tools.aaris.tech

A self-hosted collection of web-based utilities — clean, fast, no sign-up required.

Built with vanilla HTML/CSS/JS on the frontend and a Flask API on the backend,
served through Nginx and Docker Compose.

Live Site · Changelog

---

Tools

Tool	Description	Tag
Text Fixer	Fix spelling and grammar with AI — powered by Llama 3.1	ai
EPUB to PDF	Accessible EPUB to tagged PDF conversion with TTS-friendly reading order	converter
Web Novel to PDF	Export a supported web novel URL to a tagged PDF for offline TTS	converter
File Share	Upload a file, get a shareable link — auto-deleted on expiry	share
JSON Formatter	Paste messy JSON and get it pretty-printed instantly	dev
Unit Converter	Convert time, data, bandwidth, frequency, power, temperature and more	utility
Date Calculator	Add/subtract time from a date, or find the difference between two	utility
Download Time Calculator	Estimate download duration based on speed and file size	utility
Ticket Ranker	Prioritize tickets with arrow-key shortcuts and an optional fast mode for larger lists	productivity

Architecture

┌──────────────────────────────────────────────────────┐
│  Browser                                             │
└──────────────┬───────────────────────────────────────┘
               │
               ▼
┌──────────────────────────────────────────────────────┐
│  Nginx (tools container)              port 8080 → 80 │
│  ├── /              → static landing page            │
│  ├── /tools/*       → static tool HTML/CSS/JS        │
│  ├── /manifest.json → auto-generated tool list       │
│  └── /api/*         → proxy to converter:5000        │
└──────────────┬───────────────────────────────────────┘
               │
               ▼
┌──────────────────────────────────────────────────────┐
│  Flask / Gunicorn (converter container)   port 5000  │
│  ├── POST /api/fix-text          → Ollama (ai-01)   │
│  ├── POST /api/convert/epub-to-pdf → Calibre + WeasyPrint │
│  ├── POST /api/convert/web-novel-to-pdf → Royal Road + WeasyPrint │
│  ├── POST /api/share/upload      → SQLite + disk     │
│  ├── GET  /api/share/<key>/info                      │
│  ├── POST /api/share/<key>/download                  │
│  └── GET  /api/health                                │
└──────────────┬───────────────────────────────────────┘
               │
               ▼
┌──────────────────────────────────────────────────────┐
│  Ollama — ai-01 (10.10.8.10:11434)                   │
│  └── llama3.1 (8B, Q4_K_M) on Tesla M10 GPU         │
└──────────────────────────────────────────────────────┘

Quick Start

git clone git@github.com:jonarihen/tools-platform.git
cd tools-platform
cp .env.example .env
# Set a strong ADMIN_PASSWORD in .env before starting
docker compose up --build -d

The platform is available at http://localhost:8080.

Environment Variables

Variable	Default	Description
ADMIN_PASSWORD	Required	Admin password for /tools/ocistrator/ and /api/admin/*
OLLAMA_URL	http://10.10.8.10:11434	Ollama API endpoint for the Text Fixer tool
OLLAMA_MODEL	llama3.1	Model to use for text correction

Set these in .env or your shell environment before running docker compose up.

Project Structure

tools-platform/
├── docker-compose.yml
├── Dockerfile                 # Nginx + entrypoint
├── entrypoint.sh              # Auto-discovers tools → manifest.json
├── nginx.conf
├── aaris-design-language.md   # Design language reference (source of style.css)
├── public/
│   ├── index.html             # Landing page (reads manifest.json)
│   ├── style.css              # Shared AARIS design system
│   └── favicon.svg
├── converter-api/
│   ├── Dockerfile             # Python 3.12 + Calibre + WeasyPrint + Gunicorn
│   ├── app.py                 # Flask API (conversion, file share, text fixer)
│   └── requirements.txt
└── tools/                     # Each tool is a self-contained directory
    ├── text-fixer/
    ├── epub-to-pdf/
    ├── web-novel-to-pdf/
    ├── file-share/
    ├── json-formatter/
    ├── unit-converter/
    ├── date-calculator/
    ├── download-time-calculator/
    ├── ticket-ranker/
    └── add-tool-guide/

Adding a New Tool

Each tool lives in its own directory under tools/:

tools/my-tool/
├── index.html    ← self-contained frontend (HTML + CSS + JS)
└── meta.json     ← metadata for the platform

meta.json:

{
  "name": "My Tool",
  "description": "What it does in one sentence",
  "icon": "🔧",
  "tag": "utility",
  "order": 10
}

Field	Required	Description
name	Yes	Display name on the homepage
description	Yes	Short description shown on the card
icon	No	Emoji icon (defaults to 🔧)
tag	No	Category label shown on the card
order	No	Sort order on homepage (lower = first)

The entrypoint.sh script auto-discovers all tools and generates manifest.json on container startup — no other files need to change. Just add the directory, rebuild, and it appears on the landing page.

The tools/ directory is volume-mounted, so you can edit tool files without rebuilding the image. Only the manifest regeneration requires a restart.

A full guide is available at /tools/add-tool-guide/ on the live site.

Security

The platform is hardened for public-facing deployment:

• Passwords — file share passwords are hashed with scrypt (salted, constant-time comparison)
• Rate limiting — per-IP limits on all write endpoints (uploads, text fixer, slug checks)
• Prompt injection — text fixer uses a hardened system prompt + output length cap to limit abuse
• CSP — Content-Security-Policy restricts scripts, styles, fonts, and connections to known origins
• Headers — X-Content-Type-Options, X-Frame-Options, Referrer-Policy, X-XSS-Protection on all responses
• No auth on Ollama — relies on network-level access control (private subnet only)

Stack

Layer	Technology
Frontend	Vanilla HTML, CSS, JS · AARIS design system (public/style.css) · Archivo + IBM Plex Mono fonts
Backend	Python 3.12 · Flask · Gunicorn (1 worker, 8 threads)
AI	Ollama · Llama 3.1 8B (Q4_K_M) · Tesla M10 GPU
Conversion	Calibre (EPUB ingest) + WeasyPrint (tagged PDF output) + Royal Road HTML export
Database	SQLite (WAL mode) for file share metadata
Proxy	Nginx (Alpine)
Infra	Docker Compose · named volumes for persistence