imagegen

Generate images through the Azure OpenAI-compatible image generation API and save the returned base64 image data to local files.

Defaults

• Endpoint/base URL: AZURE_OPENAI_BASE_URL, or https://maritimeai-resource.openai.azure.com/openai/v1/ if unset.
• API key: AZURE_OPENAI_API_KEY is required.
• Model/deployment: AZURE_OPENAI_IMAGE_MODEL, or gpt-image-2 if unset.
• API version header: AZURE_OPENAI_API_VERSION, or preview if unset.
• Output directory: ./generated-images/ if the user does not specify a file path.

Never write API keys into files, command examples, logs, or final responses.

Standard Workflow

1. Collect the prompt. If the prompt is missing, ask for it.
2. Choose output settings. Default to 1024x1024, high, png, and n=1 unless the user specifies otherwise.
3. Run the Node CLI from the skills workspace or with an absolute script path:

node imagegen/scripts/generate-image.mjs \
  --prompt "A watercolor painting of a maritime research vessel at sunrise" \
  --output ./generated-images/ship.png

4. Report the saved file path(s), model, size, quality, and format.
5. If Azure returns an error, preserve the status code and message, but redact secrets.

Command Reference

node imagegen/scripts/generate-image.mjs \
  --prompt "A cinematic harbor at dawn" \
  --size 1536x1024 \
  --quality high \
  --format png \
  --n 1 \
  --output ./generated-images/harbor.png

Optional environment overrides:

export AZURE_OPENAI_BASE_URL="https://maritimeai-resource.openai.azure.com/openai/v1/"
export AZURE_OPENAI_IMAGE_MODEL="gpt-image-2"
export AZURE_OPENAI_API_VERSION="preview"

Useful CLI flags:

• --prompt <text>: text prompt for image generation.
• --prompt-file <path>: read the prompt from a UTF-8 file.
• --output <path>: output file path for one image, or filename prefix when n > 1.
• --output-dir <path>: directory for auto-named outputs.
• --base-url <url>: override AZURE_OPENAI_BASE_URL.
• --model <name>: override AZURE_OPENAI_IMAGE_MODEL.
• --size <WxH>: image size, such as 1024x1024.
• --quality <low|medium|high>: generation quality.
• --format <png|jpeg>: output format.
• --n <1-10>: number of images.
• --user <id>: optional end-user identifier for Azure/OpenAI tracking.
• --timeout-ms <ms>: request timeout, default 180000.
• --dry-run: validate inputs and print the redacted request without calling Azure.

GPT-image-2 Size Rules

For gpt-image-2, validate custom sizes before calling the API:

• Width and height must both be multiples of 16.
• Long edge must be no more than 3840 px.
• Aspect ratio must be no more than 3:1.
• Total pixels must be between 655,360 and 8,294,400.

Use 1024x1024 when unsure.

Reference

Read references/azure-openai-image-api.md when troubleshooting API shape, headers, endpoint variants, or model parameter constraints.