Crate rust_langgraph 

Rust LangGraph — community graph runtime for LLM apps

Rust LangGraph is an independent Rust library inspired by LangGraph (not affiliated with LangChain). It helps you build stateful, multi-actor LLM workflows with a graph execution model similar in spirit to Google’s Pregel, with checkpointing, streaming, and conditional routing.

Quick Start

use rust_langgraph::prelude::*;

#[derive(Clone, Debug, serde::Serialize, serde::Deserialize)]
struct MyState {
    count: i32,
}

impl State for MyState {
    fn merge(&mut self, other: Self) -> Result<(), Error> {
        self.count += other.count;
        Ok(())
    }
}

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let mut graph = StateGraph::new();
     
    graph.add_node("increment", |state: MyState, _config: &Config| async move {
        Ok(MyState { count: state.count + 1 })
    });
     
    graph.set_entry_point("increment");
    graph.set_finish_point("increment");
     
    let app = graph.compile(None)?;
    let result = app.invoke(MyState { count: 0 }, Config::default()).await?;
     
    println!("Final count: {}", result.count);
    Ok(())
}

Features

• Stateful Execution: Build graphs where nodes communicate through shared state
• Checkpointing: Save and resume execution at any point
• Streaming: Stream events as the graph executes
• Conditional Logic: Dynamic routing based on state
• Parallel Execution: Execute independent nodes concurrently
• Human-in-the-Loop: Interrupt and resume with human input
• LLM Integration: OpenAI, OpenRouter, Anthropic, and Ollama adapters (feature-gated)

More documentation

• README.md — full user guide (install, tutorial, API table, examples)
• AGENTS.md — cheat sheet for AI coding agents and contributors (correct crate name, features, patterns, pitfalls)

Re-exports

pub use config::Config;

pub use errors::Error;

pub use errors::Result;

pub use state::State;

pub use graph::StateGraph;

Modules

channels

Channels for state communication in graphs.

checkpoint

Checkpoint system for graph state persistence.

checkpoint_backends

Checkpoint backend implementations.

config

Configuration for graph execution.

errors

Error types for LangGraph operations.

graph

Graph builder API and compiled graph.

nodes

Node abstraction for graph execution.

pregel

Pregel execution engine and related types.

prelude

Re-exports of commonly used types for convenient imports

runtime

Runtime context for node execution.

state

State management and message handling.

types

Core types for LangGraph execution.