AnyLLM in Confluence — Chat Bar, Quick Actions, Page Assistant and Macro

This page describes every place where AnyLLM appears in Confluence (entry points, surfaces): the app page, the chat macro on a page, the assistant below the page title (byline assistant), the chat bar at the top of pages (page banner), the quick actions on selected text and on a whole page, and the fullscreen overlay. For each surface you learn how to get there, what it knows about the current page and which controls it offers.

What this does

The same AI with the same permissions runs on every surface. The surfaces differ only in size and in what they know: the app page knows no page; the macro, the page assistant, the chat bar and the quick actions know the page you are on and can include it in their answers. Everything personal — history, memory, saved choices, your own agents and skills — is the same on every surface and in Jira, because AnyLLM is one installation for both products.

Who can do this

Everyone with access to Confluence. Which surfaces are available is decided by your administration under Modules (see Modules: Switching Surfaces On and Off). Switched-off surfaces keep their menu entries but open only a short notice — except the chat bar, which disappears completely.

Before you start

• The app has been set up by your administration; otherwise you see a notice instead of the chat (see Getting Started).
• Page context is read with your permissions: what you may not see, the AI does not see either. If the administration has set AI access to Jira and Confluence to Off, no page content is sent — only the page quick actions are blocked entirely, see Page quick actions.

App page

Purpose — the full chat with the history sidebar, without page context. Permission — all users.

Steps

1. In the Confluence main navigation open Apps → AnyLLM Chat.
2. Type your question into the field Message AnyLLM … and press Enter.

Expected result — the chat opens with the sidebar on the left (Chat, Agents, Skills, History with New) and no header row. At the bottom of the sidebar you find Settings (administrators only), Switch to Jira ↗ and Open as full page ↗ (the chat on its own browser page, see Standalone full page). Export the conversation via ＋ → ↓ Export conversation (Markdown). The entry Include page context is not offered here because the app page knows no page. Confluence keeps its own page heading above the chat. The complete interface tour is on Getting Started.

Chat macro on a page

Purpose — embed the full chat directly in a Confluence page, e.g. on a team or project page; the chat knows that page. Permission — anyone who may edit the page can insert the macro; anyone who may view the page can use the chat (with their own history). Before you start — the module Chat macro is enabled.

Steps

1. Edit the page and type /AnyLLM at the position where the chat should appear.
2. Choose the macro AnyLLM Chat and publish the page.
3. Ask your question in the embedded chat.

Expected result — the chat appears in the page with a header row: ☰ (Show history / Hide history), the brand AnyLLM, the product button Jira (tooltip Open AnyLLM in Jira — opens AnyLLM in Jira), ↓ Export and + New chat (once the conversation has messages), ⤢ Fullscreen and, for administrators, Settings. On wide pages (about 900 px and more) the history sidebar opens by itself; on narrower pages open it with ☰. Include page context in the ＋ menu is switched on by default, so questions like "Summarize this page" work straight away.

Assistant below the page title

Purpose — open the chat about the current page without leaving it; the dialog is tall and narrow so the page stays visible next to it. Permission — all users. Before you start — the module Byline assistant is enabled.

Steps

1. On a page, click AnyLLM Chat in the line below the page title.
2. Ask your question.

Expected result — the chat opens in a tall, narrow dialog (declared as about one third of the window width and the full height) with the compact chat layout: header with ☰ (Show history / Hide history), the brand AnyLLM, ↓ Export and + New chat (once the conversation has messages), ⤢ Fullscreen (see Fullscreen overlay in Confluence) and Close. The history sidebar is hidden by default; ☰ opens it over the chat. Include page context is on by default.

Chat bar at the top of pages

Purpose — ask a question about the page you are reading with one click; the full chat opens as an overlay and the question is sent with the page as context. Permission — all users. Before you start — the module Chat bar (page banner) is enabled, the app is set up and the subscription is active. In all other cases the bar does not appear at all.

The bar sits above the page content and shows the field Ask AnyLLM anything — e.g. about this page …, the button Chat, an icon button with the tooltip Continue conversation (speech-bubble icon) and an icon button Dismiss (✕).

Steps: asking a question

1. Type your question into the field (up to 4,000 characters). Chat becomes active only once you have typed something.
2. Click Chat or press Enter.

Expected result — the fullscreen overlay AnyLLM opens on top of the page and your question is sent automatically with the page as context. The overlay looks like the app page (history sidebar on the left, expanded on windows of about 900 px and more). Close it with the overlay's own ✕; the field in the bar is empty again. If the app is not set up or the surface is switched off, the overlay opens with the corresponding notice and the question is not sent.

