Prepare your manuscript

Prepare source material so Arc can keep evidence close to the claims it extracts.

Use simple files first

For the current preview, use:

Markdown (.md);
plain text (.txt).

Avoid starting with PDFs, images, scans, complex exports, or heavily formatted documents. If you have a DOCX manuscript, export the chapters to Markdown or plain text before testing Arc.

Split by chapter or section

Each source file should represent a natural unit:

one chapter;
one scene sequence;
one short story;
one appendix or lore note.

Do not put a full series into a single file. Smaller files make evidence easier to inspect and make ingestion warnings easier to understand.

Use stable filenames

Good filenames help with review:

chapter-01.md
chapter-02.md
book1-chapter-03.txt
appendix-world-rules.md

Avoid names such as final_final_really_final.txt or files that change meaning between imports.

Keep manuscript and notes separate

Arc can work with notes, but the first import should usually focus on manuscript source. Mixing draft text, outline notes, discarded scenes, and private planning in one batch can make canon extraction noisy.

A practical structure is:

manuscript/
  chapter-01.md
  chapter-02.md
notes/
  world-rules.md
  character-notes.md

Import the manuscript first. Add notes later if the workflow supports the distinction you need.

Preserve the source

Do not edit the only copy of a manuscript while testing. Keep a source archive so evidence references stay meaningful.

If you revise the manuscript later, treat that as a new import or a deliberate update rather than silently replacing the original source.