Single-agent loop

An LLM-driven agent combines structured control with probabilistic reasoning. The loop remains explicit, but decisions are delegated to a model.

• Replacing deterministic decisions with LLM reasoning — Substituting rules with model-based choices.
• Implementing an LLM-driven agent — Integrating prompts, tools, and state.
• Comparing agent types — Contrasting deterministic and LLM-based approaches.