Artifacts (Editor)

Longer, written-out content from the chat lands as an artifact in a dedicated editor with version history, auto-save and download (including Word) – instead of in the middle of the conversation.

When the AI writes a letter, an e-mail, a report or another piece of fully-worded text in chat, it would flood the entire conversation. Artifacts solve this: such content does not appear inline in the chat but in a dedicated editor – as clean, editable documents. All that remains in the chat is a compact card to click. Every artifact created this way is kept permanently, can be edited by hand, keeps a version history and can be downloaded, e.g. as a Word file.

What an "artifact" is

An artifact is a standalone document with a title and content that lives on independently of any single chat message. There are two types; the overview shows the type on each entry:

• Text – running text, headings, lists, tables, quotes. Texts are edited in the built-in WYSIWYG editor.

• Interactive – interactive or visual results (e.g. small tools, 3D scenes, elaborate charts). These are view-only – not edited in the editor – and open in a sandboxed preview.

Creating an artifact

Via chat (the usual way)

Simply ask the AI for written-out content, for example:

• "Write me a letter to …"

• "Draft a job advert for …"

• "Create a summary as Markdown"

The AI recognises such requests (letter, e-mail, cover letter, application, job advert, report, summary, note, minutes, "as Markdown" …) and automatically creates the content as a text. While it writes, you see:

1. a card inside the chat message showing the title and a writing… hint,

2. the content streaming live into the editor on the right.

By hand

Open the Artifacts overview page (/texts) and click + New Text in the top right. An empty text titled New Text is created and opened straight in the editor.

Both ways – in chat and by hand – require the Create artifacts permission. Members have it by default; without it, the + New Text button is missing and Kasimir answers directly in the chat instead of creating an artifact.

Managing artifacts – the overview page

Under Artifacts (/texts) you find all artifacts you can see as a grid of cards. Each card shows the title, the type, a short excerpt and when the artifact was last changed; artifacts from projects or teams carry a Project or Team badge.

Clicking a card opens the artifact full-screen at /texts/{id}.

The editor

The editor opens in two places: as a right-hand side panel in the chat (on larger screens) and as a full-screen page at /texts/{id}. Version history and marking a passage are only available in the side panel in the chat; the full-screen page offers the space picker and Delete instead.

Formatting toolbar

Above the text area (for the Text type only) sits a toolbar:

• Heading 1 / Heading 2 / Heading 3

• Bold, Italic, Strikethrough

• Bullet list, Numbered list

• Link (prompts for the URL), Code (inline code), Quote

You can also type Markdown syntax (e.g. ## for a heading or - for a list) and it is converted automatically. Tables are supported and rendered correctly too.

Renaming

Click the title at the top and type a new one. It is saved with Enter or when you leave the field.

Automatic saving

You never need to save: every keystroke goes straight into the shared document and is stored on the server. The status line at the bottom shows the state:

Version history

Every text keeps a version history. A new version is created:

• before Kasimir changes the text for the first time in an answer – if you have edited it since the last version (person icon),

• after every answer in which Kasimir changed the text (chip icon with the model used).

Open the history in the chat side panel via the ⋮ icon (Versions) at the top of the editor. The full-screen page has no version history.

On the left you see the version list (v1, v2, …) with timestamp and origin. Pick a version to view its content on the right. You can then:

• Copy – put the version’s content on the clipboard,

• Set as current version – replaces the content in the editor (visible at once to everyone reading along) and creates a new version for it. Only the person who created the text can do this.

Kasimir edits precisely

When you ask Kasimir in the chat to change an open text (“Change paragraph 2 …”), it no longer rewrites the whole text. It finds the spot and types the change in live – like a colleague in the same document:

• Its cursor carries the name tag Kasimir.

• The part it replaces is greyed out and struck through; the new text appears word by word in front of it.

• When it is done, the changed spot lights up briefly.

• You can keep writing anywhere else meanwhile, even in the same paragraph. Only the spot Kasimir is typing in is locked for a moment.

• Ctrl+Z only undoes your own input.

The card in the chat

Below the answer there is a card with a summary (e.g. “3 changes”) and the list of changed spots. Clicking a spot jumps there in the text.

• Discard changes undoes Kasimir’s changes from that answer. Spots you have edited yourself since then stay as they are – the card tells you how many.

• If Kasimir cannot find a spot unambiguously, it skips that spot. The card shows it with the reason and the suggestion (old → new); Kasimir knows about it in the next turn.

• If live editing is briefly unreachable, the change is shown on the card, and Try again sends the request once more.

Marking a passage

When you select a passage in the text in the chat side panel, an input field appears. Your instruction then applies exactly to that selection – even if the same wording appears elsewhere.

Reading along

All members see artifacts from projects and teams live – including Kasimir’s cursor, the grey and the highlight. Only the person who created the artifact can edit it; for everyone else it is read-only (status line “Read only”).

Download

Use the download icon at the top of the editor to download the text – as Markdown (.md), as plain text (.txt) or as a Word file (.docx). This is available for the Text type only, not for interactive artifacts.

Artifacts across multiple chats

Artifacts are not tied to a single chat. In every chat the AI knows:

• the text currently open in that chat, with full content – which it can revise directly,

• a list of your most recently created texts (titles only) – whose full content it cannot see.

When you ask the AI to revise, it changes the affected spots precisely (see “Kasimir edits precisely”) and creates a new version; it only rewrites the whole text for a rework of more than about half of it. The card in the original chat message stays pinned to the state produced back then – later edits never change old chat histories.

Interactive artifacts

If the AI produces an interactive or visual result, a card with a code icon appears in the chat. Clicking it opens a full-screen preview (lightbox). For security reasons interactive artifacts run in an isolated environment; they cannot be edited in the editor and cannot be exported as Word.

Deleting an artifact

On the full-screen page (/texts/{id}), the person who created the artifact finds Delete in the top right. After a confirmation prompt the artifact is removed irreversibly, including all versions.

Limits

• At most one new text per AI answer; at most 20 changed spots.

• Kasimir only changes the open or selected text. It knows other texts by title only.

• Images cannot be inserted in the editor.

• There is no comparison (diff) between arbitrary versions.

• Only you and Kasimir edit together; members of the project or team read along.

• Ctrl+Z does not undo Kasimir’s changes – use “Discard changes” and the version history for that.

Diagram blocks

A text can contain diagrams. Either you move one over from chat (Open as artifact) or you insert an empty diagram in the editor. Every diagram block has a toggle: Preview shows the picture, Source the Mermaid code for editing.

Changes run through the same mechanics as the rest of the text — auto-save and version history. Exporting as DOCX embeds the diagram as an image. Details at Diagrams.