User Guide

1. Chat

The main interface for interacting with Zanqira. Ask questions, analyze documents, generate images, search the web, and get answers reviewed by multiple AI experts.

Sending messages

Type your message in the input bar at the bottom and press Enter or click the send button. The AI processes your message through a panel of expert modules (math, planning, coding, risk analysis, and more) and returns a synthesized answer.

Web search

Click the Search web button next to the input bar to enable web search for your next message. The AI will search the internet, gather relevant results, and incorporate them into its response. The button toggles on/off per message.

File attachments

Click the attach button or drag and drop files directly into the chat. Supported file types include documents (PDF, Word, Excel, PowerPoint, RTF, text), images, and code files. The AI will read and analyze the content of your attachments as part of its response. You can attach multiple files per message.

Image generation

Ask the AI to create an image (e.g., "draw a sunset over mountains") and it will generate one using the configured image provider. Click the generated image to view it full-size.

Chat history

All conversations are saved automatically. Access past chats from the sidebar under Your Chats. You can rename, export, or delete any conversation. Start a new chat anytime with the New Chat button.

Expert transparency

By default the AI gives you a clean final answer. You can switch to transparent mode to see how individual experts reasoned, what assumptions they made, their confidence levels, and where they agreed or disagreed. This gives you full visibility into the AI's reasoning process.

2. Voice & Live mode

Talk to the AI using your voice instead of typing. Three modes are available from the chat controls.

Voice input (microphone)

Click the microphone button to start speaking. Your speech is transcribed to text in real time. When you pause for a few seconds, the message is sent automatically. You can also click the microphone again to send immediately.

Voice Live

A hands-free voice conversation mode. Click Voice Live to start. The AI listens to you, processes your question, and responds by voice using text-to-speech. A floating panel shows the current status: Listening, Thinking, or Speaking. The real-time transcript appears below. The panel is draggable so you can position it anywhere on screen. All messages are saved to your chat history.

Live mode (camera + voice)

Click the camera button to share both your camera and microphone. The AI sees what your camera sees and can answer questions about your surroundings, objects, documents, or anything in view. A 3D avatar animates as the AI speaks. Both the camera feed and avatar panels are draggable. The AI analyzes frames periodically and can respond to what it sees in combination with what you say.

TTS settings

Customize the AI's voice in Settings → Voice & TTS. Choose a provider (OpenAI, Google, ElevenLabs, Coqui, or browser built-in), pick a voice, and adjust speaking speed. You can preview the voice before saving.

3. Meet — voice & video calls

Meet is the multi-person calling workspace on www.zanqira.us/meet. Use it for real-time voice and video with teammates while the default planner model turns the conversation into live notes—summary, working plan, decisions, and action items—so the meeting stays useful after everyone hangs up.

Starting or joining a call

1. Sign in on the platform, then open Meet in the sidebar (or visit /meet).
2. Create a new conversation/room, or open an existing one from the list.
3. Allow microphone (and camera if you want video). You can join audio-only.
4. Share the invite link (/meet?join=…) so others can enter the same room.
5. Start or join the active call from the room workspace.

Invite tokens can be pasted into native Chat apps (iOS / macOS) as a full Meet URL or raw token—the client accepts both.

In-call controls

• Microphone / camera — mute, unmute, or turn video off without leaving the room.
• Captions & transcript — speech is transcribed live and feeds the AI notes pipeline.
• Participants — see who is in the room; leave and rejoin if a peer fails to connect across networks.
• Files & media — attach supporting files to the conversation; previews open in-place when supported.

For reliable calls across different networks, administrators should configure TURN/ICE (COLLABORATION_ICE_SERVERS). Without TURN, both sides may need to be on the same network.

What the AI produces live

While the call is active, the default planner model continuously updates structured meeting intelligence from captions/transcript (and optional camera frames when vision is enabled):

Optional vision can add context from the camera (whiteboard, slides, physical docs) when enabled for the call. Notes improve when speakers name decisions and actions out loud (“We’ll ship X by Friday…”).

After the call

When the call ends, Meet runs a final consolidation pass and attaches the wrap-up to the conversation so you can reopen it later from the room history.

• Recordings — audio/video clips can be listed, previewed, and downloaded from the meeting insights panel.
• Object storage — media is stored in your configured backend (S3, MinIO, or local), depending on deployment.
• Conversation continuity — notes stay with the Meet room so chat and call context do not fork into separate tools.

Tips for better AI notes

• State decisions and owners explicitly (“Decision: …”, “Action for Alex: …”).
• Prefer one speaker at a time so captions stay clean.
• Keep the browser tab focused; grant mic/camera permissions when prompted.
• If peers cannot connect, leave/rejoin or ask an admin to enable TURN.

4. Tasks

A Kanban board to organize and track work. Open it from the sidebar or the dedicated Tasks page.

Board columns

Tasks flow through six columns:

Creating tasks

Click the + button on any column to create a new task. Give it a title and an optional description. Each task gets a unique identifier (e.g., MT-001).

Moving tasks

Drag and drop task cards between columns to update their status. The board saves automatically.

Running tasks with AI

Select a task and click Run to have the AI work on it. The AI uses the same expert pipeline as chat, processing the task description and returning results. You can view the execution history, including timestamps, status indicators, and result excerpts for each run.

