Agent progress
Show incremental progress while a multi-step graph executes.
AI Agents · Const
Streaming
LangGraph's stream API lets applications observe graph execution as it happens. Select a stream mode to receive state changes, full state snapshots, LLM tokens, custom events, checkpoints, tasks, or debug events.
How it works
LangGraph's stream method exposes graph execution as an async iterable. Choose a stream mode to receive state updates, complete state snapshots, LLM tokens, custom data, checkpoints, task events, or debugging information.
Syntax
await graph.stream(input, { streamMode: "updates" })Parameters
| Name | Type | Description |
|---|---|---|
input | Graph input | Initial graph state supplied to the stream invocation. |
config | Stream options | Options including the selected stream mode, such as updates, values, messages, custom, checkpoints, tasks, or debug. |
Returns
An async iterable of streamed graph execution output.
Key points
- Call
graph.stream(...)and consume the returned async iterable withfor await...of. - Choose
updatesfor incremental state changes orvaluesfor the full state after each graph step. - The
updatesmode reports only what changed after each execution step. - The
valuesmode reports the accumulated state snapshot after each execution step. - Use
messageswhen the application needs to render model output progressively. - Available modes cover state, model messages, custom events, checkpoints, task events, and debugging.
Common mistakes
- Use
messagesfor token-level LLM output; it is different from state streaming modes such asupdatesandvalues.
Example: Stream graph updates
import { StateGraph, Annotation } from "@langchain/langgraph";
const State = Annotation.Root({
count: Annotation<number>(),
});
const graph = new StateGraph(State)
.addNode("increment", (state) => ({ count: state.count + 1 }))
.addEdge("__start__", "increment")
.addEdge("increment", "__end__")
.compile();
const stream = await graph.stream(
{ count: 0 },
{ streamMode: "updates" }
);
for await (const chunk of stream) {
console.log(chunk);
}Runs a graph and prints each node's state update as it becomes available.
When to use it
Token streaming
Render LLM responses token by token in a chat interface.
Custom events
Send application-specific events from graph nodes to a client.
Observability
Inspect checkpoints, tasks, or detailed runtime events during execution.
Related
streamModeupdatesvaluesmessagescustomcheckpointstasksdebug