LoadArtifact [step] v1.0.0

signature: (artifact: URL | FilePath | Base64Image) -> Document effect: read-only

Errors

• LoadTimeout
• NotFound
• UnsupportedFormat

Body

    1. Detect artifact type from extension or MIME:
       - .html / .htm → parse as HTML Document
       - .png / .jpg / .webp / .gif → treat as screenshot Image
       - .fig / .sketch → treat as design file reference (metadata only)
       - .pdf → extract text + images per page
    2. For URL: fetch with a 30-second timeout. On timeout → throw LoadTimeout.
    3. For Image: pass to vision model for structural analysis; extract:
       - bounding regions of distinct UI components
       - dominant colors (hex values)
       - approximate text content
    4. For HTML: parse DOM; collect:
       - all interactive elements (buttons, links, inputs)
       - computed styles on focusable elements
       - ARIA roles and labels
    5. Return unified Document with fields: { type, raw, components[], colors[], text, metadata }
  

LoadArtifact [step] v1.0.0

signature: (artifact: URL | FilePath | Base64Image) -> Document effect: read-only

Errors

• LoadTimeout
• NotFound
• UnsupportedFormat

Body

    1. Detect artifact type from extension or MIME:
       - .html / .htm → parse as HTML Document
       - .png / .jpg / .webp / .gif → treat as screenshot Image
       - .fig / .sketch → treat as design file reference (metadata only)
       - .pdf → extract text + images per page
    2. For URL: fetch with a 30-second timeout. On timeout → throw LoadTimeout.
    3. For Image: pass to vision model for structural analysis; extract:
       - bounding regions of distinct UI components
       - dominant colors (hex values)
       - approximate text content
    4. For HTML: parse DOM; collect:
       - all interactive elements (buttons, links, inputs)
       - computed styles on focusable elements
       - ARIA roles and labels
    5. Return unified Document with fields: { type, raw, components[], colors[], text, metadata }
  

Errors

• LoadTimeout
• NotFound
• UnsupportedFormat

Body

    1. Detect artifact type from extension or MIME:
       - .html / .htm → parse as HTML Document
       - .png / .jpg / .webp / .gif → treat as screenshot Image
       - .fig / .sketch → treat as design file reference (metadata only)
       - .pdf → extract text + images per page
    2. For URL: fetch with a 30-second timeout. On timeout → throw LoadTimeout.
    3. For Image: pass to vision model for structural analysis; extract:
       - bounding regions of distinct UI components
       - dominant colors (hex values)
       - approximate text content
    4. For HTML: parse DOM; collect:
       - all interactive elements (buttons, links, inputs)
       - computed styles on focusable elements
       - ARIA roles and labels
    5. Return unified Document with fields: { type, raw, components[], colors[], text, metadata }