Library

What you see

1. HeaderDocument and version counts, a link to canon with outdated evidence when there is any, and Write a document and Upload files for editors.
2. FoldersAll documents, Library root and your folders, two levels deep, each with its count.
3. TagsColoured labels that cut across folders. Click one to show only its documents.
4. ToolbarSearch titles and file names, filter by kind (All kinds), To re-check only, and Archived once a document is archived.
5. TableTitle with the kind icon, file name and tags; Folder; Last change (version, date and who); Cited (canon facts citing it); Re-check (canon citing an older version). Click a column to sort; titles sort with their numbers in order (Episode 2 before Episode 10).

Upload documents

1. Press Upload files, or drop files anywhere on the Library.
2. Pick Markdown, text, Fountain (.fountain), Yarn (.yarn), ink (.ink), CSV, TSV, JSON or Word (.docx) files up to 4 MB, or audio recordings up to 30 minutes. A file with no blank lines (a barks list, a CSV) becomes one passage per line.
3. Open a folder first to upload straight into it.
4. Follow each file under Uploads: a new file is added, a changed file with the same name becomes its next version, and a file that reads the same as that document’s latest version is skipped (line endings and spaces at line ends don’t count). Going back to an earlier draft’s text makes a new version.
5. A whole book can go in as one file: it is split into one document per chapter in a folder named after the book, and its row shows how many chapters the AI has read so far. Tick Keep as one document to import it unsplit.
6. The AI then reads each new version in the background. Its row (and its line in the table) says “Waiting”, “The AI is reading” with a timer, “AI critic checking”, then how many facts it found. You can keep working or leave the page meanwhile.
7. Open Review to decide on the candidates the AI proposed. If a run didn’t finish, the row says why; press Try again.

Organise with folders and tags

1. Press New folder (the plus next to Folders) and type a name; use a folder’s menu for New subfolder, Rename or Delete folder.
2. Press New tag (the plus next to Tags), name it, pick a colour and press Add tag.
3. Tick documents in the table, then use Move to, Tag or Kind in the bar that appears.
4. Use a row’s menu to Open it, see its History, Edit text or Rename it.

Archive old documents

1. Tick documents and press Archive, or use Archive in a row’s menu or on the document’s page.
2. Archived documents leave the Library list, reading order and book exports. Nothing is deleted: canon citing them stays canon and keeps its citations.
3. Tick Archived to see them; press Unarchive to bring them back. Uploading a new version of one brings it back too.

Find what needs a re-check

1. Tick To re-check only, or sort by Re-check.
2. Click a document’s re-check count to open its history and see what changed.
3. Or follow the header link to every canon fact with outdated evidence.

What happens to an upload

• Sources are stored exactly as uploaded and never edited. Folders, tags, kind and title are labels on top: changing them never touches the text.
• A whole book in one file (a manuscript, a Project Gutenberg text, a Word file with heading styles) is split at its chapter headings: “Chapter 7”, “CHAPTER IV.”, “Chapter One”, Prologue, Epilogue, Part headings, Markdown or Word headings, or numbered headings. A table of contents is not mistaken for chapters. Everything before the first chapter (title page, dedication, contents) becomes one “Front matter” document. A long text with no headings at all is cut at paragraph breaks into parts of about 5,000 words. Only a Project Gutenberg header and licence are left out; the chapters’ words are kept exactly.
• Uploading the same book again matches each chapter by its place in the book: an unchanged chapter is skipped, an edited one gets a new version (canon citing it is flagged to re-check), and new chapters at the end are added. A chapter inserted in the middle moves every chapter after it up one place, so each of those gets a new version. Word comments go to the chapter they are in; the .docx of a split book is not kept.
• Kind: Chapter (a novel) or Scene (a game or screen project) is the story’s own text. Impact lists every document a change touches, grouped by kind. The AI decides it when it reads a document (story text or reference such as lore, notes or an outline). Until then, a file whose name says lore, notes, characters, bible, outline, worldbuilding, glossary, wiki or readme is Lore and everything else story text. A kind set in the Library is never changed by the AI.
• A Word file is converted to text once: headings, lists, tables and footnotes are kept, tracked changes are taken as accepted, images are dropped. Word comments become notes on the commented passage. The .docx itself stays available from the reader.
• Macro-enabled, encrypted or old .doc Word files are refused.
• Audio (MP3, WAV, M4A, OGG, WebM, FLAC) is transcribed by AI into a read-only transcript with timed passages. Recordings count towards an hourly audio limit per person and per project.
• After import, the AI proposes candidate facts with a quote. Nothing becomes canon until a person accepts it in Review.
• An upload only stores the file, so it is done in a moment; the AI reads it afterwards, up to three documents at a time, and new candidates appear in Review as each one finishes. A run cut off by a server restart shows as not finished, with Try again.

Who can do what

Viewer

Browse, search, filter and sort, and open any document.

Editor, admin, owner

Upload files and recordings, write documents, archive and unarchive, and manage folders, tags, kinds and titles.

Roles are per studio. Roles and permissions has the full list.