A trained document already works: someone opens the link, types a question, gets a cited answer. But before they type anything they meet a title, a category, maybe an intro message, maybe a disclaimer they have to accept, and — invisibly — an access rule that decided whether they got in at all. All of that lives in the document editor, and this post walks the first half of it.
Open a document and choose Edit (or Configure from its row actions). On desktop the tabs sit in a left sidebar grouped into Content, Configuration, Learning and, for super admins only, Admin. On mobile, tap the section button and pick from the Options sheet.
How saving works
The editor saves two ways. Toggles, dropdowns and radio choices save the moment you change them, showing a Saving… spinner and then a green Saved pill. Text fields — titles, labels, email addresses, intro and disclaimer text, custom prompts — plus the access level and passcode wait for Apply (which saves and keeps the editor open) or Save & Close. Try to close with unsaved changes and a dialog offers to save or discard.
If you cannot remember which tab something is on, Search settings… in the header, or Cmd+K (Ctrl+K on Windows), searches every setting by name and shows its current value.
The Checklist tab is the other shortcut. It groups settings into Essential, Recommended and Optional, shows each one's current value or a placeholder where it is empty, and filters to the rows needing attention. Clicking a row jumps to its tab. For a freshly trained document, start here.
Overview: identity and address
The Overview tab holds the document's identity. Under General, Document status controls whether the document is available to readers — the same switch as Active document on the Settings tab, so flipping either flips both. Document URL is the short identifier at the end of your share link, something like kdigo-ckd-2024. It saves when you leave the field, must be unique, and has a copy button and an Open link; if your organization has a custom domain, the copied link uses it. Document ID below it is read-only, and useful when contacting support.
Under Basic information: Title is what readers see in chat and in listings. Category opens the same searchable picker used by the inline chips in your document library. Year accepts 1900 to 2100. Language is English or Français and sets the reader-facing labels. Translations links this document to its language equivalents — click Add translation, pick another of your documents, and readers can then switch languages from within the chat.
Messages: what greets a reader
The Messages tab is where the pre-conversation experience is built, across four sub-tabs.
Introduction holds the Intro message shown in the chat sidebar or welcome area. It supports basic HTML — paragraphs, bold, italics, lists and links — with a live preview underneath. Beside it, Regenerate abstract & keywords asks the AI for a fresh summary, sample questions and keyword cloud, and shows you the result before you commit. Keywords appear as a ranked list under Keywords (N), most important first, sampled from the whole document rather than just its opening. Click Apply to replace your intro, or Cancel to keep what you have.
Composer prompt is an optional card above the chat input, shown before the first message. Type its content in Prompt above chat input — leave it empty and the card disappears. Card label gives it a short header ("Getting started"), with separate English label and French label fields. Allow user to dismiss is on by default; Hide after chat begins is off. When either is on, Dismiss persistence decides how long a dismissal sticks — All conversations for about a year, Session only for the current browser session.
Explainer is a short intro telling readers that this assistant answers only from its training material, says when it does not know, and will not make things up. It is off by default and shown once per session. Leave Explainer text empty for the built-in wording, or write up to 10,000 characters of your own. See how grounded answers work for what it is describing.
Disclaimer goes further: when Show disclaimer modal is on, readers must accept before they can use the document, with I agree and I decline buttons. Leave Disclaimer text empty for the generic default, or write your own with a live preview. Medication warning override sets whether this document shows the medication warning independent of your organization's setting: Inherit from owner, Always show, or Never show. If both explainer and disclaimer are on, the explainer comes first.
Media, helpers, and the abstract
The Images tab manages the cover image and the document's image library, extracted images included. Its Gallery Tool sub-tab picks and orders which images appear in the chat's lightbox gallery; empty the gallery and the Image gallery toggle in UI Options switches off by itself. Videos works the same way for Vimeo links. See images and video.
Helpers adds optional questions above the chat so visitors can supply structured context — patient details, for instance — that travels with their messages. Enable Helpers is off by default and everything else appears only once it is on: Wizard on first message shows a step-by-step form the first time a reader focuses the chat box, Section Title (defaulting to "Patient Information") and Description label the block, and Helper Fields is the builder. This tab saves automatically. See helpers and guided context.
Abstract generates an AI-written overview of the whole document from its processed content. Generation runs in the background with a cooldown between regenerations, and if the document is retrained afterwards the tab flags that the abstract may be out of date. Generating one does not show it to anyone — the Abstract toggle under UI Options puts the button in the chat toolbar.
Settings: who gets in, and who can find it
The Settings tab is the one with real consequences. It repeats the Active document switch, and then Access Level offers six tiles:
| Option | Who can open it |
|---|---|
| Public | Anyone with the link, no login |
| Owner admins only | Administrators of your owner group |
| Registered | Any logged-in user |
| Owner group | Members of your owner group |
| Passcode | Anyone with the link who enters the passcode |
| Token link | Anyone with a share link you generate, no login |
Two constraints apply. On plans without public sharing, Public and Passcode show a lock and open an upgrade prompt instead of changing anything. Owner admins only and Owner group stay disabled until the document has an owner, with an Owner required notice explaining why. Choosing Token link opens a manager for named tokens with optional expiry dates, use limits, copyable URLs and usage counts; the document must be saved first. See access levels explained.
Two more toggles decide where the document shows up: Show in owner library puts it in owner-scoped browse lists, Show on landing page on your organization's landing page and gallery. Both are on by default, and a hidden document is still reachable by direct URL.
Visibility & Search handles the rest. Discoverable in search lets the document appear in Docutrain search results, but only if the access level is Public too — a live status line tells you whether it currently will. Suggest related documents and Include in related suggestions control the two directions of related-document suggestions when an assistant cannot answer; both default to on. Include in sitemap appears only for public documents.
That is everything a reader meets before typing. What happens after — tone, memory, which tools appear — is in tuning the chat experience.