User Manual

1. Getting Started

2. Having a Conversation

The chat interface works like a professional AI assistant focused on financial topics.

Feature	How to use
Send a message	Type in the text box and press Enter or click the send button (↗).
New line	Press Shift + Enter to add a line break without sending.
Quick-start chips	On the welcome screen, click any of the suggested topic chips to instantly start a conversation on a common financial topic.
New conversation	Click the + button at the top of the sidebar to start a fresh conversation.
Switch conversations	Click any conversation in the left sidebar to open it. Your full history is preserved.
Delete conversation	Hover over a conversation in the sidebar and click the trash icon that appears.
Stop a response	Click the red Stop button (square icon) while Omnia AI is generating to halt the stream immediately.
Copy a response	Hover over any assistant message and click the copy icon that appears beneath it.

3. Voice Input & Speech Output

4. Attaching Files

Omnia AI can read and analyse documents you attach directly to a message.

1. Click the paperclip (📎) icon in the input bar.
2. Select one or more files from your device. Supported formats: images, PDF, TXT, CSV, JSON, Markdown, DOCX, HTML.
3. Chips appear above the text box confirming the files are attached.
4. Type your question (e.g. "Summarise this document" or "Find the compliance issues in this file") and press Send.
5. Omnia AI reads the file content and responds in context.

Note: File uploads in chat are separate from the Knowledge Base documents in the Admin Panel. Chat attachments are per-message; KB documents are permanent and available to all RAG queries.

5. Setting Up Your Workspace

A Workspace is a private, company-level environment. It holds your Knowledge Base documents and your team users. Only the Owner (the person who created the workspace) can access the Admin Panel.

6. Knowledge Base & Documents

The Knowledge Base is what makes Omnia AI answer questions specific to your organisation. When you upload a document, Omnia AI:

1. Extracts the full text from the file.
2. Splits it into small overlapping chunks.
3. Creates an AI embedding (vector) for each chunk.
4. At chat time — finds the most relevant chunks for the user's question and injects them into the AI's context window (RAG — Retrieval-Augmented Generation).

Document Statuses

Status	Meaning
processing	File is being read, chunked, and embedded. Usually completes in seconds to a few minutes depending on file size.
ready	Fully processed. All chunks are indexed and the document actively contributes to chat answers.
failed	Processing encountered an error. Hover the status badge to see the error message. Try re-uploading the file.

Allowed Roles

When uploading, you choose which roles can see a document's content in chat answers:

• Admin only — sensitive internal strategy documents.
• Admin + Employee — operational documents for staff.
• All roles — public-facing content customers can also see.

Upload Limits by Plan

Plan	Max Documents	Max File Size	Team Members
Free	2	10 MB	2
Go	5	25 MB	5
Pro	20	100 MB	20
Plus	50	200 MB	50

7. Managing Users & Roles

Add colleagues to your workspace from Admin → Users. The person must have signed in to Omnia AI at least once (so their account exists) before you can add them by email.

Role	What they can do	Documents visible
Admin / Owner	Full access — manage documents, add/remove users, change roles, view contact messages, access all admin pages.	All documents (admin, employee, customer tiers).
Employee	Chat only. No admin access. Receives detailed, operational-level answers.	Documents marked for employee or customer access.
Customer	Chat only. No admin access. Receives simplified, customer-friendly answers.	Only documents explicitly marked for customer access.

Important: Users added by you (subordinates) will not see the Admin Panel button in their sidebar — they only see the chat interface. Only you, the workspace owner, have admin access.

8. AI Models & Model Switcher

Each user can be given access to one or more OpenAI models (for example gpt-4o, gpt-4o-mini, gpt-4.1). The admin decides which models a user may use, and the user picks between them during a chat.

Assigning models to a user

Go to Admin → Users. You can assign models in two ways:

When	How
At creation	In the Add User box, tick the models the new user may use under Allowed models, choose a Default, then add the user.
Later (edit)	In the users table, click the “N models” button on a user’s row. Tick/untick models, set the default with the radio button, and click Save.

If no models are assigned, the user falls back to a standard default model so chat always works.

Switching models in chat

In the chat input bar there is a small model dropdown (similar to ChatGPT). It lists only the models that user is allowed to use, with their default pre-selected. Pick a different model at any time — the next message you send uses the selected model. The server always re-checks that the chosen model is permitted for that user.

Adding new models: the master list of selectable models lives in the app configuration (OpenAI:AvailableModels). When OpenAI releases a new model, an administrator adds its id there and it immediately becomes available to assign — no code change needed.

9. Custom AI Prompts

Prompts control how the AI behaves for your company. They are set at the company level — every user in your workspace shares the same prompts, based on your category.

1. Go to Admin → Prompts.
2. Click Create New Prompt, give it a name, and choose a type:
   • System — sets the assistant’s overall persona, tone, and rules.
   • Extraction — used when the AI needs to pull structured information out of text.
3. Write the prompt content and save. It applies to every conversation in your workspace.

You can view or delete existing prompts from the same page.

10. Data Sources & Live Feeds

Beyond uploaded documents, Omnia AI can pull knowledge automatically from external sources. This is managed by a super admin under Data Sources in the admin sidebar.

Types of data source

Type	What it does
URL Feed (XML / CSV / JSON)	Reads a structured feed from a web address and stores each record in the knowledge base. Ideal for lists such as sanctions / watch-lists (e.g. OFAC, UN). Add a User-Agent if the source requires one.
Website Scraper	Configured to read pages from a website using CSS selectors. (Ingestion engine is rolled out separately.)
File Upload	Upload a structured data file (CSV, Excel, JSON) as a one-off source.

Adding a URL feed

1. Go to Data Sources → + Add New Source → URL Feed.
2. Enter a name, choose a Category (or All Categories to share it with every company), pick the Format (XML / CSV / JSON), and paste the feed URL.
3. Optionally set a User-Agent and a sync interval, then click Create Feed Source.
4. On the dashboard, click Sync Now to fetch the data. The records become searchable in chat.

Category-aware: a data source tagged with a category is only used by companies of that same category. A source tagged All Categories is available to everyone. The same category rule applies to global documents uploaded under Global Documents.

11. Plans & Billing

Omnia AI offers multiple subscription tiers. View and upgrade from the Plans page.

Plan	Messages/Day	Docs	File Size	Users	Features
Free	10	2	10 MB	2	GPT-4o Mini, basic chat
Go	50	5	25 MB	5	File uploads, voice chat
Pro	200	20	100 MB	20	GPT-4o, image generation, faster speed
Plus	500	50	200 MB	50	GPT-4o + advanced models, priority speed, early access

To upgrade: go to Plans, select your plan, and complete the payment form. Your new limits activate immediately.

To view your current plan and usage: go to My Plan.

12. Frequently Asked Questions

13. Contact & Support