langgraph-fundamentals

Define a TypedDict state schema, write node functions that return partial state updates, wire them with add_edge or add_conditional_edges on a StateGraph, then call compile() before invoking. The same pattern applies in TypeScript using StateSchema and addNode.

Use LangGraph when you need fine-grained control over orchestration, complex workflows with branching or loops, or human-in-the-loop and persistence features. For quick prototyping or simple stateless workflows, LangChain agents or direct calls are a better fit.

Without a reducer, returning a list from a node overwrites the previous value. Annotate the field with operator.add in Python or use ReducedValue with a concat reducer in TypeScript so updates accumulate instead of replacing.

Call graph.stream with stream_mode set to messages, then iterate over the chunks where each item contains a token and metadata. For custom progress updates, use stream_mode custom with a stream writer inside nodes.

Command combines a state update and a single routing decision in one node return value. Send fans out from a conditional edge to spawn multiple parallel workers, each with its own input, and requires a reducer to aggregate their results.

Static edges that point back to earlier nodes create infinite cycles. Add a conditional edge that returns END when a termination condition is met, such as a counter exceeding a threshold.