ComfyUI API Client

A comprehensive Node.js client library for interacting with ComfyUI APIs. This powerful and flexible library provides a robust set of tools for managing ComfyUI instances, building prompts, and executing workflows with ease and type safety.

Features
Installation
Core Concepts
Basic Usage
Advanced Usage
API Reference
Configuration
Error Handling
Best Practices
Contributing
License

Features

Multi-instance Support: Manage multiple ComfyUI instances with an intuitive connection pooling system.
Type-safe Prompt Building: Construct workflows with a powerful PromptBuilder that ensures type safety for inputs and outputs.
Real-time Updates: WebSocket support for receiving live updates during workflow execution.
Robust Error Handling: Automatic reconnection and comprehensive error management.
Authentication Support: Built-in support for basic authentication (ComfyUI behind NginX with basic auth).
Event-driven Architecture: Detailed event system for monitoring execution progress and handling various states.
Flexible Execution Modes: Choose from different execution modes to optimize for your specific use case.

Installation

Install the package using npm:

npm install @saintno/comfyui-sdk

Or using bun:

bun add @saintno/comfyui-sdk

Core Concepts

ComfyApi: The main client for interacting with a single ComfyUI instance.
ComfyPool: A manager for multiple ComfyUI instances, providing load balancing.
PromptBuilder: A utility for constructing workflows with type-safe inputs and outputs.
CallWrapper: A wrapper for API calls that provides comprehensive event handling and execution control.

Basic Usage

Setting up a ComfyApi instance

import { ComfyApi } from "@saintno/comfyui-sdk";

const api = new ComfyApi("http://localhost:8188", "client-id");
await api.init().waitForReady(); // Initialize websocket and wait for the client to be ready

Creating a PromptBuilder

import { PromptBuilder } from "@saintno/comfyui-sdk";
import workflowJson from "./workflow.json"; // Get from `Save (API Format)` or `Export (API Format)` from ComfyUI Web

const promptBuilder = new PromptBuilder(
  workflowJson,
  ["checkpoint", "positive", "negative", "seed", "steps"], // Input keys
  ["images"] // Output keys
)
  .setInputNode("checkpoint", "4.inputs.ckpt_name")
  .setInputNode("positive", "6.inputs.text")
  .setInputNode("negative", "7.inputs.text")
  .setInputNode("seed", "3.inputs.seed")
  .setInputNode("steps", "3.inputs.steps")
  .setOutputNode("images", "9");

Executing a Workflow

import { CallWrapper } from "@saintno/comfyui-sdk";

const workflow = promptBuilder
  .input(
    "checkpoint",
    "SDXL/realvisxlV40_v40LightningBakedvae.safetensors",
    /**
     * Use the client's osType to encode the path
     *
     * For example, if the client's `osType` is "nt" (Windows), the path should be encoded as below
     * "SDXL\\realvisxlV40_v40LightningBakedvae.safetensors"
     */
    api.osType // This is optional, but recommended it you want to support multiple platforms
  )
  .input("positive", "A beautiful landscape")
  .input("negative", "blurry, text")
  .input("seed", 42)
  .input("steps", 20);

new CallWrapper(api, workflow)
  .onStart((promptId) => console.log(`Task ${promptId} started`))
  .onProgress((info, promptId) =>
    console.log(`Task ${promptId} progress:`, info)
  )
  .onFinished((data, promptId) =>
    console.log(`Task ${promptId} finished:`, data)
  )
  .run();

You can use api.getCheckpoints() to get the list of available checkpoints.

const client = new ComfyApi("http://localhost:8188");
await client.getCheckpoints().then(console.log);

Advanced Usage

ComfyUI Manager

Support interact with ComfyUI-Manager extesion, require ComfyUI-Manager to be installed in the ComfyUI instance.

const api = new ComfyApi("http://localhost:8189").init();
/**
 * Should wait for the client to be ready for checking supported extensions
 */
await api.waitForReady();

if (api.ext.manager.isSupported) {
  /**
   * Get the list of all extensions
   */
  await api.ext.manager.getExtensionList().then(console.log);
  //More methods are available, check the `api.ext.manager` reference for more details
}

ComfyUI Crystools

Support listen event from ComfyUI-Crystools extension for tracking system resources, require ComfyUI-Crystools to be installed in the ComfyUI instance.

