Search interface

The search interface is the main page of your portal (your-account.lynkora.pro). It is what your company's staff see during daily use.

#The 3-column interface

On a phone the interface switches to a single column with a bottom navigation bar — see LiKE on a phone.

#Portal header

A signed-in user has, on the right of the header, the notification bell (an unread badge; click it to open the feed) and the account menu under the avatar:

• Bookmarks — articles marked with a star; the list opens as a panel on the right, and an article from it opens right here;
• My Profile;
• Management — entry to the admin panel (everyone except a Guest);
• Logout.

For the Guest role this is the main entry to bookmarks and notifications: the admin panel is not available to them.

#How search works

1. The user types a question in natural language (in any of the 25+ languages)
2. LiKE finds relevant fragments via semantic search (by meaning, not by keywords)
3. The AI produces a concise answer based on the sources found
4. Below the answer, links to the source articles are shown (clickable)
5. Click an article → it opens in the center column
6. An article from a connected source has an "Open on the site" icon next to it — it opens the original page in a new tab. The main click still opens the article in LiKE

Search starts only on your command — with the "Ask a question" button or the Enter key.

#Search scope — a single space

Search is always performed within one active space. This preserves the accuracy of answers (context from different spaces is not mixed) and privacy (the content of each space is strictly isolated).

The active space is selected in the space switcher in the left sidebar:

Switching the space changes the search scope — LiKE searches only within the articles of the selected space.

If you already have an answer and then change the space, collection or article language, search does not restart on its own. The previous answer is removed: it belonged to a different scope. In its place a "Search scope changed" hint appears with your question and an "Ask again" button.

#How to get a more detailed answer

Ask for it in the question itself. Words such as "in detail", "detailed", "full list", "complete list", "step by step", "everything about" are a signal that search recognises before it calls the model, and from there it behaves differently:

• several times more documentation fragments go into the answer — where an ordinary question is served by the top of the results, a detailed one gets a substantially larger slice of the document;
• the answer length ceiling is raised so the explanation is not cut off mid-sentence;
• the model is instructed to unfold what it found rather than summarise it: every documented step with its place in the interface, every condition, limit and role requirement.

The signal is understood in every interface language — "подробно", "ausführlich", "en détail" work exactly like "in detail".

This costs tokens. Both sides grow: more documentation goes in, and a longer answer comes out — such a request consumes noticeably more tokens than an ordinary one. Ask for detail where you need it: on "how much is Pro" the detailed mode only spends quota.

Length follows the documentation, not the request. If your knowledge base says three lines on a topic, a detailed answer will not be longer — the assistant will say plainly where the documentation stops instead of padding it out.

#AI answer vs Articles

• AI answer — concise, to the point, synthesized from several sources
• Articles — full documents for a deep study of the topic

They complement each other: the AI gives a quick answer, while the articles provide context and detail.

#When an AI answer is unavailable

If the organization's monthly or daily AI-answer allowance is used up (or what remains is not enough for a whole answer), an anonymous visitor to the public portal keeps searching without AI:

• a question from the search bar finds matching articles by meaning; in the center, instead of an AI answer, there is a "Found in articles" block: a quote from the best-matching article with an "Open article" button and a "More on this topic" list;
• the right column shows "The AI assistant is resting";
• nothing is said to the visitor about limits or purchases.

Organization members see a notice that the limit is used up (see below).

#Actions when reading an article

When an article is open in the center column, its header offers the following actions:

If the article comes from a connected source, there is a "Source" line under the title with the address of the original page — the link opens it in a new tab.

At the bottom of the article there is a "Was this article helpful?" block with a quick rating ( / ). Ratings help you understand which articles work well and which ones need improvement.

Bookmarks and helpfulness ratings are available to authenticated users; they are not shown to anonymous readers of the public knowledge base.

Bookmarked articles are marked with a star next to the title: in Articles (list and tree), in the portal's article catalog, in quick search (⌘K / Ctrl+K), in the AI answer's Sources and in Recommended materials.

#If the AI-answer limit is used up

The question field is closed in advance, before sending. Above it you can see which limit ran out ("Daily AI answer limit reached" or "Monthly AI answer allowance used up") and when it comes back — in your time zone.

Article search always works and does not consume the allowance: the "Find an article" button or ⌘K / Ctrl+K opens search across the articles of the current space.

The owner sees what can be done and a button to go to the plan. Other members see a hint to contact the administrator and a "Notify administrator" button: the owner and administrators get one notification per day with the number of people who reported.

#Cross-language search

LiKE supports search in 25+ languages. Cross-language search: a question in one language → an answer from articles in another.

Which language the answer comes in:

1. If a language is selected in "My Profile" → "AI Answer Language" — in that language, whatever language the question was asked in.
2. If it is set to "Follow the question" — in the language of the question, which is detected automatically.
3. If the question's language can't be detected (a very short query) — in the interface language.

In the widget and for visitors who are not signed in there is no profile, so points 2 and 3 apply.

#Article catalog (left panel)

1. Grouping by sections (from the TOC — Table of Contents)
2. Filter by article language
3. Filter by source (LiKE Editor, Wiki.js, Confluence)
4. Click an article → it opens in the center column
5. Source badges next to the title

#Recommendations and history (right panel)

1. Recommended articles — only the articles the AI answer is based on; if there is no answer, the list is empty
2. Query history — your previous questions and answers in this organization; the history is stored in your account and is the same on a computer, a phone and in the app
3. Click a history item → it repeats the query

#Access control

Configuration: Settings → Organization → Access mode