llm-fallback-chains

What problem does it solve?

This Skill addresses the unreliability of LLM services by providing robust fallback strategies, ensuring your AI applications remain available and functional even during outages or performance degradation of primary models.

Core Features & Use Cases

• Multi-Provider Failover: Automatically switch to alternative LLM providers when the primary one fails.
• Quality Degradation: Scale down to less powerful but more available models to maintain service.
• Circuit Breakers: Prevent overwhelming failing providers and allow them to recover.
• Health Monitoring: Proactively check the status of all LLM endpoints.
• Use Case: An e-commerce chatbot relies on a premium LLM for detailed product descriptions. If that LLM becomes unavailable, this skill can seamlessly switch to a faster, cheaper model for basic queries or provide a cached response, preventing a complete service outage.

Quick Start

Implement a fallback chain that prioritizes Anthropic's Sonnet, then GPT-4o, then Haiku, and finally GPT-4o-mini, with a circuit breaker threshold of 5 failures and a reset time of 60 seconds.