entviz 0.11.0

Visualize high-entropy values (keys, signatures, UUIDs, addresses) as SVG diagrams a human can compare at a glance.

entviz

Entviz is a tool for visualizing high-entropy values (like cryptographic keys, UUIDs, or blockchain addresses) into a grid of colored shapes and text, making it easy for humans to compare them. This is its reference implementation in python. There are also implementations in javascript/react and rust.

{ width="320" }

This repository itself, rendered as an entviz of its own root commit hash.

Install

pip install entviz      # or: uv add entviz

Render any value to an SVG on stdout:

entviz "550e8400-e29b-41d4-a716-446655440000" --ar 1:1 > id.svg

The library has one runtime dependency (lxml) and emits SVG only — no rasterizer required.

Comparing two entvizes

Note: This is interim guidance. A dedicated comparison UI (a React component that aligns two entvizes and highlights differences) is planned; until then, follow the manual procedure below.

entviz is built for comparison, not memorization. To check whether two values are the same:

• Render both the same way. Same point size, same font, same background, shown side by side at the same scale. Differences in scale, zoom, font, or surrounding color can hide or fake a difference.
• Reject on any visible difference. The check is asymmetric: any visible difference in text, color bar, surround pattern, blank positions, ellipse, or quartile marks means the values are different. A match is never proof of identity — it only means no difference was found at the resolution you looked.
• Compare every channel, not just one. A habituated reader who checks only one landmark (e.g. the color bar) is the easiest to fool. Scan the text, the color bar (each band carries a letter w/g/r/b/k), the surround rings, the blank-cell positions, and the overlays.

The fingerprint of marker

For inputs longer than 512 bits, the text channel can't show every character, so the entviz is labeled fingerprint of <type>(<length>): in bold dark red. On these entvizes:

• The first 8 and last 8 text cells are real characters from the start and end of your value (use these to recognize and spot-check it).
• The 4 middle cells (neutral background, framed) are a hash readout in hex — they are not characters from your value. Two different long inputs can therefore share many visible cells and still be different; trust the full picture, and remember that the whole input is bound into every color/shape channel through the fingerprint.

Developer Quickstart

Prerequisites

• uv (manages the Python interpreter, virtualenv, and dependencies). uv will fetch a suitable Python (>=3.10) for you.

Installation

1. Clone the repository:

   git clone https://github.com/dhh1128/entviz.git
   cd entviz
   

2. Create the environment from the lockfile:

   uv sync
   

Running Tests

uv run pytest

To prove the supported Python floor locally (CI runs the full 3.10/3.11/3.12 matrix):

uv run --python 3.10 pytest

Running the CLI

entviz is installed as a console entry point:

uv run entviz "your-entropy-string" --ar 1:1 --fs 12

Regenerating the gallery

uv run python scripts/gallery.py

Regenerating the social preview card

The GitHub "social preview" image (the Open Graph card that unfurls when the repo URL is shared on Slack, X, LinkedIn, etc.) is a reproducible asset. The card embeds an entviz of this repo's own root commit SHA — the tool applied to itself — which is the same entviz shown at the top of the docs.

uv run python scripts/social_card.py

This (re)writes three files into docs/assets/, alongside the other docs images:

• root-commit-entviz.svg — the entviz of the root commit hash (also embedded at the top of docs/index.md);
• social-card.svg — the composed 1280×640 card (vector source of truth);
• social-card.png — the PNG you upload (under 1 MB).

The PNG is rendered with cairosvg (a dev dependency) using the DejaVu font fallback baked into the font stack, so it reproduces without bundling fonts.

Uploading it (one-time, manual): GitHub → repo Settings → General → Social preview → Edit → upload docs/assets/social-card.png.

Reusing the template across repos: the palette, layout, and typography are shared family constants; only the KNOBS block at the top of scripts/social_card.py (repo name, owner, tagline, language, and the MARK) changes per repo. See the module docstring for details.

Cutting a release

python3 scripts/release.py --patch -m "what changed"

The script is self-guarding: it switches to main, fast-forwards to origin/main, refuses a dirty tree or unpushed local commits, runs the full test suite, regenerates the gallery, then bumps the version, commits, and pushes a vX.Y.Z tag. It is pure-stdlib, so it works from any directory (it operates on the repo root regardless of cwd) — python3 /path/to/entviz/scripts/release.py … is equivalent. (uv must be on PATH, since the script shells out to uv run for the tests and gallery.)

The pushed tag triggers .github/workflows/release.yml, which re-runs the tests on the tagged commit, publishes the sdist + wheel to PyPI via Trusted Publishing (OIDC — no API token), and creates a GitHub Release with the same artifacts attached.

See scripts/release.py for bump options (--minor, --major, --set X.Y.Z, --no-bump) and the versioning convention.

Project Structure

• src/entviz/: Core library.
  • entropy.py: Entropy parsing and normalization.
  • layout.py: Grid layout calculations.
  • colors.py: Color selection and conversion.
  • shapes.py: Edge shape definitions.
  • pipeline.py: End-to-end render pipeline.
  • app.py: CLI entry point (entviz).
  • __init__.py: SPEC_VERSION (algorithm/spec) and __version__ (library) — the single source of truth for both.
• scripts/: Maintenance tooling (gallery.py, release.py).
• tests/: Unit tests.
• docs/: Specification, gallery, and assets.

Documentation

For the full specification of the Entviz algorithm and design goals, see docs/spec.md.

Methodology & AI Collaboration

This repository follows a specific methodology for high-quality software development, especially when collaborating with AI:

• AI Behavioral Rules: See AGENTS.md for mandatory rules on comprehension, TDD, and intent tracking.
• Intent Layer: We track design decisions and goals in this.i.
• TDD: Strict Test-Driven Development is required for all changes.