LLM features fail in more ways than ordinary code. A provider times out, a model returns malformed output, a tool server goes down, or an agent repeats the same step without making progress. Reliable systems expect this. They retry within limits, fall back to another model or a cached answer, stop calling a failing dependency with a circuit breaker, and cap how long a run may go on.
Prompts matter too: a prompt treated as an interface, with defined inputs and outputs, behaves more consistently than a loose instruction. These guides cover building an agent with model fallback, the harness that handles failures around the model, detecting stuck loops, semantic caching with fallback, and circuit breakers for MCP servers.
The library lists all guides with filters for topic, level, category and reading time.