获取开发者密钥

此密钥为平台API操作唯一标识，务必保证不要泄露

❗️登录OneThingAI平台，进入账户中心，选择开发者密钥自助创建密钥

ComfyOne简介

ComfyOne可以将多个ComfyUI应用合并统一调度使用，支持多工作流可以并发执行功能 ，可自行管理应用实例，上下线实例能力

域名: https://pandora-server-cf.onethingai.com

必须请求头信息

ComfyOne 整体使用流程

也可以使用以下的sdk来操作ComfyOne API

python sdk （需要使用更复杂的任务调度的，可以使用sdk提供的节点scheduler 模块来实现scheduler 例子）

typescript sdk (例子参考 examples/main.ts)

版本更新说明

### v1.1.5版本 + 优化程序 + 限制每个用户最大连接 ws 数

v1.1.4 版本

• 支持自定义工作流任务

v1.1.3 版本

• 修复查询任务状态接口偶尔出现 404 问题

• 支持查询历史任务列表接口

v1.1.2 版本

• 修复调度工作流节点 ID为非数字问题

• 修复部分任务失败未通知问题

• 任务状态接口返回任务耗时

v1.1.1 版本

• 支持系统环境变量输入

• 修复任务执行后不更新状态问题

• 修复部分失败任务消息通知丢失问题

• 任务超时无法被调度，自动失败能力

v1.1.0 版本

• 支持 WebSocket 任务状态通知

• 支持运行中任务也可以取消

• 任务执行错误返回执行错误节点

注意：只保留 3 天内的任务和结果数据，如有需要麻烦自行在有效期内保存好

系统变量

系统变量，在执行前系统会自动更新为对应的数据

{COMFYONE:TASK_ID} 任务ID
{COMFYONE:WORKFLOW_ID} 工作流ID

示例

提交任务信息

// 提交任务的参数为
{
  "workflow_id": "test123",
  "inputs":[
    {"id": "8", "task_id": "{COMFYONE:TASK_ID}", "workflow_id":"{COMFYONE:WORKFLOW_ID}"}
  ],
  ...
}
// {"code":0,"msg":"success","data":{"task_id":"task456"}}

真实提交到后端为

{
  "inputs":[
    {"id": "8", "task_id": "task456", "workflow_id":"test123"}
  ],
  ...
}

开放接口

GET 实例列表

GET /v1/backends

返回示例

200 Response

{
  "code": 0,
  "msg": "string",
  "data": [
    {
      "name": "string",
      "tasks": 0,
      "is_live": true,
      "is_down": true
    }
  ]
}

返回结果

返回数据结构

状态码 200

POST 添加实例

POST /v1/backends

Body 请求参数

{"instance_id": "string" }

如何获取实例 ID

1. 点击我的应用

2. 找到创建好的 ComfyUI 应用

3. 在应用 ID 处，点击复制按钮

请求参数

返回示例

200 Response

{
  "code": 0,
  "msg": "string",
  "data": {
    "name": "string",
    "tasks": 0,
    "is_live": true,
    "is_down": true
  }
}

返回结果

返回数据结构

状态码 200

PATCH 上下线实例

PATCH /v1/backends/{name}

Body 请求参数

{
  "state": "up"
}

请求参数

枚举值

返回示例

200 Response

{
  "code": 0,
  "msg": "string",
  "data": {
    "name": "string",
    "tasks": 0,
    "is_live": true,
    "is_down": true
  }
}

返回结果

返回数据结构

状态码 200

DELETE 删除实例

DELETE /v1/backends/{name}

请求参数

返回示例

200 Response

{
  "code": 0,
  "msg": "string",
  "data": {
    "name": "string",
    "tasks": 0,
    "is_ive": true,
    "is_down": true
  }
}

