Home > Tasks > File > ImageTableTask

ImageTableTask

Generate structured grids of AI-generated images for style comparisons, product variations, and concept exploration.

Category: File Model: Image Generation Output: HTML/Markdown
⚙️ ExecutionConfig.json
{
  "rows": ["Astronaut", "Robot"],
  "columns": ["Impressionist", "Synthwave"],
  "image_prompt_template": "A portrait of an {row}, {column} style",
  "base_style": "cinematic lighting, 8k",
  "output_directory": "portraits_grid",
  "image_format": "png"
}
→
👁️ Session UI Gallery
Impressionist Synthwave
Astronaut
Robot

✔ 4 images generated

✔ HTML table saved to portraits_grid/image_table.html

Workspace Artifacts

Explore the generated images and HTML reports directly from the task workspace.

Execution Configuration

Field Type Description
rows * List<String> Row labels for the image table (descriptive text or image file paths).
columns * List<String> Column labels for the image table (descriptive text or image file paths).
image_prompt_template * String Prompt template using {row} and {column} as placeholders.
base_style String Base style or context to apply to all generated images.
output_directory String Relative path for generated images. Default: "generated_images".
image_format String Format: 'png', 'jpg', or 'jpeg'. Default: "png".

Type Configuration (Static)

Field Default Range
parallel_generation 2 1 - 10
image_width 512 64 - 2048 px
image_height 512 64 - 2048 px

Task Lifecycle

  1. Validation: Checks for required labels and valid image format.
  2. Initialization: Resolves output directory and initializes ImageProcessingAgent.
  3. Generation: Iterates through the grid, replacing placeholders and generating images in parallel.
  4. Export: Saves individual images, generates a standalone image_table.html, and a Markdown summary.

Embedded Execution (UnifiedHarness)

Use the UnifiedHarness to run this task programmatically in headless environments or CI/CD pipelines. 

import com.simiacryptus.cognotik.plan.tools.file.ImageTableTask
import com.simiacryptus.cognotik.plan.tools.file.ImageTableTask.*

fun generateComparisonGrid(harness: UnifiedHarness, projectDir: File) {
    // 1. Define Runtime Input
    val executionConfig = ImageTableTaskExecutionConfigData(
        rows = listOf("Astronaut", "Robot"),
        columns = listOf("Impressionist", "Synthwave"),
        image_prompt_template = "A portrait of an {row}, {column} style",
        base_style = "cinematic lighting",
        output_directory = "style_comparison"
    )

    // 2. Run Task
    harness.runTask(
        taskType = ImageTableTask.ImageTable,
        typeConfig = ImageTableTaskTypeConfig(parallel_generation = 4),
        executionConfig = executionConfig,
        workspace = projectDir,
        autoFix = true
    )
}

Prompt Segment

The following instructions are injected into the LLM context:

ImageTable - Generate a table/grid of AI-generated images
  ** Specify row labels in the 'rows' array
  ** Specify column labels in the 'columns' array
  ** Provide an image_prompt_template using {row} and {column} placeholders
  ** Optionally specify a base_style for consistent styling
  ** Generates an HTML table displaying all images with labels