aihubmix-image

name	aihubmix-image
description	Generate images with Aihubmix APIs, especially openai-compatible gpt-image-2 and GPT-Image-1 image generation. Supports prompt files, OpenAI-compatible image generation, reference-image editing in openai mode, optional zimage native mode, output downloads, dry-run payload inspection, and saved JSON responses.

Aihubmix Image

Use this skill to generate images through Aihubmix.

Core Rules

• Never store real API keys in shared files.
• Prefer AIHUBMIX_API_KEY or a private --env-file .env over --api-key.
• Use the bundled script instead of rewriting API callers.
• Default to --mode openai for gpt-image-2 or GPT-Image-1 compatible generation.
• Use --mode zimage only when the user asks for Aihubmix native /v1/models/{model_path}/predictions workflow.
• Save generated images into the current project unless the user asks for preview-only output.
• If reference images are used, state each image's role and treat them as edit/reference inputs in openai mode.
• If network/DNS fails in a sandboxed environment, rerun the same command with the required network permission.

Files

• Script: scripts/aihubmix_image.py
• User README: README.md
• Usage reference: references/usage.md
• Prompting reference: references/prompting.md
• Config template: assets/.env.example
• Example prompt: assets/example_prompt.txt

Read references/usage.md when you need command examples or option details. Read references/prompting.md when preparing complex prompts or when a prompt is refused.

Standard Workflow

1. Decide whether the request uses direct prompt text or a prompt file.
2. Decide mode:
   • openai: default image generation via /v1/images/generations.
   • zimage: native Aihubmix prediction workflow via /v1/models/{model_path}/predictions.
3. Decide whether reference images are needed and label their role.
   • In openai mode, reference images trigger /v1/images/edits.
   • In zimage mode, currently avoid --reference-image.
4. Put the API key in AIHUBMIX_API_KEY, a private .env file, or use the user-provided key only for the command.
5. Run:

python3 <skill_dir>/scripts/aihubmix_image.py \
  --env-file .env \
  --prompt-file <prompt.txt> \
  --name <output_name> \
  --out-dir <output_dir> \
  --size 1536x864 \
  --quality auto \
  --save-response

6. Add one or more --reference-image <path-or-url> options if needed.
7. Inspect the generated image before judging success.
8. If the image fails the user's constraints, revise the prompt and regenerate rather than manually hiding major generated-image errors.

Troubleshooting

• Missing API key: set AIHUBMIX_API_KEY or pass --api-key.
• HTTP 401: the token is invalid.
• Network/DNS failure: rerun with network permission.
• Policy refusal: rewrite into a neutral, non-harmful, non-identifying, or clearly educational/demo form; see references/prompting.md.
• Reference image ignored: try fewer references and make each role explicit in the prompt.