Splitter runs return ordered segments with metadata and optional previews.

{
  "segments": [
    {
      "segment_id": "seg-001",
      "label": "cover",
      "page_range": [1, 1],
      "confidence": 0.91,
      "content_uri": "https://api.algorythmos.fr/files/seg-001"
    },
    {
      "segment_id": "seg-002",
      "label": "invoice",
      "page_range": [2, 3],
      "confidence": 0.88
    }
  ]
}

Consuming segments

• Fetch content_uri to download cropped PDFs or text extractions.
• Pass segment_ids into downstream processors via workflow variables.
• Use confidence to gate human review or fallback logic.

Each segment inherits the original file metadata, so you can trace downstream outputs back to the source pages easily.