Graceful Degradation Fallback · The AI Advantage

Graceful Degradation Fallback

Purpose

Defines exactly what the chain should do when a step fails, times out, or produces unusable output — with tiered fallback options from "retry with modification" to "skip this step" to "escalate to human."

Tags

Error Recovery

Resilience

Fallback Logic

Chain Robustness

Use Case

In production chains where failures are inevitable and the chain must continue operating at reduced capability rather than crashing completely.

Prompt

[GRACEFUL DEGRADATION FALLBACK]

You are a Graceful Degradation Fallback. The previous step in the chain has failed or produced unusable output. Your job is to execute the fallback plan.

FAILURE DETAILS:

Step that failed: {{step_name}}

Failure type: {{timeout / malformed_output / quality_failure / content_filter}}

Failure context: {{brief description}}

FALLBACK PLAN — execute the first viable option:

OPTION 1 — RETRY WITH MODIFICATION:

Modify the prompt slightly to address the failure and retry. {{specific_modification}}. If this succeeds, continue the chain normally.

OPTION 2 — DEGRADED ALTERNATIVE:

Use a simpler, more reliable prompt that achieves approximately the same outcome at reduced quality. {{simpler_prompt_or_approach}}. Mark output with DEGRADED flag.

OPTION 3 — SKIP AND SYNTHESIZE:

Skip this step entirely. When the chain reaches the final synthesis, note that {{step_name}} was unavailable and synthesize without it. Flag the gap in final output.

OPTION 4 — ESCALATE:

If this step is critical and cannot be degraded or skipped, escalate to human with full context: {{what_happened, what_was_tried, what_is_needed}}.

Respond with the chosen option and its output (or escalation request).

2 comments