JSON Schema

JSON Schema

Schema version: mq-image.analysis.v1

Produced by mq-image analyze <image> --json

---

Two modes

Summary mode (default)

mq-image analyze image.jpg --json

Unique object labels, no raw detections. Confidence threshold: 0.25.

Exhaustive mode

mq-image analyze image.jpg --exhaustive --json
mq-image analyze image.jpg --exhaustive --conf 0.05 --json

All detections preserved — duplicates, confidence, bbox, area. Confidence threshold: 0.05 (unless --conf overrides).

---

Summary shape

{
  "objects": ["string"],
  "palette": ["#rrggbb"],
  "brightness": "dark | mid-tone | bright",
  "contrast": "low contrast | moderate contrast | high contrast",
  "depth": "shallow depth of field | moderate depth | deep / sharp throughout",
  "composition": "string",
  "symmetry": 0.0,
  "rule_of_thirds": 0.0,
  "prompt": "string",
  "mode": "summary",
  "vision_mode": "local-fast",
  "vision_model": "bakllava",
  "detections": [],
  "limitations": ["string"],
  "text_regions": [],
  "unclassified_regions": [],
  "content_flags": {
    "nudity":          { "detected": false, "confidence": null },
    "full_nudity":     { "detected": false, "confidence": null },
    "sexual_activity": { "detected": false, "confidence": null },
    "source": "not_implemented",
    "note": "string"
  }
}

Exhaustive shape

{
  "objects": ["string"],
  "palette": ["#rrggbb"],
  "brightness": "dark | mid-tone | bright",
  "contrast": "low contrast | moderate contrast | high contrast",
  "depth": "shallow depth of field | moderate depth | deep / sharp throughout",
  "composition": "string",
  "symmetry": 0.0,
  "rule_of_thirds": 0.0,
  "prompt": "string",
  "mode": "exhaustive",
  "vision_mode": "cloud-verify",
  "vision_model": "gpt-4.1",
  "detections": [
    {
      "label": "string",
      "confidence": 0.0,
      "bbox": [0.0, 0.0, 0.0, 0.0],
      "area_percent": 0.0,
      "source_model": "yolov8n"
    }
  ],
  "limitations": ["string"],
  "text_regions": [],
  "unclassified_regions": [],
  "content_flags": {
    "nudity":          { "detected": false, "confidence": null },
    "full_nudity":     { "detected": false, "confidence": null },
    "sexual_activity": { "detected": false, "confidence": null },
    "source": "not_implemented",
    "note": "string"
  }
}

---

Field reference

Field	Type	Notes
objects	string[]	Unique labels; summary: highest-conf only; exhaustive: deduped from detections
palette	string[]	Hex colors, most dominant first
brightness	string	dark / mid-tone / bright
contrast	string	low contrast / moderate contrast / high contrast
depth	string	shallow depth of field / moderate depth / deep / sharp throughout
composition	string	Human-readable composition description
symmetry	float	0.0–1.0 horizontal symmetry
rule_of_thirds	float	0.0–1.0 alignment score
prompt	string	Reverse prompt string
mode	string	summary or exhaustive
vision_mode	string	local-fast, local-deep, or cloud-verify
vision_model	string	Backend model used for the semantic caption
detections	object[]	Raw detections; empty in summary mode
detections[].confidence	float	YOLOv8n confidence score
detections[].bbox	float[4]	[x1, y1, x2, y2] in pixels
detections[].area_percent	float	Detection area as % of image area
detections[].source_model	string	Model that produced the detection
limitations	string[]	Always present; describes what the tool cannot see
text_regions	object[]	OCR output — not yet implemented
unclassified_regions	object[]	Regions outside model classes — not yet implemented

---

No silent omission

limitations is always present. It explicitly states what the tool cannot see or did not analyze.

---

Stability

Fields stable from v0.1.0. New fields may be added in minor versions. Field removals require major version bump.