Coronelli ◉

A literary atlas builder — upload a book and explore its world in 3D.

Coronelli is inspired by the craft of historical cartographers: it reads a novel section by section, finds every named place, maps spatial relationships and visual qualities, and builds a navigable world you can orbit and inspect.

Upload → Catalog → Extract → Synthesize → Atlas Editor → 3D Explorer

• Import .txt, .md, and text-bearing .pdf files
• LLM pipeline identifies places, containment hierarchies, spatial claims, and visual details
• Every extracted candidate is held for human review before it becomes canonical
• Edit the atlas — rename entities, merge duplicates, assign hierarchy, add relationships
• Explore in an interactive 3D scene: zone rings, floating nodes, force-directed layout

Installation

Requires Python 3.12+. Install with uv:

uv tool install coronelli

Then launch:

coronelli

Coronelli opens in your browser at http://localhost:8000. Your atlas data is stored locally:

Set your OpenAI API key before running extraction:

# in your shell profile, or in %LOCALAPPDATA%\Coronelli\.env  (Windows)
#                                ~/Library/Application Support/Coronelli/.env  (macOS)
OPENAI_API_KEY=sk-...

Requirements (for development)

• Python 3.12+, uv
• Node 20+, pnpm 9+
• OpenAI API key

Setup

# Backend
cd apps/api
uv sync --extra dev

# Frontend
pnpm install

Copy .env.example to apps/api/.env and set OPENAI_API_KEY.

Running

# Backend  (http://localhost:8000)
cd apps/api
uv run uvicorn app.main:app --reload

# Frontend  (http://localhost:5173)
pnpm dev

Tests

cd apps/api && uv run pytest -q
pnpm --filter web test --run

Cost and caching

Estimated cost is shown before each pipeline stage. Repeated extraction on unchanged sections is served from cache — no new API call is made.

License

Source-available, personal non-commercial use only. See LICENSE.