Long-Running Agents: State, Resumption, and the Database You Need First
The plumbing that turns a demo agent into a product agent. State, resumption, idempotency, and why you need a database before you need a graph framework.
Search for a command to run...
Articles tagged with #agents
The plumbing that turns a demo agent into a product agent. State, resumption, idempotency, and why you need a database before you need a graph framework.
Every framework wants to sell you a team of specialised agents. Most teams would be better served by one smarter agent. Here is the honest line between a good multi-agent system and a badly abstracted one.
ReAct, reflection, plan-and-execute. Three agent patterns the papers love. An honest look at which ones survive contact with production and which are expensive optional extras.
Every agent framework wraps the same small loop. Here it is, written by hand, so you know what you are actually buying when you adopt LangChain or LangGraph.
Every 'agent framework' wraps the same small idea: the model produces structured output that names a function, and your code runs the function. Here is the idea, stripped of the marketing.
The difference between a chatbot and an agent is two things: a set of tools it can call, and a loop. That's it. Everything else is plumbing. Here is why the plumbing is hard.