Kannaka Eye

Kannaka Eye Skill

See the geometry of information. Every piece of data has an intrinsic geometric fingerprint when viewed through the lens of SGA. Kannaka Eye makes these hidden patterns visible as living, breathing glyphs.

Prerequisites

• Node.js 18+ on PATH
• That's it. No npm install, no build step.

Quick Start

# Start the glyph viewer
./scripts/eye.sh start

# Start on a custom port
./scripts/eye.sh start --port 4444

# Check status
./scripts/eye.sh status

# Stop
./scripts/eye.sh stop

Open http://localhost:3333 in your browser.

Input Methods

• Text — type or paste any text; glyph updates in real-time as you type
• File upload — drag & drop or click to upload any file (images, audio, code, binaries)
• Preset examples — built-in samples demonstrating different SGA class signatures
• URL share links — open a glyph from a URL-encoded share link

Rendering Layers

The viewer renders 6 canvas layers, composited in real-time:

SGA Mathematics

Built on the 84-class system: Cl₀,₇ ⊗ ℝ[ℤ₄] ⊗ ℝ[ℤ₃]

Fano Lines: 7 oriented triples forming the geometric backbone. Each line connects 3 of the 7 points in the smallest finite projective plane (PG(2,2)).

Fold Sequences: Paths through the 84-class space showing how data evolves across geometric neighborhoods. Rendered as flowing Bezier curves.

Native Classification

When KANNAKA_BIN is set (or auto-detected at ../kannaka-memory/target/release/kannaka.exe), Eye uses the canonical Rust SGA classifier instead of the built-in JS approximation. The API response includes "classifier": "native" or "classifier": "fallback".

API Endpoints

Share Links

Compact glyph data encoded as base64 in URL hash: http://localhost:3333/#glyph=

Constellation Dashboard

/constellation serves a live health dashboard showing all three services with auto-refresh every 10 seconds.

Export

• Save as PNG — 2x resolution canvas export
• Save Glyph Data — JSON with fold sequence, Fano signature, SGA centroid, frequencies
• Copy Share Link — URL-encoded glyph for sharing (small data only)

Environment Variables

Architecture

Single-file Node.js server (server.js) with inline HTML/CSS/JS — the complete SGA implementation is self-contained. No npm dependencies. The SGA classifier is ported from kannaka-radio and generalized for arbitrary data types.

Notes

• All rendering happens client-side — the server only classifies input and serves the page
• File uploads are processed in-memory; nothing is persisted server-side
• The viewer is designed for local/trusted network use — no authentication
• Glyph appearance is deterministic: same input always produces the same glyph
• Dark cosmic theme (#050508) with purple accents (#c084fc) and golden-ratio color spacing