iLook Face Analysis MCP

เครื่องมือวิเคราะห์รูปหน้า ความสมมาตร คะแนนฟี และคำติชมรูปลักษณ์ที่ได้รับอนุญาตสำหรับผู้ใหญ่ ผ่าน Streamable HTTP MCP

ดูบน ilook.fit

เอกสาร

iLook API v1

Send one authorized adult image to POST /api/v1/score as multipart/form-data.

For a single free-tool report, use POST /api/v1/tools/{tool} with a concrete supported tool key such as jawline_rating or golden_ratio_face. The MCP server also exposes named focused tools alongside score_face.

Required fields

  • image: JPEG, PNG, or WebP, maximum 5 MB.
  • adult_and_authorized=true: adult and lawful-use attestation.
  • Issued agent keys authenticate with Authorization: Bearer <api-key>.

Response

The API returns an analysis id, rubric version, model, overall score out of 100, five visible subscores, face shape context, a 0–100 Golden Ratio phi score, one-line explanations, summary, confidence, and a safety disclaimer.

iLook API vs general-purpose computer-vision APIs

iLook is a purpose-built appearance-feedback API. General-purpose computer-vision APIs can support broad image tasks; iLook instead returns a small, versioned scorecard with explicit limits and does not offer identity recognition, biometric matching, medical outputs, or sensitive-trait inference. Use iLook for consented profile-photo feedback, styling workflows, image-quality coaching, and agent demonstrations.

Implementation examples

Python REST workflow

Use a multipart request to https://www.ilook.fit/api/v1/score with Authorization: Bearer <api-key>, adult_and_authorized=true, and a JPEG, PNG, or WebP image no larger than 5 MB. Keep the source image in memory for the request only. If you retain a permitted downstream result, store analysis_id, rubric_version, and the resulting fields, never the image bytes in logs, traces, or queue payloads.

MCP workflow

Initialize POST /mcp, list tools, then call score_face with image_base64, mime_type, adult_and_authorized: true, and an optional focused measurement_key. Use named focused tools when only one output is needed. Do not put base64 image bytes in prompts, logs, queue payloads, or traces.

measurements returns the free-tool report blocks. Each block uses score, label, analysis, and stable details rows.

Stable measurement rows

  • face_shape_detector: Primary Shape, Primary Match, Runner-Up Shape, Runner-Up Match, Length-to-Width Cue, Widest Area, Jawline Curve, Styling Direction.
  • jawline_rating: Left Estimate, Average, Right Estimate, Front-Photo Jaw Angle, Definition, Taper, Symmetry, Visible Jaw Width, Visible Lower Face.
  • canthal_tilt: Left Eye Tilt, Right Eye Tilt, Average Tilt, Left/Right Difference, Category.
  • eye_shape: Opening Shape, Opening Ratio, Eye Spacing, Eye Color Estimate, Corner Orientation, Mouth Symmetry, Eye Framing, Styling Ideas, Brow Notes.
  • face_age_cues: Photo Presentation, Presentation Direction, Top Factors.
  • facial_symmetry: Overall Symmetry Score, Eye Symmetry, Eyebrow Symmetry, Nose Symmetry, Mouth Symmetry, Jaw Symmetry, Straight-On Photo Check, Feature-Level Proof.
  • golden_ratio_face: Golden Ratio Score, Face Length-to-Width, Facial Thirds, Facial Fifths.
  • attractiveness_test: Overall Score, Symmetry, Facial Thirds Balance, Facial Fifths Balance, Golden Ratio, Jawline Measurement, Face Shape Measurement.
  • psl_score: PSL-Style Range, Geometry Score, Symmetry, Visible Thirds, Facial Fifths, Golden-Ratio Proximity, Conversion Basis.
  • facial_ratios: FWHR-Style Upper-Face Ratio, Visible Face Length-to-Width Ratio, Jaw-to-Cheekbone Width Ratio, Pupil-Distance-to-Face-Width Ratio, Eye Spacing Ratio, Average Canthal Tilt, Midface Height Share, Lower-Face Height Share, Mouth-to-Face Width Ratio, Nose-to-Mouth Width Ratio, Nose-Length-to-Face-Height Ratio, Lip-Height-to-Mouth-Width Ratio.
  • color_analysis: Season Verdict, Runner-Up Season, Palette Temperature, Value, Chroma, Contrast, Sample Colors, Lighting Caveat.
  • skin_analysis: Texture Evenness, Tone Evenness, Apparent Hydration, Shine, Visible Photoaging Signs, What's Working, Focus Areas, Lighting Caveat.
  • hairstyle_finder: Face Shape, Best Overall Direction, Recommended Hairstyles, Beard/Framing Notes, Styles to Use Carefully.

Agent discovery