返回结果

返回数据结构

状态码 200

POST 创建工作流

POST /v1/workflows

Body 请求参数

示例请求数据

{
    "name": "test",
    "inputs": [
        {
            "id": "5",
            "name": "width",
            "type": "number"
        },
        {
            "id": "5",
            "name": "height",
            "type": "number"
        }
    ],
    "outputs": [
        "9"
    ],
    "workflow": {
        "3": {
            "inputs": {
                "seed": 156680208700286,
                "steps": 20,
                "cfg": 8,
                "sampler_name": "euler",
                "scheduler": "normal",
                "denoise": 1,
                "model": [
                    "4",
                    0
                ],
                "positive": [
                    "6",
                    0
                ],
                "negative": [
                    "7",
                    0
                ],
                "latent_image": [
                    "5",
                    0
                ]
            },
            "class_type": "KSampler"
        },
        "4": {
            "inputs": {
                "ckpt_name": "AWPainting_1.4.safetensors"
            },
            "class_type": "CheckpointLoaderSimple"
        },
        "5": {
            "inputs": {
                "width": 512,
                "height": 512,
                "batch_size": 1
            },
            "class_type": "EmptyLatentImage"
        },
        "6": {
            "inputs": {
                "text": "beautiful scenery nature glass bottle landscape, , purple galaxy bottle,",
                "clip": [
                    "4",
                    1
                ]
            },
            "class_type": "CLIPTextEncode"
        },
        "7": {
            "inputs": {
                "text": "text, watermark",
                "clip": [
                    "4",
                    1
                ]
            },
            "class_type": "CLIPTextEncode"
        },
        "8": {
            "inputs": {
                "samples": [
                    "3",
                    0
                ],
                "vae": [
                    "4",
                    2
                ]
            },
            "class_type": "VAEDecode"
        },
        "9": {
            "inputs": {
                "filename_prefix": "ComfyUI",
                "images": [
                    "8",
                    0
                ]
            },
            "class_type": "SaveImage"
        }
    }
}

请求参数

枚举值

请求示例

const fetch = require('node-fetch');
const workflowName = 'test';
const inputs = [{id: 10, type: 'image', name: 'image'}...];
const outputs = ['13'];
const workflowData = {"10": {inputs: {image: "", upload: "image"},"class_type": "LoadImage"},...};
const run = async function createWorkflow(name, inputs, outputs, workflow){
    const url = 'https://pandora-server-cf.onethingai.com/v1/workflows'
    const resp = await fetch(url, {
      headers: {
        "accept": 'application/json',
        "content-type": 'application/json',
        "Authorization": `Bearer ${process.env.ONETHINGAI_TOKEN}`
      },
      body: JSON.stringify({ name, inputs, outpus, workflow }),
      method: "POST"
    });

    return await resp.json()
}
console.log(run(workflowName, inputs, outputs, workflowData))

返回示例

200 Response

{
  "code": 0,
  "msg": "Success",
  "data": {
    "id": "667d3a99c8e1118ea12be2b9"
  }
}

返回结果

返回数据结构

状态码 200

如何获取工作流信息

1. 打开 ComfyUI 界面，点击如图1所示的设置

2. 在设置界面找到开启开发模式，勾选，如图2所示

3. 配置好的工作流后，点击保存API格式，如图3所示

4. 打开下载的文件，拷贝里面的所有内容，这就是工作流的真实信息

图1

图2

图3

PATCH 修改工作流

PATCH /v1/workflows/{id}

Body 请求参数

{
  "inputs": [
    {
      "id": "5",
      "name": "width",
      "type": "number"
    }
  ]
}

请求参数

请求示例

const fetch = require('node-fetch');
const workflowId = '667d3a99c8e1118ea12be2b9';
const inputs = [{id: 10, type: 'image', name: 'image'}...];
const outputs = ['13'];
const workflowData = {"10": {inputs: {image: "", upload: "image"},"class_type": "LoadImage"},...};
const run = async function createWorkflow(workflowId, inputs, outputs, workflow){
    const url = `https://pandora-server-cf.onethingai.com/v1/workflows/${workflowId}`
    const resp = await fetch(url, {
      headers: {
        "accept": 'application/json',
        "content-type": 'application/json',
        "Authorization": `Bearer ${process.env.ONETHINGAI_TOKEN}`
      },
      body: JSON.stringify({ inputs, outpus, workflow }),
      method: "PATCH"
    });

    return await resp.json()
}
console.log(run(workflowId, inputs, outputs, workflowData))

返回示例

200 Response

{
  "code": 0,
  "msg": "Success",
  "data": {
    "id": "667d3a99c8e1118ea12be2b9"
  }
}

