Quickstart: User — chat in 10 minutes

Goal

In about 10 minutes you will have signed in to your organization’s GENIE.AI assistant, asked a real question, inspected the sources behind the answer, switched the chat to your preferred language, and submitted feedback on the response.

You only need a browser — GENIE.AI runs entirely in the cloud, there is nothing to install.

Prerequisites

• The chat URL your organization gave you (for example https://genie.example.gov).
• A valid SSO account. GENIE.AI delegates authentication to Keycloak, so the same username and password you use for other internal tools will work — including brokered identity providers (Microsoft Entra ID, Google Workspace, Okta) if your administrator has wired them up.
• A modern browser with cookies enabled and JavaScript turned on (Chrome, Firefox, Edge, or Safari, current version).
• Pop-ups allowed for the chat URL and the Keycloak URL (the sign-in page opens in a redirect you must allow on first visit).

Step 1 — Open the chat and sign in

1. Open the chat URL in your browser. The first time you visit, a splash screen shows for a few seconds, then the app loads.
2. You are redirected to your organization’s Keycloak login page. Enter your corporate credentials (or pick the federated identity provider your admin enabled — Google, Microsoft, Okta, etc.).
3. After successful authentication, you are redirected back to the chat. The user icon in the top-right corner is now active, indicating you are signed in.
4. If your account has the admin role, the Administration icon (grid) appears in the navbar. Regular users do not see this icon.

Verify — type a single character in the chat box and delete it. The placeholder text “Type your query here…” reappears. You are now signed in.

Sign-in loop? If you sign in successfully but the page bounces back to the login screen, your browser is blocking third-party cookies. Either allow cookies for both the chat URL and the Keycloak URL, or ask your admin to enable Keycloak’s same-site cookie configuration.

Step 2 — Ask your first question

The chat area takes up the main panel of the screen. Above it is the response area, below it the input row.

1. Click the input box at the bottom of the screen. The placeholder reads “Type your query here…”.
2. Type your question in your own language. Examples:
   • “How do I apply for a building permit?”
   • “What documents do I need to renew my national ID?”
   • “Tell me about the affordable housing program.”
3. Send your message in either of two equivalent ways:
   • Press Enter inside the box.
   • Click the Send button on the right.
4. While the assistant thinks, a small spinner with the label “Thinking…” appears. As soon as tokens stream back, the answer renders in a bot message bubble. The current time appears under the bubble.

Verify — the bubble contains a full sentence (or a polite refusal such as “I cannot answer”). A timestamp appears under it.

Tip — multi-line questions. Press Shift+Enter to insert a line break without sending. The input box has four rows; long questions scroll inside it.

Step 3 — Read the sources behind the answer

GENIE.AI grounds every answer in documents from the organization’s knowledge base. Each bot message carries two transparency signals directly under the text:

• Confidence — a percentage shown next to a brain icon, e.g. Confidence: 78%. The score is the mean of the reranker scores for the passages used; higher = more strongly supported by the documents.
• Grounded vs. AI-generated — if the answer is fully grounded in library documents you see the confidence chip. If the assistant chose not to use any document, you see a sparkle icon with the label “AI-generated — not based on library documents”. Both are honest signals; neither means the answer is wrong.

For the full list of source passages, open the right sidebar by clicking the chevron at the screen edge (or use the toggle). The section “Related Documents” shows the top-N documents that informed the answer, with:

• Title and Document Name — the source file the chunk came from.
• URL — the clickable address if the source is an external link.
• Labels — taxonomy tags the retriever matched (e.g. Taxes, Identity).
• Confidence — the per-document score, on the same 0–100% scale.

Click any card to open the source. You can collapse the panel with the Collapse sidebar button at its top.

Why this matters. “Confidence: 24%” or an AI-generated flag tells you to double-check the answer before acting on it. “Confidence: 92%” with three matching documents tells you the answer is well supported.

Step 4 — Switch language

The interface and the answers can both be in your language.

1. The language selector lives in the top-right navbar (on mobile it sits next to the hamburger button).
2. Click it and choose your language from the dropdown. Supported locales out of the box include English, French, Spanish, Arabic, Bengali, German, Indonesian, Swahili, Mandarin, Thai, Portuguese, Russian — your deployment may expose a subset (your admin controls this via the VUE_APP_AVAILABLE_LOCALES config).
3. The whole UI re-renders immediately — buttons, labels, and menus — and the placeholder changes from “Type your query here…” to the equivalent in your locale.
4. Answers are translated on the fly. If the assistant cannot translate a passage (rare technical terms, names), the original wording is preserved in parentheses.

Verify — the placeholder text appears in the language you selected; the brand name in the navbar still reads “GENIE.AI”.

Locale not in the list? Contact your admin. They can add a locale by extending VUE_APP_AVAILABLE_LOCALES and shipping the corresponding translation file. The system already ships 14 locales; expanding is a configuration change, not a release.

See Restrict active locales for the canonical deployer reference.

Step 5 — Manage your conversation history

Your conversations live in the left sidebar under the second tab, “Saved Chats” (the History icon).

The Saved Chats tab has four sub-tabs at the top: All Chats, Folders, Starred, and Archived. Inside each:

• Click a conversation title to reopen it — the chat history scrolls back and you can continue the conversation.
• Hover a row to reveal action icons: Star, Rename, Move to folder, Archive, and Delete.
• Use the search box at the top to filter by title.
• Folders let you group related conversations. Click + Create Folder to add one, then drag conversations into it (or use the Move Chat action).

