Skip to content

Fallback Cascade Command

Usage

/amplihack:cascade <TASK_DESCRIPTION>

Purpose

Execute fallback cascade pattern for resilient operations. Graceful degradation from optimal → pragmatic → minimal ensures reliable completion.

EXECUTION INSTRUCTIONS FOR CLAUDE

When this command is invoked, you MUST:

  1. Import the orchestrator:
import sys
from pathlib import Path
sys.path.insert(0, str(Path.cwd() / ".claude/tools/amplihack"))
from orchestration.patterns.cascade import run_cascade
  1. Execute the pattern:
result = run_cascade(
    task_prompt="{TASK_DESCRIPTION}",
    fallback_strategy="quality",  # or "service", "freshness"
    timeout_strategy="balanced",  # or "aggressive", "patient"
    working_dir=Path.cwd()
)
  1. Display results:
  2. Show final result and cascade level reached
  3. Explain any degradation from optimal
  4. Report which fallback succeeded
  5. Report session_id for traceability

  6. Link to logs: .claude/runtime/logs/cascade_<timestamp>/

  7. Manual fallback (if orchestrator unavailable):

  8. Read workflow: .claude/workflow/CASCADE_WORKFLOW.md
  9. Execute steps manually with TodoWrite tracking

When to Use

Use for operations with multiple viable approaches:

  • External API calls (primary service, backup service, cached fallback)
  • Code generation (GPT-4, Claude, cached templates)
  • Data retrieval (database, cache, defaults)
  • Complex computations (exact algorithm, approximation, heuristic)

Cost-Benefit

  • Cost: 1.1-2x execution time (only on failures)
  • Benefit: 95%+ reliability vs 70-80% single approach
  • ROI Positive when: Operation reliability > availability requirements

Task Description

Execute the following task with fallback cascade:

{TASK_DESCRIPTION}

Configuration

The workflow can be customized by editing .claude/workflow/CASCADE_WORKFLOW.md:

  • Timeout strategy: Aggressive (5/2/1s), Balanced (30/10/5s), Patient (120/30/10s)
  • Fallback types: Service, Quality, Freshness
  • Degradation notification: Silent, Warning, Explicit, Interactive
  • Number of cascade levels: 2-4

Success Metrics

From research (PR #946):

  • Reliability Improvement: 95%+ vs 70-80% single approach
  • Graceful Degradation: 98% of failures handled successfully
  • User Impact: 90%+ users unaware of fallbacks occurring