返回结果

返回数据结构

状态码 200

DELETE 删除工作流

DELETE /v1/workflows/{id}

请求参数

请求示例

const fetch = require('node-fetch');
const workflowId = '667d3a99c8e1118ea12be2b9';
const run = async function createWorkflow(workflowId){
    const url = `https://pandora-server-cf.onethingai.com/v1/workflows/${workflowId}`
    const resp = await fetch(url, {
      headers: {
        "accept": 'application/json',
        "content-type": 'application/json',
        "Authorization": `Bearer ${process.env.ONETHINGAI_TOKEN}`
      },
      method: "DELETE"
    });

    return await resp.json()
}
console.log(run(workflowId))

返回示例

200 Response

{
  "code": 0,
  "msg": "Success",
}

返回数据结构

状态码 200

GET 工作流列表

GET /v1/workflows

请求参数

请求示例

const fetch = require('node-fetch');
const run = async function listWorkflows(page, pageSize){
    const url = `https://pandora-server-cf.onethingai.com/v1/workflows?page=${page}&page_size=${pageSize}`
    const resp = await fetch(url, {
      headers: {
        "accept": 'application/json',
        "content-type": 'application/json',
        "Authorization": `Bearer ${process.env.ONETHINGAI_TOKEN}`
      },
      method: "GET"
    });

    return await resp.json()
}
console.log(run(1, 10))

返回示例

200 Response

{
  "code": 0,
  "msg": "string",
  "data": {
    "list": [
      {
        "id": "string",
        "name": "string",
        "inputs": [
          {
            "id": null,
            "name": null,
            "type": null
          }
        ],
        "outputs": [
          "string"
        ],
        "total_nodes": 0,
        "updated_at": "string",
        "created_at": "string"
      }
    ],
    "pagination": {
      "page": 1,
      "page_size": 10,
      "total": 1
    }
  }
}

返回结果

返回数据结构

状态码 200

GET 单个工作流

GET /v1/workflows/{id}

请求参数

请求示例

const fetch = require('node-fetch');
const workflowId = '667d3a99c8e1118ea12be2b9';
const run = async function createWorkflow(workflowId){
    const url = `https://pandora-server-cf.onethingai.com/v1/workflows/${workflowId}`
    const resp = await fetch(url, {
      headers: {
        "accept": 'application/json',
        "content-type": 'application/json',
        "Authorization": `Bearer ${process.env.ONETHINGAI_TOKEN}`
      },
      method: "GET"
    });

    return await resp.json()
}
console.log(run(workflowId))

返回示例

200 Response

{
  "code": 0,
  "msg": "Success",
  "data": {
    "id": "667d3a99c8e1118ea12be2b9",
    "name": "test",
    "inputs": [
      {
        "id": 10,
        "name": "image",
        "type": "image"
      }
    ],
    "outputs": [
      "13"
    ],
    "total_nodes": 5,
    "workflow": "{\"10\":{\"inputs\":{\"image\":\"9b6811d2da10e59ac2b830bf57c5eaf9ca96374e0be00aeafbbbf4cad14e1a8c.jpg\",\"upload\":\"image\"},\"class_type\":\"LoadImage\"},...\"13\":{\"inputs\":{\"images\":[\"12\",0]},\"class_type\":\"PreviewImage\"}}",
    "updated_at": "2024-07-08T02:15:02.231Z",
    "created_at": "2024-07-08T02:15:02.231Z"
  }
}

返回结果

返回数据结构

状态码 200

POST 生成任务

POST /v1/prompts

Body 请求参数

以下是使用创建工作流示例，并将潜空间设置为 1024x1024 的任务

{
  "workflow_id": "667d3a99c8e1118ea12be2b9",
  "inputs": [
    {
      "id": "5",
      "params": {
        "width":  1024,
        "height": 1024
      }
    }
  ],
  "free_cache": false
}

