Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/UnkleFunk/HouseMusicSwarm-/llms.txt

Use this file to discover all available pages before exploring further.

The Image Generation Agent is a specialist for producing, editing, and compositing high-quality images. It selects the right model for each task based on quality requirements and available API keys, runs a mandatory QA pass after every generation, and performs one automatic correction before final delivery. All outputs are saved to mnt/{product_name}/generated_images/ and the file path is always included in the response.

Tools

GenerateImages

Generates one or more images from a text prompt. Supports gemini-2.5-flash-image, gemini-3-pro-image-preview, and gpt-image-1.5. Up to 4 variants per call. Accepts a wide range of aspect ratios (1:1, 16:9, 9:16, 4:3, 3:4, and more). Saved to mnt/{product_name}/generated_images/.

EditImages

Edits an existing image based on a natural language instruction. Accepts a local path, URL, or previously generated image name as input. Supports the same three model options as GenerateImages, with up to 4 output variants. Ideal for targeted changes like style transfers, background swaps, or adding/removing elements.

CombineImages

Combines two or more images into a single composition following a text instruction. Use this whenever you need to place elements from one image into another — for example, applying a company logo onto a product photo. Requires at least two image references. Supports Gemini and OpenAI models.

RemoveBackground

Removes the background from an image using the Pixelcut model via fal.ai, producing a transparent PNG. Accepts a local path, URL, or generated image name as input. Requires FAL_KEY.

Supported providers

The agent supports three model families. Each tool validates that the requested model is available given your configured API keys, and will report a clear error if a required key is missing.
Requires GOOGLE_API_KEYTwo Gemini image models are available:
  • gemini-2.5-flash-image — default for most tasks. Fastest high-quality option, well-suited for iterative workflows and rapid variants.
  • gemini-3-pro-image-preview — precision-first. Use this for text-heavy images, complex product compositions, high-fidelity brand assets, or any prompt with many competing constraints.
Gemini models support the full aspect ratio set: 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9.

Model selection logic

The agent follows these guidelines when choosing a model:
  1. Default: gemini-2.5-flash-image for most generation and editing tasks
  2. Upgrade to gemini-3-pro-image-preview for: text-heavy images, complex compositions, brand assets, precise editing, large prompts with many style directives
  3. Use gpt-image-1.5 only when OpenAI is explicitly requested or for multi-model comparison
  4. If a requested model is unavailable, the agent switches to a compatible one and explains why

Example prompts

  • “Generate a hero image for my startup’s landing page”
  • “Remove the background from this product photo”
  • “Combine these 3 images into a grid composition”
  • “Create two variants: one with Gemini Flash and one with the Pro model”
  • “Edit this photo to use a dark, cinematic style”
  • “Generate a clean product shot on a marble surface with soft studio lighting”
Both GOOGLE_API_KEY and FAL_KEY are optional but unlock different capabilities. GOOGLE_API_KEY enables GenerateImages, EditImages, and CombineImages with Gemini models. FAL_KEY is required exclusively for RemoveBackground. If you only set one key, those specific tools that require the missing key will return an informative error while the others continue to work.

Build docs developers (and LLMs) love