Steps: continuing your last conversation

1. Click the icon button Continue conversation.

Expected result — the overlay opens straight into the conversation you last worked in (the one with the most recent activity, regardless of pinning). While it loads, the input is locked and the chat shows Opening your last chat …; if you start typing or sending during that time, your own action takes precedence. Without any history an empty chat opens. Your own Include page context setting applies here (it is not forced on). If the history cannot be read, you see History could not be loaded. or Conversation could not be loaded.

Dismiss hides the bar for the current view only; on the next page load it is back.

If the overlay cannot open — on some surfaces the answer is shown directly below the bar instead. You can ask follow-up questions there (they continue the same conversation), copy the last answer with Copy (confirmed with Copied ✓), and Continue in chat → takes you to the app page, where the conversation is in your history. In this mode Continue conversation simply takes you to the app page. Answers shown in the bar are short by design (about 800 tokens), use no Jira/Confluence lookups, actions or web search, and must complete within 18 seconds.

Quick actions on selected text

Purpose — explain, summarize, translate or improve a passage of text with one click, without opening the chat. Permission — all users. Before you start — the module Text-selection quick actions is enabled. The selected text is sent along with your request, so these actions work even when AI access to Jira and Confluence is set to Off.

Steps

1. Select text on a page.
2. In the context menu that appears, click AnyLLM Chat. The dialog shows your selection as a quotation (Selected text) above four buttons.
3. Click Explain, Summarize or Improve writing — or, for a translation, click Translate: the first click reveals the selector into (English, German, French, Spanish; German is preselected in the English interface, English in the German one). Choose the language and click Translate a second time.

Expected result — while the request runs, all buttons are locked and Working … appears next to a spinner. The result is shown as formatted text; below it you find Copy (turns into Copied ✓), Continue in chat → and Close. Every result is also saved as a new conversation in your history. Continue in chat → takes you to the app page, where you open that conversation from the History list — the chat does not open on it automatically; the button is not shown if the app page cannot be resolved. Up to 6,000 characters of the selection are taken into account; answers are capped at about 800 tokens, use no Jira/Confluence lookups, actions or web search, and must complete within 18 seconds. Images in a result are not loaded; they appear as the link view external image.

Page quick actions

Purpose — summarize a whole page, extract its action items or translate it, without opening the chat. Permission — all users; the page is read with your permissions. Before you start — the module Page quick actions (••• menu) is enabled and AI access to Jira and Confluence is not Off.

Steps

1. On a page, open the ••• menu (top right) and click AnyLLM Chat.
2. Click Summarize page, Extract action items or Translate page (two clicks, with the into selector as above).

Next to the three actions there is Open full chat →, which takes you to the app page.

Expected result — as with selected text: Working … while running, then the result with Copy / Copied ✓, Continue in chat → and Close; the result is saved as a new conversation in your history. At most 8,000 characters of the page text are sent; longer pages are cut off and the AI is told so.

If the administration has set AI access to Jira and Confluence to Off, the dialog shows Page access for the AI is disabled by your admin — page actions are unavailable. and the three action buttons are greyed out; Open full chat → still works. Actions on selected text keep working.

Fullscreen overlay in Confluence

Purpose — get the full-size chat without leaving the page. Permission — all users.

Steps

1. In the macro or in the assistant below the page title, click ⤢ Fullscreen (tooltip Open the chat fullscreen) in the header — or use Chat / Continue conversation in the chat bar.

Expected result — the chat opens as a fullscreen overlay titled AnyLLM on top of the page. It looks like the app page: history sidebar on the left (expanded on windows of about 900 px and more), no header row, export via ＋ → ↓ Export conversation (Markdown). Include page context refers to the page you opened the overlay from. Close the overlay with its own ✕ — you are still on the page. Only if the overlay cannot be opened on a surface does ⤢ Fullscreen take you to the app page instead.

There is no separate module toggle for the overlay in Confluence: the overlay counts as part of the surface you opened it from — for the chat bar this is Chat bar (page banner).

Notices when a surface is unavailable

Menu entries of switched-off surfaces cannot be hidden per installation; they stay visible and open only the notice.

Common problems

More messages: Troubleshooting.

Related topics

• Getting Started
• Chatting with the AI
• Prompts, Skills, Memory and Model Parameters (＋ Menu) — Include page context
• Managing Chat History: Pin, Fork, Archive, Delete
• AnyLLM in Jira
• Modules: Switching Surfaces On and Off
• General Settings: Language, Default Model, AI Access, Rate Limit, Site Knowledge — AI access
• Troubleshooting · FAQ