请求参数

请求示例

const fetch = require('node-fetch');
const params = {
  workflow_id: '667d3a99c8e1118ea12be2b9',
  inputs: [
    {
      id: "10",
      params: {
        image: 'https://pandora-server-cf.onethingai.com/v1/files/inputs/9b/68/9b6811d2da10e59ac2b830bf57c5eaf9ca96374e0be00aeafbbbf4cad14e1a8c.jpg'
      }
    }
  ]
}
const run = async function prompts(params){
    const url = 'https://pandora-server-cf.onethingai.com/v1/prompts'
    const resp = await fetch(url, {
      headers: {
        "accept": 'application/json',
        "content-type": 'application/json',
        "Authorization": `Bearer ${process.env.ONETHINGAI_TOKEN}`
      },
      body: JSON.stringify({ ...params }),
      method: "POST"
    });

    return await resp.json()
}
console.log(run(params))

返回示例

200 Response

{
  "code": 0,
  "msg": "Success",
  "data": {
    "taskId": "667d3a99c8e1118ef12be2b9"
  }
}

返回结果

返回数据结构

状态码 200

POST 创建工作流任务

POST /v1/prompts_workflow

Body 请求参数

说明： 该接口是创建工作流+生成任务的合并接口，在不需要复用工作流的情况下可使用该接口

示例请求数据

{
    "name": "test",
    "inputs": [
      {
        "id": "5",
        "params": {
          "width":  1024,
          "height": 1024
        }
      }
    ],
    "outputs": [
        "9"
    ],
    "workflow": {
        "3": {
            "inputs": {
                "seed": 156680208700286,
                "steps": 20,
                "cfg": 8,
                "sampler_name": "euler",
                "scheduler": "normal",
                "denoise": 1,
                "model": [
                    "4",
                    0
                ],
                "positive": [
                    "6",
                    0
                ],
                "negative": [
                    "7",
                    0
                ],
                "latent_image": [
                    "5",
                    0
                ]
            },
            "class_type": "KSampler"
        },
        "4": {
            "inputs": {
                "ckpt_name": "AWPainting_1.4.safetensors"
            },
            "class_type": "CheckpointLoaderSimple"
        },
        "5": {
            "inputs": {
                "width": 512,
                "height": 512,
                "batch_size": 1
            },
            "class_type": "EmptyLatentImage"
        },
        "6": {
            "inputs": {
                "text": "beautiful scenery nature glass bottle landscape, , purple galaxy bottle,",
                "clip": [
                    "4",
                    1
                ]
            },
            "class_type": "CLIPTextEncode"
        },
        "7": {
            "inputs": {
                "text": "text, watermark",
                "clip": [
                    "4",
                    1
                ]
            },
            "class_type": "CLIPTextEncode"
        },
        "8": {
            "inputs": {
                "samples": [
                    "3",
                    0
                ],
                "vae": [
                    "4",
                    2
                ]
            },
            "class_type": "VAEDecode"
        },
        "9": {
            "inputs": {
                "filename_prefix": "ComfyUI",
                "images": [
                    "8",
                    0
                ]
            },
            "class_type": "SaveImage"
        }
    }
}

请求参数

枚举值

返回示例

200 Response

{
  "code": 0,
  "msg": "Success",
  "data": {
    "taskId": "667d3a99c8e1118ef12be2b9"
  }
}

返回结果

返回数据结构

状态码 200

GET 查看任务的状态

GET /v1/prompts/{id}/status

请求参数

返回示例

200 Response

{
  "code": 0,
  "msg": "string",
  "data": {
    "status": "pendding",
    "message": "string",
    "position": 0,
    "images": [
      "string"
    ],
    "text": [
      "string"
    ],
    "tags": [
      "string"
    ]
  }
}

返回结果

返回数据结构

状态码 200

枚举值

错误信息

添加节点错误信息：Node xxx#123 节点执行错误信息

GET 查看历史任务列表

GET /v1/prompts

请求参数

返回示例

200 Response

{
  "code": 0,
  "msg": "success",
  "data": {
    "list": [
      {
        "task_id": "676526a12a9f535e8d324b26",
        "state": "finished",
        "backend": "test",
        "workflow_id": "676404f6ad6f29dad4eb5aa6",
        "prompt_id": "b8f1a085-1db1-4551-91de-8b5cab516587",
        "dispatched_at": "2024-12-20 08:12:37",
        "created_at": "2024-12-20 08:11:13"
      }
      ...
    ],
    "pagination": {
      "total": 20,
      "page": 1,
      "page_size": 10
    }
  }
}

