Skip to content

faint069/midjourney-api

 
 

Repository files navigation

midjourney-api

基于 Discord 的 Midjourney API。

添加 Midjourney 违禁词入口 issue

项目集成 Demo 参考:issue31

UML

sequenceDiagram
    participant ThirdServer
    participant APIServer
    participant DiscordAPI

    ThirdServer->>APIServer: 请求接口触发任务
    APIServer->>APIServer: 放入任务队列
    APIServer->>DiscordAPI: 调接口触发绘画任务
    APIServer-->>ThirdServer: 返回是否触发成功

    DiscordAPI->>DiscordAPI: 触发Midjourney bot绘画任务
    DiscordAPI->>DiscordAPI: 监听MidJourney bot消息
    DiscordAPI-->>ThirdServer: 返回监听实时消息
    DiscordAPI-->>APIServer: 清除队列任务
Loading

使用条件

  1. 确保程序启动环境能访问 Discord
  2. 已有 Midjourney、Discord 账户
  3. 创建 Discord 频道并添加机器人,参考教程 Midjourney|如何集成到自己的平台

安装启动

git clone
pip install -r requirements.txt

将文件.env.template重命名为.env,并填入参数值:

USER_TOKEN=用户token
BOT_TOKEN=机器人token
GUILD_ID=服务器ID
CHANNEL_ID=频道ID
CALLBACK_URL=回调地址,默认http post请求,用于接收 midjourney 作图进度和结果的服务

直接启动

# 启动监听机器人
python task_bot.py
# 启动http服务
python server.py

更新

git pull

# 启动监听机器人
python task_bot.py
# 启动http服务
python server.py

docker 启动

填写 start.sh-e 后的环境变量,直接启动:

sh start.sh

或者本地构建镜像:

# 构建镜像
sh build.sh
# 启动容器
sh start.sh

更新

docker rmi kunyu/midjourney-api:1.0
sh start.sh

接口swagger文档:http://127.0.0.1:8062/docs

midjourney-api 提供接口:

  • /v1/api/trigger/imagine:触发绘画任务(图生图,Prompt 前加上图片链接即可)
  • /v1/api/trigger/upscale:U
  • /v1/api/trigger/variation:V
  • /v1/api/trigger/reset:重绘
  • /v1/api/trigger/upload:上传图片
  • /v1/api/trigger/describe:通过上传图片名,生成 Prompt
  • /v1/api/trigger/message:发送图片消息,返回图片链接,用于图生图功能

使用

imagine

文生图

curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/imagine' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "prompt": "a cute cat"
}'

图生图,需带上图片 URL

curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/imagine' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "prompt": "a cute cat",
  "picurl": "https://xxxxxx/xxxxxxxxxxxx.jpg"
}'

upscale

curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/upscale' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "index": 1,
  "msg_id": "xxxxxxxxxx",
  "msg_hash": "xxxxx-xxx-xxxx-xxxx-xxxxxx",
  "trigger_id": "xxxxxxxxxx"
}'
  • index: 图片索引,取值:1、2、3、4
  • msg_id: imagine 绘画完成后回调报文 id 字段
  • msg_hash: imagine 绘画完成后回调报文 attachments[0].filename.split("_")[-1].split(".").[0]
  • trigger_id: imagine 绘画完成后回调报文 trigger_id 字段

variation

curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/variation' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "index": 2,
  "msg_id": "xxxxxxxxxx",
  "msg_hash": "xxxxx-xxx-xxxx-xxxx-xxxxxx",
  "trigger_id": "xxxxxxxxxx"
}'

reset

curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/reset' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "msg_id": "xxxxxxxxxx",
  "msg_hash": "xxxxx-xxx-xxxx-xxxx-xxxxxx",
  "trigger_id": "xxxxxxxxxx"
}'

describe

  1. 先上传图片
curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/upload' \
  -H 'accept: application/json' \
  -H 'Content-Type: multipart/form-data' \
  -F '[email protected];type=image/jpeg'
  1. 根据返回的图片文件名,调用 describe
curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/describe' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "upload_filename": "b56ca21a-5fbe-40b4-89ab-6e0aa732f561/9231228408.jpg",
  "trigger_id": "9231228408"
}'
  • trigger_id 先用 upload 返回的 trigger_id
  • upload_filename upload 返回的文件名

message

describe 一样,先 /v1/api/trigger/upload 上传图片,然后根据返回文件名,发送消息:

curl -X 'POST' \
  'http://127.0.0.1:8062/v1/api/trigger/message' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "upload_filename": "560a1e26-36a2-4d5f-a48d-9dd877642b51/7185811546.jpg"
}'

发送图片后,会返回图片链接。 该链接用于以图生图中,拼接 Prompt 形如 图片URL Prompt,调用 /v1/api/trigger/imagine

功能

  • imagine
  • upscale
  • variation
  • reset
  • describe
  • 图生图(获取到上传图片的链接)
  • 敏感词过滤上报
  • 任务队列(内存存储,不希望引入外建,可加入异常落盘)
  • tests

enjoy it

About

midjourney in discord api.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 96.3%
  • Shell 2.7%
  • Dockerfile 1.0%