Project files

Matemium separates creative memory, authoring source, input media, and generated output. The desktop interface presents these as focused views even when the underlying workspace contains several files.

project/
├── project.json
├── scenes.py
├── helpers.py
├── brief/
│   ├── description.md
│   ├── passport.json
│   ├── roadmap.json
│   ├── tape.md
│   ├── narration.md
│   └── timestamps.json
├── assets/
│   ├── images/
│   ├── video/
│   └── audio/
└── output/

Not every production path uses every artifact.

scenes.py

The primary visual authoring file. It defines one or more CanvasScene classes using CanvasBuilder.

Both you and the agent edit this visible Python source. The internal SheetDSL representation is compiler data, not the normal product authoring format.

helpers.py

Optional project-specific composition functions. Put repeated subject recipes here instead of adding topic-specific methods to the generic engine.

Examples:

• Build a comparison row for two graphs
• Add a labeled circuit component
• Create a recurring timeline event
• Produce coordinates for a subject-specific diagram

Brief artifacts

description.md

The stable explanation goal, audience, and constraints.

passport.json

Rare-changing production identity and selected production path.

roadmap.json

Current phase, progress, blockers, branch, and phase notes.

tape.md

Approved or provisional content intended to appear on the visual reasoning tape.

Narration and timing files

Used by voice-synthesis and custom-voice paths. Exact artifacts depend on the selected workflow.

Assets

Imported images, video, and audio are grouped by media type. Prefer importing through the desktop interface so the project retains stable paths.

Do not reference temporary downloads or files that may disappear from outside the workspace.

Output

Rendered videos and exports belong in output history. They are generated artifacts and should not be manually edited as if they were authoring source.

Section markers

Large scenes.py files can use visual section fences:

# ---DIV: Introduction---
def part_introduction(tape):
    ...

# ---DIV: Comparison---
def part_comparison(tape):
    ...

The desktop editor can present these as navigable sections without fragmenting the project into many source files.

Backups and version control

Project workspaces are ordinary local data. Back up important projects and use version control when appropriate. Generated media can be large, so decide separately whether output belongs in your repository.