General Settings: Language, Default Model, AI Access, Rate Limit, Site Knowledge

This page explains the site-wide base settings of AnyLLM on the settings page General: the interface language (UI language), the default model (preselected model), the AI access policy for Jira and Confluence (tool level, read/write permissions of the AI), the hourly rate limit per person and the site knowledge (organisation background for the AI). The third card of the same page — web search and image generation — has its own page: Enabling Web Search and Image Generation.

What this does

General holds the settings that shape every chat for every user in Jira and Confluence alike. They are part of the one shared configuration of the installation — there is nothing to repeat in the other product.

Who can do this

Product administrators (Jira or Confluence administration) on the settings page AnyLLM Chat — Settings — see Opening the Settings and Initial Setup.

Before you start

At least one saved connection with an API key (Setting Up LLM Connections); otherwise the Default model list is empty.

Interface language

Purpose — decide which language the AnyLLM interface uses.

Effects

• Applies to the chat on all surfaces, the settings pages, the Confluence quick actions and chat bar, the dashboard gadget and the labels on archived chat pages in Confluence (those are German only if this setting is German (Deutsch) — see Managing Chat History).
• Menu titles in Jira and Confluence (e.g. AnyLLM Chat, AnyLLM Chat — Settings) always follow the person's own Atlassian language, whatever you choose here.
• The settings page switches its language immediately when you change the selection — even before you click Save. Users see the change the next time they open the chat.
• Messages generated on the server (validation and error texts) stay English in both languages, with one exception: the notice about a missing endpoint approval is shown in the interface language.

Default model

Purpose — the model preselected for everyone who has not chosen a model themselves.

Steps

1. Under Default model, pick a model. The list groups the models by connection name and offers the models of all saved connections — including connections without a stored API key.
2. Choose a model whose connection has a key.
3. Click Save.

Expected result — new users and users who never changed the model get this model. A person's own choice (remembered per person, the same in Jira and Confluence) takes precedence.

AI access to Jira & Confluence

Purpose — decide what the AI may do inside Jira and Confluence, for every user and every agent.

Effects

• The AI never sees more than the person asking; every lookup runs with that person's Atlassian permissions. Even with Read + write nothing is written without a click on ✓ Execute (Letting the AI Search, Read and Create in Jira and Confluence).
• With Off, page and issue context is not attached to messages at all, and the page quick actions in Confluence (summarize page, extract action items, translate page) show "Page access for the AI is disabled by your admin — page actions are unavailable." with their buttons greyed out. Actions on selected text keep working, because the selection travels with the request.
• An agent's Jira/Confluence tools setting can only narrow this policy, never widen it (Providing Agents, Prompts and Skills).
• Web search and image generation are independent of this setting; they are enabled separately (Enabling Web Search and Image Generation).
• Tightening the policy takes effect immediately: when you leave Read + write, clicking ✓ Execute on a suggestion card that is still open fails with "Write actions are not enabled by the admin."

Rate limit

Purpose — cap how many requests one person may send per hour, as protection against cost and misuse.

Steps

1. Enter a number from 0 to 100,000 in Rate limit: max requests per user and hour (0 = off).
2. Click Save.

How the limit is counted

• Per person, not site-wide — and one counter covers Jira and Confluence together.
• The hour is a fixed window that starts with the person's first request and ends one hour later; then a new window starts. Blocked attempts do not extend the window.
• Every send counts: questions in the chat, ↻ Regenerate answer, Edit & resend, skills, the Confluence quick actions and chat bar, and each ↻ Refresh of a dashboard gadget (counted for the person who clicks).
• If the counter cannot be read for technical reasons, the request is allowed rather than blocked.

Expected result — once the limit is reached, the person sees "Hourly limit reached (N requests/hour) — please try again later." and can continue after the window has passed.

Site knowledge

Purpose — give the AI background about your organisation that it should know in every conversation: project keys, space conventions, team terminology.

Steps

1. Write the text into Site knowledge (for the AI) — at most 4,000 characters.
2. Click Save.

Example: "Jira projects: ABC = product development, OPS = operations. Confluence: the DOCS space holds the customer documentation. We call sprints 'waves'."

Effects — the text is sent to the AI service with every request on every surface, in Jira and Confluence alike, and thereby shapes the answers of all users. Changes apply from the next request.

⚠️ Warning: Do not put secrets, credentials or personal data into the site knowledge. Everything you write here goes to your AI service and influences every user's chat.

Common problems

How Save works on this page

Save on General saves the whole configuration in one go — including any connection cards you have edited but not yet saved under Connections, together with API keys typed there, and the module toggles. If a new endpoint address is involved, the Atlassian approval dialog opens first; declining it means nothing is saved (Managing Endpoint Approvals). The status next to the button reads Saving …, then Saved. or Save failed.

Related topics

• Administration & Configuration
• Opening the Settings and Initial Setup
• Setting Up LLM Connections
• Managing Endpoint Approvals
• Providing Agents, Prompts and Skills for Everyone
• Enabling Web Search and Image Generation
• Letting the AI Search, Read and Create in Jira and Confluence — the user side of AI access
• Permissions, Roles and Data Visibility
• Troubleshooting
• FAQ