Help Center

1. Start with the source

Paste notes or import a UTF-8 TXT or Markdown file. Save the source before generating drafts. Markdown H1 and H2 headings split the text into up to 20 candidate articles per generation; fenced code stays literal. Repeating generation from the same source skips sections already added, and oversized sections are reported.

Importing a file keeps your pasted draft. Save source captures the notes you submit; new edits made while it prepares are kept for your next save. Switching centers keeps the import with its original center. If you reload or restore the workspace first, prepare the source again in the loaded workspace.

Drafts preserve source text and link to exact excerpts. This is deterministic section extraction, with no AI model, invented answers or automatic fact-checking. PDFs, scanned documents, web addresses and cloud connectors are not supported inputs.

2. Edit, link and review

Set each article's title, stable address, short description, category and keywords. Edit the Markdown body and choose which articles to include. Link to another included article using [[article-address]], or select a related article. HTML and external URLs remain literal text.

Check the wording and exact source excerpts, then enter a reviewer name and note. Article changes and replacement source versions clear affected reviews. Reviewer names are local labels you enter; they are not verified identities, account sign-ins or team permissions.

3. Save a reviewed release

A release freezes the included articles at that point in time. Unreviewed content, broken links, links to excluded articles and self-links block release creation. Later draft edits do not change earlier releases. Renaming an article address adds a moved-article page in a later export; old addresses cannot be reassigned to a different article.

Use the reader to check the release, follow internal links and print an article or save it as a PDF. Download the website ZIP, public JSON or separate article review CSV. The ZIP includes article pages, local search, categories, themes and print styles. Enter the final HTTPS directory URL on the release card to add canonical URLs, public article descriptions, a sitemap and article documentation. Leave it blank for a noindex offline review package; the address is kept only in this tab.

4. Share the right files

Extract the whole ZIP and open index.html, or upload the complete extracted directory to your chosen static host. Keep its scripts, styles and article directory together. Read the included PUBLISHING.txt and PUBLICATION.json before upload. A directory URL can include a subpath; the host-root robots policy and permanent moved-address redirects need configuration on your host. A noindex request is not access control. Search requires JavaScript; ordinary article links work without it. Search matches all query words, ranks title and keyword matches, and filters by category. It does not provide semantic or AI search.

Exports omit internal source records, reviewer notes and excluded drafts. An included article can still contain private text, so inspect every public page before sharing. Removing an article affects future exports only. You must replace or delete older copies on your host yourself.

Keep drafts and backups private

The workspace is saved in this browser's IndexedDB, with no cloud sync, uploads, analytics or model calls. The host receives ordinary page requests. Anyone who can use this browser profile may be able to access the saved workspace. Clearing site data or losing this browser profile can remove drafts that were not backed up.

Export a full backup to preserve source notes, drafts and releases; it contains private workspace data. Restore checks its structure and source hashes and asks before replacing current work. Competing-tab warnings help prevent stale writes. If browser storage is unavailable or full, keep the page open and export a backup of recoverable work.

Preview limits and unfinished features

One workspace supports 5 centers. Each center allows 30 source versions, 80 articles and 10 saved releases. Source text is limited to 100,000 characters; article bodies to 20,000; each article to 10 citations and 10 related links. Text files may be up to 1 MB, and the workspace or backup up to 20 MB. A release can retain up to 100 historical address pages.

This preview has no direct or Google account sign-in, account-owned cloud data, team editing, hosted publishing, custom-domain setup, subscription or paid access. Hosting, shared accounts and billing need separate implementation and verification. Source hashes help detect changed backup text; they do not prove that an answer is true or safe to publish.