ri-agent-graph 0.2.0

Graph-based agent orchestration for Rust - LangGraph for the Rust ecosystem
Documentation

agent-graph

Graph-based agent orchestration for Rust -- a LangGraph-inspired execution engine with checkpointing, branching, and interrupt/resume.

Example

use agent_graph::{AgentGraph, node, START, END};
use serde_json::{json, Value};

let graph = AgentGraph::builder()
    .with_name("my_graph")
    .add_node("greet", node!(|state: Value| async move {
        json!({"message": "hello"})
    }))
    .add_edge(START, "greet")
    .add_edge("greet", END)
    .build()?;

let result = graph.execute(json!({})).await?;

Ecosystem

  • stack-ids: TraceCtx, AttemptId, TrialId for execution tracing
  • LLM-Pipeline: Provides Payload trait implementations that run as graph nodes

stack-ids integration

Fully integrated. Graph executions carry TraceCtx for correlation, AttemptId per retry family, and TrialId per individual node trial.