What Literess can do

Inside a document she sits in the left sidebar. Everywhere else she is the bubble in the bottom-right corner. What she sees in each place is under What does Literess see?

• Comment on a block, and apply the fix a workflow comment suggests.
• Show a block’s versions, switch the active one, or write a new version by hand.
• Show the document’s history: who changed what, and when.
• Set the document’s translation style, or switch a glossary or style guide on or off for it.
• Create a glossary or style guide from the conversation, add or edit glossary terms, merge one glossary into another. A glossary or guide she creates while a document is open is switched on for that document straight away; ask her to leave it off and it stays in your library only. This works from a glossary’s or style guide’s own page too, not only from a document, and the page shows her changes as she makes them, with no reload.

• Build a workflow, edit a saved one, run it on any document you name (the open one by default), stop a run, delete a workflow you no longer need.
• Run a built-in template, such as Quality checks or Fiction & literary, without saving it first. Customize on her card opens the template’s run dialog.
• Watch a run and report back in the same conversation when it finishes, fails, or pauses for a review.
• With the builder open, her edits land on your canvas with an Apply button. See Building & running your own.

• Create a document or a project, move documents between projects, set a document’s CRM status.
• Invite a teammate by email; change workspace defaults and your settings.

• Search your documents and past chats by meaning, read a document, or answer a focused question about one. “My piece about the festival” is enough.
• Search your translation memory for how a term or phrase was rendered before.
• Read a glossary or style guide and quote it back.
• Search the web when a question calls for it (below).

• Open a panel (glossary, style guide, comments), a page (CRM, billing, settings), or another document. When she opens another document, the conversation continues there.

She also remembers your preferences between chats (Memory & preferences) and can work through a longer goal step by step (Allow All & goal runs).

Before anything that changes your documents or library, she shows a confirmation chip with what she is about to do and, for a run, the estimated cost. Press it and she proceeds. Leave it and nothing happens.

On Starter and Pro, a message about your translation costs the credits her answer takes, and the range shows under the send button before you send. Block translations she writes are part of that cost, with nothing charged per word on top.

On Free, a block translation she writes because you asked costs 3 credits per source word when she writes it, whether you approve it or not. One she suggests on her own costs nothing unless you approve it, and then its price shows on Confirm.

When one reply proposes several actions, they arrive on one card. Approve all runs them together, each row keeps its own reject button, and Cancel all turns everything down. The cost is the same whether you approve actions one by one or together.

When she proposes a new version of a block that already has one of her proposals waiting, the older one is marked Replaced and only the newer one can be approved. A proposal to change glossary terms lists each term on her card: the ones she would add, the translations she would change, and the ones she would remove.

When the card holds several block edits on one document, Review in one screen opens them on the review screen: each block’s current text beside her version, which you can edit before you accept it. Accepting there costs the same and gets the same reply from her as approving on the card. Edits she already applied under Allow All show up there too, with Keep or Revert to previous.

While she works through a longer task, the review screen keeps up with her. A new edit to a block whose card offers options joins that card as one more option. Any other edit arrives as its own card, marked New, in place of her older card for that block or at the end. The card you are on stays in view. What you accept or reject reaches her before her next step, so one rejection changes how she handles the rest.

Once she has applied an edit, her card offers Review changes, even for a single edit. Edits she applies under Allow All are also listed under the reviews pending list, for example Literess · 2 changed.

While that document’s review screen is open, her card shows no review link, because her edits are already on the screen. It comes back when you leave.

Opening a panel or a page, or changing a setting on a screen you have open, needs no chip: it happens as soon as you ask.

To let her apply her chips without waiting for your click, turn on Allow All (Allow All & goal runs).

Allow All covers what you ask her to do. When she reaches out on her own, what she proposes still waits for your click, unless you also turn on Allow All on her own in Settings → Notifications. Taking the chat to another document always asks.

On the dashboard, she knows your name and plan, which page you are on, and which runs are going on any of your documents. She looks up your documents, projects and their CRM statuses when you ask about them.