The first tab, Knowledge Areas, is the browsable catalog of services your admin has published (e.g. Taxes, Identity, Business Registration). Picking a service there adds it as Query Context — the assistant will then prefer documents in that area. Click the ✕ next to a context chip to remove it.

Want to start fresh? Click the + button (top-left of the input area) to start a new conversation. If you have unsaved messages, GENIE.AI prompts you to save them first.

Verify — start a new chat, send one message, then save it with the Save Chat icon (the floppy disk next to the +). Title the chat, pick a folder, and reopen it from the All Chats sub-tab.

Step 6 — Give feedback on the answer

Your feedback is the single most useful signal for improving the assistant. Each bot message has a small Feedback button underneath.

1. Click Feedback below the answer.
2. A dialog opens with two ways to rate the answer:
   • Thumbs up (positive) or thumbs down (negative), each with a human-icon picker for the colour tone.
   • A 1–5 rating scale: 1 Useless, 2 Slightly Helpful, 3 Moderately Helpful, 4 Very Helpful, 5 Life Changing.
3. Optionally type a one-line reason in the Additional comments… box (e.g. “Answer is out of date”, “Wrong department cited”).
4. Click Submit. A confirmation toast appears: “Thank you for your feedback!”.

Verify — the dialog closes; the rating and comment are saved against the specific bot message. Admins see aggregated feedback in the Analytics dashboard; nothing is shared with other users.

Why thumbs AND a rating scale? Thumbs are fast. The 1–5 scale is more diagnostic — a thumbs-down “Slightly Helpful” tells operators the answer was on-topic but weak, not wrong.

Step 7 (optional) — Bring your own document

Some deployments let users upload their own files (PDFs, Word documents, images, links) so the assistant can answer against your material in addition to the shared knowledge base.

1. In the chat area, click the paperclip / upload icon (if enabled by your admin).
2. Pick a file from your computer. The file is scanned (ClamAV) and queued for ingestion.
3. Wait until the file’s status flips from Pending to Ingested. This can take a minute for short files, several minutes for large PDFs — the ingestion log (visible in Admin → Document Repository) shows progress.
4. Once ingested, ask a question that references the document. The assistant will cite it in Related Documents.

If your admin has not enabled user uploads, the upload icon is hidden.

Privacy. Uploaded files are stored in your organization’s ArangoDB and are visible to admins for moderation and to other authorized users only when the admin explicitly shares the file with a knowledge area. Files are scanned for malware before they enter the pipeline.

Keyboard shortcuts

GENIE.AI follows common chat-app conventions so the keyboard does the heavy lifting once you learn them.

Screen-reader users: every interactive element exposes an aria-label — the hamburger is “Toggle sidebar”, the user icon is “User profile”, the logout icon is “Log out”, and so on. The chat window itself is a polite live region (aria-live="polite"), so new bot messages are announced without stealing focus.

Troubleshooting

FAQ

Where does my data live? Your chat history and uploaded files are stored in your organization’s GENIE.AI deployment — ArangoDB for conversations, document-repository for files. Nothing is sent to GENIE.AI upstream servers or to the language-model provider beyond the query text needed to generate the answer. See the architecture overview for the data flow.

Who can see my questions? The system operators (your organization’s admins) can read conversation logs and feedback for the purpose of moderation, debugging, and improvement. Other end users cannot see your conversations. Anonymous, aggregated metrics (counts of questions per day, satisfaction scores) appear in the analytics dashboard.

Is the answer always right? No. GENIE.AI grounds answers in your documents and shows the confidence score for each one, but the LLM can still make mistakes — especially on questions that fall outside the knowledge base. Treat a low confidence or AI-generated flag as a reason to verify before acting. The feedback you submit helps the operators find and fix bad answers.

Can I delete a conversation? Yes. Hover any conversation in the Saved Chats tab and click the trash icon. Confirm in the dialog. Deletion is immediate and irreversible — the conversation and its feedback are removed from the database (GDPR-compliant).

Can I export a chat? Click the Export Chat icon (a document icon next to the save button). Choose a filename and click Export. The chat is written to a PDF in your browser’s download folder.

Does the chat cost anything? GENIE.AI is sovereign software licensed under Apache 2.0 — there is no per-user fee to your organization. Your deployment’s compute and storage costs are the only costs involved.

What about my language? The chat UI is translated into 14 locales out of the box. The assistant answers in your selected language and translates documents on the fly. If a term has no clean translation, the original is kept in parentheses.

Where is the mobile app? A Flutter mobile client is available — see the Mobile app section. The web app also adapts to phones and tablets; the navbar collapses into a hamburger menu below 768 px.

Related

• Welcome — one-pager — the 30-second version.
• Concepts in 5 minutes — the query → embedding → retriever → reranker → LLM mental model.
• Glossary — RAG, embedding, reranker, BM25, OPEA, and the rest of the alphabet soup.
• FAQ — the 12 questions deployers ask first.
• Where to next? — paths for each persona.
• Knowledge base — content guidance — what the documents your assistant reads look like.
• Architecture overview — the system map behind the chat screen.
• Privacy & data handling — what is stored, where, and for how long.
• Report a problem with these docs — spotted a typo or a UI label that has drifted? Open an issue or MR.