返回结果

返回数据结构

状态码 200

枚举值

POST 取消生成

POST /v1/prompts/{id}/cancel

请求参数

返回示例

200 Response

{
  "code": 0,
  "msg": "string"
}

返回结果

POST 上传图片

POST /v1/files

Body 请求参数

file: string

请求参数

返回示例

200 Response

{
  "code": 0,
  "msg": "string",
  "data": {
    "path": "string"
  }
}

返回结果

返回数据结构

状态码 200

任务通知接口

GET /v1/ws

该接口为 websocket 协议，支持长链接通知，每个用户最多只有 15 个连接

注：由于接口是在有任务信息时才会发送消息，空闲情况下会无信息发送，这样会导致部分 websocket 的库认为读消息超时了，这里建议使用websocket 的原生 ping pong 来保活，可以避免被库给关闭了

ping pong的示例代码

ctx, cancel := context.WithCancel(context.Background())
maxLastPongMessage := int32(0)
ws.SetPongHandler(func(_ string) {
    atomic.AddInt32(&maxLastPongMessage, -1)
})
// 定时发 ping 消息
go func() {
	defaultInterval := 5 * time.Second
	ticker := time.NewTimer(defaultInterval)
	defer ticker.Stop()
	for {
		select {
		case <-ticker.C:
			err := ws.WriteControl(websocket.PingMessage, nil, time.Now().Add(time.Second))
			if err != nil {
				klog.V(3).InfoS("websocket send ping message has failure", "error", err, "clientIp", clientAddr, "uid", uid)
				cancel()
				return
			}
            // 超过 3 次没有响应 pong 消息了
            if atomic.AddInt32(&maxLastPongMessage, 1) > 3 {
                cancel()
                return
            }
			ticker.Reset(defaultInterval)
		case <-ctx.Done():
			return
		}
	}
}()

调用示例

const ws = new WebSocket('wss://pandora-server-cf.onethingai.com/v1/ws', {
    headers: {
        'Authorization': `Bearer ${process.env.ONETHINGAI_TOKEN}`
    }
});

// 接收消息的回调
ws.onmessage = (event) => {
    console.log(data);
    const data = JSON.parse(event.data);
    if (data.type == 'pendding') { 
        // TODO: 需要处理任务等待执行
        console.log(`任务 ${data.task_id} 等待执行, 当前位置: ${data.data.current}`);
    }
    if (data.type == 'progress') { 
        // TODO: 需要处理任务正在执行
        console.log(`任务 ${data.task_id} 正在执行, 进度: ${data.data.process}%`);
    }
    if (data.type == 'finished') {
        // TODO: 需要处理任务已完成
        if (data.data.success) {
          // TODO: 任务处理全部完成,可以通过认为状态接口获取输出内容
        }
    }
    if (data.type == 'error') {
        // TODO: 需要处理任务执行出错
        console.error('任务执行出错:', data.data.message);
    }
}

// 连接建立时的回调
ws.onopen = () => {
    console.log('WebSocket 连接已建立');
}

// 连接关闭时的回调
ws.onclose = () => {
    console.log('连接已关闭');
}

// 发生错误时的回调
ws.onerror = (error) => {
    console.error('WebSocket 错误:', error);
}

消息通知数据结构

返回消息示例

{ "type": "pendding","taskId": "667d3a99c8e1118ef12be2b9", "data": {"message": "等待中...", "current": 1}}
{ "type": "progress","taskId": "667d3a99c8e1118ef12be2b9", "data": {"message": "执行中...", "process": 0}}
....
{ "type": "progress","taskId": "667d3a99c8e1118ef12be2b9", "data": {"message": "执行中...", "process": 30}}
....
{ "type": "finished","taskId": "667d3a99c8e1118ef12be2b9", "data": {"message": "完成", "success": true}}

数据模型

Workflow

{
  "class_type": "string",
  "inputs": {}
}

属性

TaskStatus

{
  "status": "pendding",
  "message": "string",
  "position": 0,
  "images": [
    "string"
  ],
  "text": [
    "string"
  ],
  "tags": [
    "string"
  ]
}

属性

枚举值

Response

{
  "code": 0,
  "msg": "string"
}

属性