Maliang Image — AI Image Generation & Editing

Provider: Maliang API (nano.djdog.ai)

Model: Google Gemini (image generation & editing)

Pricing: $0.60/image, free $6 credit on first use (~10 images)

What it does

Generate images from text descriptions or edit existing images using AI.

• No images provided → text-to-image generation
• Images provided → AI image editing / inpainting (1–10 reference images, user-provided only)

Security & Data Handling

• API Key: Auto-provisioned on first use via POST /api/v1/provision. The key is stored only in your local environment (MALIANG_API_KEY). You can revoke it at any time.
• Images: Only images explicitly provided by the user are sent to the API. The skill never reads files unless the user specifies the file path. All uploads go over HTTPS.
• No data collection: The API processes your request and returns results. Images are not stored permanently on the server.

Inputs needed

Auto-Provision (First-time Setup)

On first use (when MALIANG_API_KEY is not set), automatically provision a new account:

curl -s -X POST "https://nano.djdog.ai/api/v1/provision"

Response:

{
  "success": true,
  "data": {
    "api_key": "sk-...",
    "short_code": "ML-XXXXXX",
    "balance": 6.0,
    "currency": "USD",
    "recharge_url": "https://nano.djdog.ai/recharge/ML-XXXXXX",
    "message": "Your API key has been created with $6 initial credit (~10 images)..."
  }
}

Save the returned api_key as MALIANG_API_KEY for all subsequent requests.

Tell the user their short code (ML-XXXXXX) and initial balance — they'll need the short code to recharge later.

Balance Check

After each task, check remaining balance:

curl -s "https://nano.djdog.ai/api/v1/balance" \
  -H "Authorization: Bearer ${MALIANG_API_KEY}"

Response includes short_code for recharge reference. If balance is low, remind the user:

> Your balance is $X.XX. To recharge, visit: https://nano.djdog.ai/recharge/{short_code}

Workflow

Step 0 — Ensure API key

If MALIANG_API_KEY is not set, run Auto-Provision first (see above) and store the returned key.

Step 1 — Determine mode

• If the user provides one or more images (file paths, URLs, or pasted base64): edit mode
• Otherwise: generate mode

Step 2 — Prepare images (edit mode only)

For each image the user provides:

1. If it is a local file path, read and base64-encode it.
2. If it is a URL, download it first, then base64-encode.
3. Strip any data:image/...;base64, prefix — the API accepts raw base64.
4. Verify each image is under 10 MB after decoding.
5. Maximum 10 images total.

Step 3 — Submit task

Generate mode — call:

curl -s -X POST "https://nano.djdog.ai/api/v1/generate" \
  -H "Authorization: Bearer ${MALIANG_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "<user prompt>",
    "aspect_ratio": "<ratio, default 1:1>"
  }'

Edit mode — call:

curl -s -X POST "https://nano.djdog.ai/api/v1/edit" \
  -H "Authorization: Bearer ${MALIANG_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "<user editing instruction>",
    "image": "<base64 string or array of base64 strings>",
    "aspect_ratio": "<ratio, optional>"
  }'

Both return:

{
  "success": true,
  "data": {
    "task_id": "...",
    "status": "pending",
    "created_at": "..."
  }
}

Extract task_id from the response.

Step 4 — Poll for result

Poll every 3 seconds, up to 120 seconds max:

curl -s "https://nano.djdog.ai/api/v1/tasks/${TASK_ID}" \
  -H "Authorization: Bearer ${MALIANG_API_KEY}"

Response data.status values:

Step 5 — Deliver result

When status is completed:

• If image_url is present: show the URL to the user (preferred).
• If only image_base64 is present: save to a local file and show the path.

Output format

Image generated successfully!
URL: https://...
Aspect ratio: 1:1
Prompt: "<original prompt>"

Guardrails

• Never fabricate task IDs or image URLs. Only use values from API responses.
• Never poll more than 40 times (120 seconds). If not completed, tell the user it is still processing and provide the task ID for manual checking.
• Do not send images larger than 10 MB to the edit endpoint.
• Do not send more than 10 images to the edit endpoint.
• If the API returns 402 (INSUFFICIENT_BALANCE), tell the user their balance is low and they need to recharge.
• Prompt max length is 4000 characters. If the user's prompt is longer, ask them to shorten it.

Failure handling

Examples

Text-to-image:

User: Generate a cute orange cat sitting on a windowsill at sunset, anime style
→ Auto-provision if no API key → POST /api/v1/generate with prompt → poll for result → return image URL

Image editing:

User: Change the background of this photo to a beach scene [attaches photo]
→ Base64-encode the photo
→ POST /api/v1/edit with prompt + image, poll for result, return image URL

Multi-image editing:

User: Combine these character designs into one group portrait [attaches 3 images]
→ Base64-encode all 3 images
→ POST /api/v1/edit with prompt + image array, poll for result, return image URL