http://akjamie.github.io/tags/resilience/ Recent content in Resilience on Jamie's Blog Hugo en-us Wed, 27 May 2026 00:00:00 +0000 http://akjamie.github.io/post/2026-05-27-llm-exception-handling-deep-dive/ Wed, 27 May 2026 00:00:00 +0000 http://akjamie.github.io/post/2026-05-27-llm-exception-handling-deep-dive/ <h1 id="exception-handling-for-llm-calls-in-production-agentic-systems">Exception Handling for LLM Calls in Production Agentic Systems</h1> <blockquote> <p>A deep dive into <code>agent/auxiliary_client.py</code> from Hermes Agent. Written for engineers building production agentic applications who need to go beyond &ldquo;catch Exception and retry.&rdquo;</p> </blockquote> <hr> <h2 id="why-llm-exception-handling-is-different">Why LLM Exception Handling Is Different</h2> <p>Most backend services talk to one upstream. When it fails, you retry or return an error.</p> <p>LLM-backed agents are different in three ways that make naive exception handling dangerous:</p> <ol> <li> <p><strong>The call is expensive.</strong> A failed LLM call that retried five times against a dead endpoint burned five API round-trips, five timeouts, and potentially five billing events — before the user saw anything.</p>