const api = new ComfyApi("http://localhost:8189").init();
/**
 * Should wait for the client to be ready for checking supported extensions
 */
await api.waitForReady();

if (api.ext.monitor.isSupported) {
  /**
   * Listen to the system monitor event
   */
  api.ext.monitor.on("system_monitor", (ev) => {
    console.log(ev.detail);
  });
  /**
   * Current monitoring data
   */
  console.log(api.ext.monitor.monitorData);
}

Connection Pooling

import { ComfyPool, EQueueMode } from "@saintno/comfyui-sdk";

const pool = new ComfyPool(
  [
    new ComfyApi("http://localhost:8188", "node-1"),
    new ComfyApi("http://localhost:8189", "node-2"),
  ],
  //   "PICK_ZERO", Picks the client which has zero queue remaining. This is the default mode. (For who using along with ComfyUI web interface)
  //   "PICK_LOWEST", Picks the client which has the lowest queue remaining.
  //   "PICK_ROUTINE", Picks the client in a round-robin manner.
  EQueueMode.PICK_ZERO
);

pool.run(async (api, clientIdx) => {
  // Your workflow execution logic here
});

// Or execute multiple jobs in parallel
pool.batch([
  (api) => executeWorkflow(api, params1),
  (api) => executeWorkflow(api, params2),
  // ...
]);

Authentication

const api = new ComfyApi("http://localhost:8188", "client-id", {
  credentials: {
    type: "basic",
    username: "your-username",
    password: "your-password",
  },
});

API Reference

ComfyApi

constructor(host: string, clientId?: string, opts?: { credentials?: BasicCredentials })
queuePrompt(number: number, workflow: object): Promise<QueuePromptResponse | false>
getQueue(): Promise<QueueResponse>
getHistories(maxItems?: number): Promise<HistoryResponse>
getSystemStats(): Promise<SystemStatsResponse>
uploadImage(file: Buffer | Blob, fileName: string, config?: { override?: boolean, subfolder?: string }): Promise<{ info: ImageInfo; url: string } | false>

ComfyPool

constructor(clients: ComfyApi[], mode: EQueueMode = EQueueMode.PICK_ZERO)
run<T>(job: (client: ComfyApi, clientIdx?: number) => Promise<T>, weight?: number): Promise<T>
batch<T>(jobs: Array<(client: ComfyApi, clientIdx?: number) => Promise<T>>, weight?: number): Promise<T[]>

PromptBuilder

constructor(prompt: T, inputKeys: I[], outputKeys: O[])
setInputNode(input: I, key: DeepKeys<T>): this
setOutputNode(output: O, key: DeepKeys<T>): this
input<V = string | number | undefined>(key: I, value: V): PromptBuilder<I, O, object>

CallWrapper

constructor(client: ComfyApi, workflow: T)
onPreview(fn: (ev: Blob, promptId?: string) => void): this
onPending(fn: (promptId?: string) => void): this
onStart(fn: (promptId?: string) => void): this
onFinished(fn: (data: Record<keyof T["mapOutputKeys"], any>, promptId?: string) => void): this
onFailed(fn: (err: Error, promptId?: string) => void): this
onProgress(fn: (info: NodeProgress, promptId?: string) => void): this
run(): Promise<Record<keyof T["mapOutputKeys"], any> | undefined | false>

Configuration

The library supports various configuration options, including:

Setting up multiple ComfyUI instances
Configuring authentication credentials
Choosing execution modes for the connection pool

Refer to the individual class constructors for specific configuration options.

Error Handling

The library provides comprehensive error handling through the event system. Use the onFailed method of CallWrapper to catch and handle errors during execution.

Best Practices

Always use PromptBuilder to ensure type safety when constructing workflows.
Utilize ComfyPool for managing multiple ComfyUI instances to improve reliability and load distribution.
Implement proper error handling using the provided event callbacks.
Use batch method of ComfyPool for parallel execution of multiple workflows.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Fork the repository
Create your feature branch (git checkout -b feature/AmazingFeature)
Commit your changes (git commit -m 'Add some AmazingFeature')
Push to the branch (git push origin feature/AmazingFeature)
Open a Pull Request

License

This project is licensed under the MIT License. See the LICENSE file for details.

This README provides a comprehensive overview of the ComfyUI API Client library, including detailed usage examples, API references, and best practices. It should give users a solid understanding of how to use and configure the library for their needs.

tctien342/comfyui-sdk