Workflow DAG Engine

Visual workflow orchestration with 7 node types and conditional branching.

Overview

The Workflow Engine lets you build multi-step agent pipelines as a Directed Acyclic Graph (DAG). Each node performs an action (LLM call, tool execution, condition check, etc.) and passes output to the next node.

Node Types

Type	Description	Config
input	Entry point, receives input text	-
llm	Call an LLM with a prompt	prompt, model
tool	Execute a built-in tool	toolName, toolArgs
condition	Branch based on JS expression	condition, trueBranch, falseBranch
code	Execute JavaScript code	code
delay	Wait for specified milliseconds	delayMs
output	Format and return result	outputFormat

Quick Start

```ts import { workflowEngine } from "quark-agent";

// Create a workflow const wf = workflowEngine.createWorkflow("data-pipeline", "Process and summarize");

// Add nodes const inputNode = workflowEngine.addNode(wf.id, { type: "input", name: "raw_data", config: {}, position: { x: 0, y: 0 } }); const llmNode = workflowEngine.addNode(wf.id, { type: "llm", name: "summarize", config: { prompt: "Summarize this: ${raw_data}" }, position: { x: 200, y: 0 } }); const outputNode = workflowEngine.addNode(wf.id, { type: "output", name: "result", config: { outputFormat: "markdown" }, position: { x: 400, y: 0 } });

// Connect nodes workflowEngine.addEdge(wf.id, { from: inputNode.id, to: llmNode.id }); workflowEngine.addEdge(wf.id, { from: llmNode.id, to: outputNode.id });

// Execute const result = await workflowEngine.run(wf.id, "Some input data to process"); console.log(result.outputs.result); ```

Conditional Branching

ts const conditionNode = workflowEngine.addNode(wf.id, { type: "condition", name: "check_length", config: { condition: "context.input.length > 100", trueBranch: longSummaryNode.id, falseBranch: shortReplyNode.id, }, position: { x: 300, y: 100 } });

API Endpoints

Method	Path	Description
GET	/api/workflows	List workflows
POST	/api/workflows	Create workflow
GET/PUT/DELETE	/api/workflows/:id	CRUD
POST	/api/workflows/:id/run	Execute
POST	/api/workflows/:id/nodes	Add node