---

name: python-api description: "Routes programmatic use of Stitcher, AffineStitcher, image lists, feature masks, component settings, and API-default inspection for the stitching package." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Python API

Use this sub-skill when the user wants to call the package from Python instead of the shell.

Typical triggers

• "How do I use Stitcher?"
• "I want to stitch loaded NumPy arrays instead of filenames."
• "How do I make an affine panorama?"
• "What are the default settings?"
• "How do I pass masks or tune matching in code?"

What this sub-skill owns

• from stitching import Stitcher, AffineStitcher
• Stitcher.DEFAULT_SETTINGS and AffineStitcher.AFFINE_DEFAULTS
• Images.of and the filename-vs-NumPy input split
• Feature detection/matching defaults and tuning knobs
• Camera, warper, cropper, seam-finder, blender, and timelapse options
• StitchingError and StitchingWarning patterns that a caller should handle

What this sub-skill does not own

• Shell command building; use cli.
• Verbose file interpretation and matches-graph analysis; use diagnostics.
• Install/import setup; use the root installation notes.

Start with these bundled files

• API reference for verified signatures and choices.
• Workflow recipes for copyable Python snippets.
• Troubleshooting for common mistakes and recovery steps.
• Inspect defaults helper to print a JSON summary of installed defaults and signatures.

Fast routing hints

• If the prompt names Stitcher, AffineStitcher, or Images.of, read the API reference first.
• If the prompt asks for the closest Python equivalent to a CLI command, use the workflow recipes.
• If the prompt is about dropped images or mask mismatches, hand off to diagnostics.

Common success criteria

A good answer from this sub-skill should include:

• Exact signatures or settings names when relevant.
• A concrete Python snippet that works with filenames or loaded arrays.
• The expected output file or return value shape.
• A clear error-handling or warning-handling path for invalid inputs.