koala-diagrams 0.1.0

Koala is a DSL and SVG compiler for structured concept diagrams.

Koala

🇺🇸 English | 🇪🇸 Español

Koala is a DSL for generating diagrams from structured text.

The project is designed around a simple idea: the same source file should be able to drive multiple layouts and visual styles without rewriting the content itself.

Quick Start

Main entry point:

• cli.py
• installed command: koala

Install locally in a virtual environment:

python3 -m venv .venv
./.venv/bin/pip install -r requirements.txt
./.venv/bin/pip install -e . --no-build-isolation

If you prefer a user-facing install instead of a repo-local venv:

pipx install .

Basic commands:

koala themes
koala layouts
koala typographies
koala compile docs/examples/tree.txt --layout tree
koala compile docs/examples/radial.txt --layout radial --theme jungle --size square
koala inspect docs/examples/tree.txt
koala validate docs/examples/radial.txt --strict
koala config-path

Available subcommands:

• compile: render a source file to SVG
• inspect: show metadata, warnings, and resolved settings
• validate: validate parsing and resolved settings; with --strict it fails on warnings
• themes: list available themes
• layouts: list available layouts
• typographies: list available typography presets
• config-path: show the expected user config path

User config:

• default path: ~/.config/koala/config.toml
• fallback path: ~/.koala.toml

Example:

[tool.koala]
default_layout = "tree"
default_theme = "academic"
default_typography = "default"
default_size = "a4_landscape"
default_text_align = "left"
default_show_node_numbers = true
default_output_mode = "next_to_input"

Supported config keys:

• default_layout
• default_theme
• default_typography
• default_size
• default_text_align
• default_show_node_numbers
• default_output_mode: next_to_input, desktop, cwd
• default_output_dir

Output behavior:

• by default, output goes next to the source file and is named <input_stem>.<layout>.svg
• --output writes to an explicit SVG path
• --output-dir writes to a specific folder
• --desktop writes to ~/Desktop when present, otherwise falls back to the input folder

Current capabilities

• Parse hierarchical concept trees from .txt and .docx
• Render the same document as tree, synoptic, synoptic_boxes, or radial
• Apply themes and typography presets
• Select page size from the CLI
• Read optional render metadata directly from the source file with @...
• Fit the final scene into SVG output

Syntax overview

Nodes use hierarchical numbering:

• 1, 2, 3 are roots
• 1.1, 1.2 are children of 1
• 1.1.1 is a child of 1.1
• 0 is an optional super-root

Example:

1 DSL Platform
Defines the main concept.

organizes -> 1.1 Core
Contains parser, models, and validations.

renders -> 1.2 Render
Generates SVG output.

Optional semantic kind:

main:: 1 DSL Platform
hl:: 1.2 Users Table
hl:: contains -> 1.2.1 Foreign Key
note:: 1.3 Note
focus:: 1.4 Main Insight

Built-in kinds currently available:

• universal: note, warn, soft
• theme-owned: main, hl, focus

main is intended mainly for the diagram root. When the root node uses main::, built-in themes give it a more prominent color treatment and a thicker border.

Built-in themes currently available:

• default
• terracotta
• jungle
• frutal
• academic

Optional metadata:

@layout tree
@theme terracotta
@typography default
@text-align justify
@size square
@show-node-numbers false
@output-dir output/metadata_demo

Metadata is optional and follows this precedence:

1. explicit CLI flag
2. metadata in the source file
3. built-in defaults

Node text alignment is now left by default. If you want justified text, use @text-align justify.

For the full syntax, see docs/syntax.md.

CLI Reference

Current page presets:

• a4: 210 x 297 mm
• a4_landscape: 297 x 210 mm
• square: 210 x 210 mm

Output:

• by default, next to the source file, named <input_stem>.<layout>.svg
• it can also go to Desktop, another folder, or an explicit file path via --output

Layouts

Current supported layouts:

• tree: top-down hierarchy with adaptive parent widths and page-aware fitting
• synoptic: synoptic chart with brackets and no relation labels
• synoptic_boxes: synoptic chart with boxes and relation labels
• radial: center-out mind-map style layout

For implementation details, see docs/layouts.md.

Architecture

The project is split into three layers:

• core/: parsing, DSL models, input loading
• layout/: shared measurement plus one geometry engine per layout
• render/: theme resolution, render context, viewport fitting, and SVG backend

The separation of concerns is:

• core understands the language
• layout computes geometry and connector paths
• render resolves presentation and draws the scene

For the full architecture, see docs/architecture.md.

Authoring tips for better visuals

These suggestions are practical heuristics for getting cleaner diagrams with the current engines.

General

• Keep titles short and concept-oriented; move explanation into body text
• Use relations only when they add value; too many relation labels create noise
• Prefer balanced sibling groups over very uneven branching when possible
• Split very long paragraphs into tighter prose; box measurement is line-based and compact prose produces better wrapping
• Use metadata for layout/theme/size when a file is meant to be rendered in a specific way

tree

• Best for classic hierarchies and explanatory maps
• Works especially well when upper levels are conceptual and lower levels carry detail
• Good fit for a4_landscape and square
• If a parent has many children, keep child titles compact so the tree can stay readable at larger sizes
• If the map is shallow and very wide, square often helps force a denser composition

synoptic

• Best for grouped classifications and outline-like content
• Prefer short titles and very light body text
• Works best when each level behaves like a category refinement
• a4_landscape is currently the safest page size because the internal geometry is still column-oriented
• Use fewer relation labels in the source; they are intentionally suppressed in this layout

synoptic_boxes

• Best for left-to-right teaching diagrams or process decompositions
• Keep each sibling set conceptually parallel
• Use short-to-medium body text, not large paragraphs
• a4_landscape is currently the safest page size
• square is already supported, but there is no dedicated compaction pass yet, so content should be tighter

radial

• Best for central topics with several balanced branches
• Try to keep first-level branches similar in importance and size
• Avoid one giant branch plus many tiny ones if you want a clean circular composition
• square and a4_landscape usually give the best radial balance
• If the root concept is very short and children are rich, radial can use the page very efficiently

Page-size-specific advice

a4

• Better for taller trees or vertically deep hierarchies
• Good when you want more stacked reading rhythm
• Less ideal for wide synoptic structures unless the content is compact

a4_landscape

• Current default
• Best general-purpose preset
• Safest choice for synoptic and synoptic_boxes
• Also strong for medium-width tree documents

square

• Best when you want a denser, more centered composition
• Strong option for radial
• Useful for tree when the content would otherwise spread too far horizontally
• Requires tighter writing for synoptic and synoptic_boxes until those engines get aspect-ratio-specific optimization

Recommended examples

• docs/examples/tree.txt
• docs/examples/radial.txt

Current status

The project is already useful for:

• rapid concept-map prototyping
• trying multiple layouts from the same content
• testing visual semantics with kind::
• shipping self-describing demo files through metadata

There is still room to improve:

• layout-specific adaptation for synoptic and synoptic_boxes under portrait and square pages
• richer DSL syntax
• stronger validations
• automated visual regression checks