Plugin Tasks

Besides Midjourney, the gateway ships with plugin channels: Suno music, Kling/Hailuo/Vidu video, Veo, Jimeng, and more. Two calling patterns exist:

Pattern	Applies To	Flow
Generic task	Most plugins (model-driven)	POST /v1/tasks submit → GET /v1/tasks/{id} poll → fetch artifacts
Native passthrough	Suno, Kling, etc.	Vendor-native paths /suno/*, /kling/* — parameters match the vendor's official API

Generic Task Endpoints

Submit a Task

POST /v1/tasks

curl https://test.kailin.com.cn/v1/tasks \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-token" \
  -d '{
    "model": "hailuo-video",
    "prompt": "Bioluminescent jellyfish in the deep ocean",
    "duration": 6
  }'

Parameter	Type	Required	Description
model	string	Yes	Plugin model name — decides the routing
Other fields	-	No	Plugin-specific parameters (prompt, image URL, duration, aspect ratio…), passed through as-is

Response:

{ "task_id": "3e2c1a9b-xxxx-xxxx-xxxx" }

Query a Task

GET /v1/tasks/{task_id}

curl https://test.kailin.com.cn/v1/tasks/3e2c1a9b-xxxx \
  -H "Authorization: Bearer sk-your-token"

{
  "task_id": "3e2c1a9b-xxxx",
  "status": "SUCCESS",
  "data": {
    "video_url": "https://oss.example.com/v/xxx.mp4"
  }
}

status values: SUBMITTED / IN_PROGRESS / SUCCESS / FAILURE (some plugins extend their own states — trust the response).

Task Artifacts

Endpoint	Description
GET /v1/tasks/{key}/artifacts	List artifacts of a task (audio/video/image files)
GET /v1/tasks/{key}/artifacts/{artifact_key}/content	Download a specific artifact (binary stream)

curl "https://test.kailin.com.cn/v1/tasks/3e2c1a9b/artifacts" \
  -H "Authorization: Bearer sk-your-token"

curl "https://test.kailin.com.cn/v1/tasks/3e2c1a9b/artifacts/audio-0/content" \
  -H "Authorization: Bearer sk-your-token" \
  --output track.mp3

Suno (Music)

SunoAPI-format passthrough:

Endpoint	Description
POST /suno/submit/{action}	Submit a task — action is a SunoAPI action (music, lyrics, caption, …), parameters match SunoAPI docs
GET /suno/fetch/{task_id}	Query one task
POST /suno/fetch	Batch query — body takes ["id1", "id2"]

curl https://test.kailin.com.cn/suno/submit/music \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-token" \
  -d '{ "prompt": "An upbeat summer pop song", "make_instrumental": false, "wait_audio": false }'

The data field mirrors the raw Suno task structure. With wait_audio: true, the call waits until the audio is ready and returns audio URLs directly.

Kling (Video)

Kling-native passthrough:

Endpoint	Description
POST /kling/v1/videos/text2video	Text to video
POST /kling/v1/videos/image2video	Image to video
GET /kling/v1/videos/text2video/{task_id}	Query text-to-video task
GET /kling/v1/videos/image2video/{task_id}	Query image-to-video task

curl https://test.kailin.com.cn/kling/v1/videos/text2video \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-token" \
  -d '{
    "model_name": "kling-v1",
    "prompt": "Morning forest, sunlight through the leaves",
    "duration": 5,
    "mode": "std"
  }'

Parameters and responses match the Kling official API (model_name / mode / aspect_ratio / cfg_scale, etc.) — just swap the domain for the gateway.

Built-in Plugin Overview

The plugins below all submit via POST /v1/tasks (model set to the matching name) and query via GET /v1/tasks/{task_id}. Their parameters mirror the upstream vendor APIs and pass through untouched:

Plugin	Capability	Key Parameters
Hailuo Video	Text/image to video (MiniMax Hailuo)	prompt, first-frame image, duration
Hailuo H3	H3-model video tasks	Same as above
Google Veo	Veo video tasks (Gemini/Vertex channels)	prompt, duration, aspect ratio
Doubao	Jimeng video/image (ByteDance)	prompt, image, resolution
Alibaba Bailian	Tongyi Wanxiang image/video tasks	prompt, model, size
xAI	Grok Imagine video (generate/edit/extend)	prompt, source task ID
Vidu	Text/image to video	prompt, first-frame image, duration
Jimeng	Jimeng tasks	prompt, image

caution

Plugin endpoints pass vendor formats through — invalid parameters surface upstream errors directly. Always double-check the parameter docs in the Console's model gallery before submitting. Result URLs are temporary; re-host files promptly once SUCCESS.