Graph

A Graph workflow is a Directed Acyclic Graph (DAG) consisting of nodes and edges, with each node representing a task and each edge indicating the sequence of execution. Unlike the linear sequence of workflows in a pipeline, a graph provides a more complex and flexible structure, allowing for workflows to have multiple predecessors and successors. This makes a Graph workflow suitable for use cases where the processing logic isn't strictly linear.

The Graph workflow provides a powerful way to manage complex workflows that can't be adequately represented in a linear sequence, making it easier to build, understand, and debug intricate sequences of operations. The shared context between nodes simplifies data passing and communication between different parts of the workflow.

Example

import type { Metadata } from "https://deno.land/x/scribeai/types.ts";
import {
  Graph,
  runWorkflow,
  Scribe,
} from "https://deno.land/x/scribeai/mod.ts";

// Create a new Scribe instance
const scribe = new Scribe();

// Create a new graph
const myGraph = await scribe.createGraph<MyCtx, Metadata>({
  name: "test",
  version: "1.0.0",
  tags: ["test"],
  metadata: { key: "value" },
  ops: {
    // Set up a dummy runFor operation workflow (a Task in this case)
    runFor: async (_ctx, next) => {
      await next();
    },
  },
});

// Run the graph
await runWorkflow(myGraph, ctx);