|type.?error)"
    type: validation
    severity: medium
    action: escalate

  timeout_errors:
    id: "timeout"
    pattern: "(timeout|timed.?out|deadline.?exceeded)"
    type: timeout
    severity: medium
    action: retry

  resource_errors:
    id: "resource"
    pattern: "(not.?found|404|missing|does.?not.?exist)"
    type: permanent
    severity: high
    action: escalate

# Flow integration
flow_integration:
  description: "How to add error handling to flow YAML"

  example: |
    # In flow command definition
    flow:
      name: my-flow
      error_handling:
        strategy: retry_then_escalate
        retry:
          max_attempts: 3
          backoff_strategy: exponential
        escalation:
          channels: [cli, issue_comment]
        checkpoint:
          auto_checkpoint: true
          checkpoint_on: [phase_start]

# Ralph integration
ralph_integration:
  description: "Error handling in Ralph loops"

  on_iteration_failure:
    - increment_failure_count: true
    - check_retry_budget: true
    - checkpoint_state: true
    - analyze_error_pattern: true

  recovery_actions:
    transient: retry_with_backoff
    validation: adjust_and_retry
    permanent: skip_or_escalate
    timeout: extend_timeout_and_retry

# Examples
examples:
  - name: "Standard flow error handling"
    strategy: retry_then_escalate
    classification:
      categories:
        - id: "api_error"
          pattern: "(API|api).*(error|failed)"
          type: transient
          action: retry
      default_category: unknown
    retry:
      enabled: true
      max_attempts: 3
      backoff_strategy: exponential
      initial_delay_ms: 1000
    escalation:
      enabled: true
      trigger_after: 3
      channels: [cli, issue_comment]

# References
references:
  research:
    - "@.aiwg/research/findings/REF-001-production-agentic.md"
  implementation:
    - "#110"
  related:
    - "@agentic/code/frameworks/sdlc-complete/schemas/flows/hitl-gate.yaml"
    - "@agentic/code/addons/ralph/schemas/reflection-memory.json"