In a document, she reads the page you are on: each block’s source text and active translation, the comments on those blocks, and the glossaries and style guides switched on for the document. She opens blocks on other pages when she needs them.

She also knows which block you are on, which panels are open and how many blocks you have selected. “This block” means the one you are on.

In the workflow builder, she sees the steps on your canvas and their settings, saved or not.

On any page, she sees the buttons, links and tabs on your screen, and can highlight one to show you where it is.

When a button is greyed out, she sees why, for example that no target language is picked yet in the upload dialog.

In a dialog or on the review screen, she reads what it shows and can change its settings for you:

• In a merge: which glossary you picked, which side wins each clash, and the counts in the preview.
• In a glossary or style guide generation: the name, the languages, and which suggested terms or rules are kept.
• On the review screen: which review you are on, how many items are left, and its layout, suggestions and model.

The button that finishes the job, such as starting a merge or a generation or accepting on the review screen, is yours to press: she sets the dialog up and tells you which button it is. With Allow All on, she presses it for you when you ask.

If she changes a glossary or style guide while a window about it is open, a notice says the window may be out of date. Reopen it to see her change.

On a desktop, the dialogs she can work on (merging, building, generating or filling a glossary or style guide, extracting terms, adding a language, and their reviews) leave her button clickable: open her and she sits beside the dialog. With a keyboard, the last stop in those dialogs is Open Literess assistant. Both stay usable, and Escape closes whichever one your cursor is in. Under any other dialog her button dims with the page, so open her first and then the dialog. On a phone the dialog fills the screen, so open her before or after.

She keeps a short record of your recent actions in Transept: the documents and pages you open, your edits, what you accept or reject in a review, changes to your glossaries and style guides, the files you import, the workflows you run and how they end, and when you step away. She reads it before she answers, so you can ask "why did that fail?" or "is this better?" without explaining what you did.

Each action is one line. When a line matters, she opens it for the detail, such as a block's text before and after your edit. The record covers the last seven days, and when a long chat is summarized, what you did goes into the summary. Your teammates don't see it.

She writes first only when something can’t wait:

• a run, import or export that failed
• a job that finished while you were somewhere else
• a spot where you seem stuck, such as undoing the same block again and again, or turning down her suggestions one after another
• your first translated document or your first finished run
• a document someone shared with you
• work due this week, when you open the dashboard

Anything else she notices, such as a faster way to do what you keep doing by hand, waits until you next write to her, and she brings it up only if it fits what you asked.

She sends at most two of these messages a day, at least an hour apart. If two go unanswered, she stays quiet until you write to her or a week passes. Once she has raised a topic, it doesn’t come back for a week, or for a month if you dismissed her message.

Tell her what you don’t want to hear about, for example “don’t mention glossaries”, and she remembers it (Memory & preferences).

She knows which page you are on, and can highlight the button or card she is talking about.

She doesn’t write while you are chatting with her, unless something failed, and she doesn’t repeat a point she just made.

Anything she proposes on her own waits for your click, even with Allow All on, unless you also turn on Allow All on her own (see below). Turn these messages off at Settings → Notifications → "Literess check-ins".

Paste, drop, or attach up to five files per message. She sees images natively (“what does this screenshot say,” “translate this menu”) and reads the text of documents (PDF, DOCX, EPUB, TXT, MD, HTML) you send for context.

Yes. When a question calls for it, she searches the live web as part of answering: a current fact, how a phrase is really used in the target language, a term of art, a cultural reference. There is no button to press and nothing to set up; a “Searching the web” pill shows while she looks, with the query she is running.

The pages she drew on appear as numbered sources under her reply, so you can check her reading yourself. Searching never spends credits.

Yes, once per reply, with her last message, whenever you are not looking at her chat:

• Elsewhere in Transept: a bubble next to her icon, with a chime.
• With the Transept tab in the background: a notification from your browser, if you allowed notifications.
• With Transept closed: a push notification, if push is on in Settings → Notifications.

A long goal run notifies you once, when it ends. Her replies don’t go to the Notifications panel: that one holds mentions and system events.