LLMOps is the operational discipline for LLM-based systems: versioning prompts and contexts (not just models), eval-gated releases, tracing and cost governance,
The operational discipline for LLM-based systems: versioning prompts and contexts (not just models), eval-gated releases, tracing and cost governance, provider and model change management, and incident response for behavioral regressions.
It adapts MLOps to a world where behavior changes through prompts and provider updates rather than retraining. That makes regression evals and rollback paths, not retraining pipelines, the core safety net of production LLM systems.