tracefinity/tracefinity

Generate custom gridfinity bins with AI, from photos of your tools

Tracefinity – AI‑assisted design of Gridfinity storage bins

What it does – You snap a photo of your tools laid out on a sheet of paper, upload it to Tracefinity, and an AI model automatically extracts the silhouette of each tool. Those silhouettes become editable vector shapes that you can arrange into custom Gridfinity‑compatible bins and export as STL/3MF for 3‑D printing.

Why it matters – Designing tool organizers by hand is tedious. Tracefinity removes the manual tracing step with on‑device or cloud‑based vision models, letting makers quickly generate printable storage that fits exactly around their real tools.

Key features (as described in the README)

  • AI‑powered tracing – Local saliency models (IS‑Net, BiRefNet‑Lite, InSPyReNet) or Google Gemini generate black‑and‑white masks from photos.
  • Paper‑corner calibration – Detects the sheet edges (U2‑Net) to scale the image correctly, even when tools overhang.
  • Tool library – Save traced tools for reuse across multiple projects.
  • Bin builder – Drag‑and‑drop tools onto a Gridfinity layout, with automatic snap‑to‑grid, cut‑out clearance, chamfers, and optional finger‑hole cutouts.
  • Export options – STL, 3MF (multi‑colour), SVG, and a live three.js preview.
  • Automatic naming – Optional Ollama or OpenRouter vision model can suggest tool names.
  • Authentication & multi‑user – Native login, 2FA, proxy or open modes, all configurable via environment variables.
  • Single‑container Docker – Frontend and backend run together; also available via Docker‑Compose or Helm charts for Kubernetes.
  • Resource‑aware – Supports ARM devices (Raspberry Pi) and optional CUDA acceleration for faster inference.

Quick start (Docker)

# Run with the built‑in CPU model (no API key)
docker run -p 3000:3000 -v ./data:/app/storage ghcr.io/tracefinity/tracefinity

# Or use Google Gemini (replace with your key)
# docker run -e GOOGLE_API_KEY=your-key …

Open http://localhost:3000 and create an admin account on first launch.

How tracing works

  1. Place tools on A4/Letter/A3/Tabloid paper (tools may spill over the edges).
  2. Photograph from directly above.
  3. Upload; adjust the four detected corners for scale.
  4. The selected tracer (local model or Gemini) returns a mask.
  5. Convert the mask to polygons via OpenCV contour extraction.
  6. Edit, label, and save the tools.
  7. Assemble tools into a bin layout and export for 3‑D printing.

Model options (environment variable TRACERS)

Model Speed (CPU) RAM Quality
IS‑Net (default) ~0.8 s 2 GB Good – fastest, lowest memory
BiRefNet‑Lite ~3.6 s 8 GB Best – handles reflections
InSPyReNet ~2.8 s 6 GB Good – Apple Silicon support

Set GOOGLE_API_KEY to switch to Gemini (gemini-3.1-flash-image-preview by default). Remote providers (fal.ai, Replicate) are also supported via FAL_KEY or REPLICATE_API_TOKEN.

Installation alternatives

  • Docker‑Compose – define a service with volume ./data:/app/storage and optional GOOGLE_API_KEY.
  • Helm chart – helm install tracefinity oci://ghcr.io/tracefinity/charts/tracefinity with persistence and env values.
  • From source – Python 3.11+ backend, Node 20+ frontend, pnpm for dependencies, then make dev to run backend on :8000 and frontend on :4001.

License

MIT – free to use, modify, and redistribute.

Bottom line – Tracefinity is a practical, AI‑driven tool for makers who want custom‑fit Gridfinity bins without hand‑drawing each part. It bundles the necessary vision models, a web UI, and export pipelines into a single Docker image, making it easy to self‑host on a laptop, Raspberry Pi, or Kubernetes cluster.

Related

  • Project
    Comfy-Org/ComfyUI

    ComfyUI is a modular, node-graph based engine that lets visual creators build and run AI workflows for images, video, audio, 3D, and text, with broad support for state-of-the-art open-source models and efficient local execution on consumer hardware.

  • Project
    huangserva/3DCellForge

    An AI-powered 3D model studio that converts images into interactive 3D models using multiple AI providers and provides a professional inspection and presentation workspace.

  • Project
    mo-browser-apps/icons

    MōIcons is a macOS‑only desktop app that uses the OpenAI image‑generation API to create app icons. You type a short description (and can add a reference image), receive three AI‑generated variants, refine the chosen one, and export a fully‑packed `.icns` bundle with all required sizes.

  • Project
    amagine-ai/Amagine3D

    Amagine3D is an open‑source web app that uses an LLM‑driven “3‑D‑native Agent” to turn product descriptions, images and dimensions into fully editable CAD models (build123d Python source) and manufacturing files (STEP, STL, 3MF). It validates wall thickness, clearances, and motion paths, supports multi‑part enclosures with hinges, and lets users tweak parameters in the UI. The stack combines React/Vite, an Express server, OpenAI‑compatible Codex, and Python CAD kernels (build123d + Open CASCADE).

  • Project
    no8d/ComfyUI-NO8D-controls

    ComfyUI‑NO8D‑controls is a MIT‑licensed add‑on for the ComfyUI image‑generation workflow. It provides a suite of custom nodes—prompt libraries, LoRA stacking, an LLM‑backed prompt generator, a unified generate/inpaint/outpaint node, image compositing, grid and title utilities, and dataset‑saving tools—plus a UI for configuring an OpenAI‑compatible or Ollama LLM. Install by cloning into `ComfyUI/custom_nodes` and pip‑installing `requirements.txt`.