5. Notes

Capture thoughts quickly with text or voice. The AI can then analyze your notes to extract action items and visualize topics as mind maps.

Creating notes

Open the Notes page from the sidebar. Type a quick text note or press the record button to capture a voice note. During voice recording you'll see a real-time transcription and a waveform visualization. Press stop when finished.

Calendar view

Notes are organized by date. Switch between Week, Month, and Day views using the toggle at the top. Navigate with the arrow buttons or jump to today with the Today button.

Note cards

Each note shows a title, preview text, tags, and a pin indicator. Voice notes display a waveform with a play button and duration. Click any note to open it for editing.

AI analysis

Open a note and click Analyze. The AI reads your note and produces:

• Action items — extracted tasks with checkboxes you can mark as done
• Mind map — a visual hierarchy of topics and connections
• Full analysis — a detailed summary with insights

Managing notes

Pin important notes so they always appear at the top. Search notes by text using the search bar. Delete notes by swiping on touch devices or using the delete button in the detail view.

6. Workflows

Build multi-step automations visually. Combine AI tasks, external tools, human approvals, and conditional logic on a drag-and-drop canvas.

Creating a workflow

Open the Workflows page and click New Workflow. Give it a name. The canvas opens with a palette of node types on the left.

Node types

Building flows

Drag nodes from the palette onto the canvas. Click a node to configure its properties (title, AI model, expression, timeout, etc.). Connect nodes by drawing a line from one node's output port to the next node's input port. Use the Condition node to create branches: the "true" output goes one way, "false" goes another.

Running & testing

Click Run to execute the workflow from its trigger. View results in the Execution History tab, which shows each run with step-by-step logs, timestamps, and status (success or error). Click Test to validate the workflow without executing it.

Saving & managing

Workflows save to the cloud automatically. You can rename, duplicate, or delete workflows from the list view. Zoom the canvas in or out with the + and − controls.

7. Sidebar & navigation

The sidebar is your navigation hub for all features.

Your Chats

A list of all past conversations. Click any chat to resume it. Use New Chat to start fresh.

Video chats

Past Live mode sessions (camera + voice) are listed separately so you can revisit them.

Quick actions

• Memories — view everything the AI remembers from past conversations
• World State — the AI's accumulated knowledge and beliefs
• Tasks — open the Kanban board
• MCP Marketplace — browse and enable plugins

Expert panel

See how many experts are active and their scaling status. Quick buttons let you Scale Up or Scale Down the number of expert modules. Toggle Auto-Scale to let the system manage experts based on demand. Open Model Settings or Platform Settings directly from here.

8. MCP Marketplace

Browse and enable tool integrations. The Marketplace lists available plugins (shell commands, file operations, web browsing, GitHub, Slack, and more). Each plugin shows its name, description, and supported platforms.

Toggle a plugin on to make it available in workflows and chat. Toggle it off to disable it. Enabled plugins expand what the AI can do—for example, enabling the shell plugin lets the AI run commands on a connected agent, and the browser plugin lets it interact with web pages.

9. Settings

Open Settings from the top-right corner. Here you can personalize the platform to your needs.

Profile

Upload or change your profile photo. Choose between light and dark theme. Switch the interface language (English or Russian). You can also delete your account permanently from here.

Voice & TTS

Pick a text-to-speech provider and voice for AI responses in Voice Live and Live modes. Adjust the speaking speed. Preview the voice before saving.

API Keys

Store your own LLM API keys (OpenAI, Claude, Gemini, etc.) securely with encryption. Set token usage limits per key with reset periods (hourly, daily, weekly, monthly) and monitor usage history.

Ollama servers

Connect your own Ollama server for fully local, private AI. Add the server URL, select default models for chat, coding, vision, and image generation. Test the connection to make sure everything works.

Expert models

Choose a default AI provider and model for all experts, or override specific experts individually. For example, use a fast model for coding tasks and a more capable model for complex reasoning.

Platform

Fine-tune how the AI reasons. Enable or disable features like debate & critique (where a critic reviews expert opinions), plan mode (create a plan before answering), web search, and image generation. Select which experts to activate and choose multi-expert verification modes.

Security

Enable two-factor authentication (2FA) for extra account security. Scan a QR code with your authenticator app and save backup codes. View connected devices and active sessions.

Channels

Configure messaging channel integrations (Telegram, Discord, Slack, Line, WhatsApp) so you can interact with the AI from your preferred messaging app.

10. Desktop & mobile apps

macOS apps

Two native macOS apps are available from the Downloads page:

• AGI Chat — a lightweight chat client for conversations with the AI
• AGI Dev — a full IDE with built-in AI assistance, code editing, terminal, and file browser

Download the DMG, open it, and drag the app to your Applications folder. The apps connect to the platform automatically.

Agent packages

Install the Zanqira agent on any machine (macOS, Linux, or Windows) to extend the AI's capabilities to that device. The agent auto-connects to the platform and appears in your sidebar. Available as DMG, PKG, DEB, and EXE from Settings → Agent Packages.

11. Messaging channels

Connect Zanqira to the messaging apps your team already uses. Once configured in Settings → Channels, you can chat with the AI directly from:

• Telegram — via a bot token
• Discord — via a bot in your server
• Slack — via a Slack app
• Line — via a Line channel
• WhatsApp — via webhook

Each channel can be enabled or disabled independently. Set allowlists to restrict access to specific users or groups.