A Fallback Chain is a systematic strategy for managing failures by defining ordered, pre-validated sequences of actions when a primary tool or path fails.
Core Principle
Move from ad-hoc error handling to a deterministic flow:
- Attempt primary tool.
- Attempt pre-validated alternatives or cached results.
- Escalate to human/system if all alternatives fail.
Implementation Rules
- Pre-validation: All fallback mechanisms MUST be tested under realistic failure conditions; an untested fallback is not a safety net.
- Explicit Signaling: Systems must explicitly signal any degradation (e.g., partial results, fallback usage) to downstream consumers to prevent them from treating degraded output as authoritative.
- Escalation Decision:
- Low Risk/Reversible: Degrade and continue.
- High Risk/Irreversible: Abort and escalate.