Luminus is a photography workspace that makes your entire library discoverable through semantic search, so you can move from memory to selection in seconds.

It features:

1. Semantic search that surfaces images by content and meaning

Roadmap:

1. Workflow with photography projects
2. Support RAW files

Development

Prerequisites

• Rust (stable toolchain)
• Bun (package manager)
• Tauri CLI v2
• uv (optional)

Setup

1. Clone the repository

   git clone <repo-url>
   cd luminus

2. Download CLIP model from hugging face and convert to ONNX

uvx --from "optimum[onnxruntime]" optimum-cli export onnx --model laion/CLIP-ViT-B-32-laion2B-s34B-b79K .models/laion-CLIP-ViT-B-32-laion2B-s34B-b79K

3. Split the CLIP model into vision and text models

uvx --from "onnx" python3 scripts/split-clip.py

2. Install frontend dependencies

   cd desktop
   bun install

3. Run in development mode

   cargo tauri dev

Useful commands

# Build for production
bun tauri build

# Lint frontend
npx @biomejs/biome check --write src/

# Lint + format Rust (from desktop/src-tauri/)
cargo fmt
cargo clippy -- -D warnings

Disclaimer

The non learning parts of this project (UI, DB CRUD, etc.) were developed with heavy assistance from AI coding tools, primarily Claude Code and OpenAI Codex. The ML/AI components (model selection, embedding pipeline, and inference logic) were designed and implemented by hand with guidance by online resources and tools.