🤖 Resilient AI Agent: GPT & Gemini Automatic Failover Chain

This n8n workflow creates a highly resilient AI Agent capable of handling API failures, rate limits, or outages by automatically failing over to alternative language models like GPT and Gemini. It uses a stateful counter (fail_count) initialized in the Agent Variables node to track attempts. A LangChain Code node (Fallback Models) dynamically selects the next model from a chain of connected AI nodes based on the fail count, routing the prompt to the appropriate one.

The core execution happens in the AI Agent node, which processes your custom prompt using the selected model. On success, the workflow completes seamlessly. On error, it triggers the error output, loops back to increment the fail count, and retries with the next model. This continues until all models are exhausted, at which point it fails gracefully. Ideal for production environments where reliability is paramount.

Benefits include zero-downtime AI operations, reduced manual intervention, and scalability across multiple providers. Use cases span healthcare patient management (e.g., resilient symptom analysis or report generation despite outages), customer support chatbots, data analysis pipelines, and any agentic AI workflow needing high availability. Compatible only with self-hosted n8n using community nodes; setup takes 3-5 minutes by adding and credentialing your AI models.