First Run

You have Chvor installed and the server is running. This guide walks you through your first interaction — from opening the UI to watching the Brain Canvas react to a live conversation.

What you see at localhost:5173

When you open http://localhost:5173, Chvor loads with two main areas:

• The Brain Canvas — a full-screen, node-based visualization that takes up most of the viewport. It starts mostly dark, with a few idle nodes representing your AI’s core systems (memory, tools, active channels).
• The Chat Panel — a collapsible panel on the right side where you type messages and read responses.

If this is a fresh install, the canvas will be sparse. That is normal. It populates as you interact with the system, add tools, connect channels, and build up conversational memory.

The Brain Canvas explained

The Brain Canvas is not decorative — it is a live map of your AI’s state. Each node represents something real:

Node Type	What It Represents	Visual Cue
Core	The central LLM processing node	Large, always visible
Memory	Long-term memory storage	Pulses when memories are read or written
Channel	A connected communication channel (Web, Telegram, etc.)	Lights up on incoming messages
Tool	An MCP tool or custom skill	Flashes when invoked during a response
Skill	A custom skill you have defined	Distinct color, linked to Core

Edges between nodes show data flow. When a message arrives on a channel, you can watch it travel from the Channel node through Core, branch out to Tools or Memory, and flow back as a response.

Hover over any node to see its current status. Click a node to open its detail panel — this is where you can inspect tool call logs, memory entries, or channel configuration.

Starting your first conversation

1. Click the Chat Panel on the right side (or press / to focus it).
2. Type a message. Something simple is fine:

Hello! What can you do?

3. Press Enter or click the send button.

The AI responds using whichever LLM provider you configured during installation. The response streams in token-by-token, and you can see the Core node on the canvas pulse as it generates.

Try something that uses tools

If you want to see the canvas light up more dramatically, ask something that triggers tool usage:

What time is it right now?

or

Remember that my favorite programming language is Rust.

The first query invokes a tool (you will see the Tool node flash). The second writes to memory (the Memory node pulses). Both show data flowing across the canvas in real time.

Watching the canvas light up

As the conversation progresses, notice how the canvas evolves:

• New edges appear when the AI accesses a tool or memory for the first time in a session.
• Node intensity increases proportionally to how active that component is.
• The Core node color shifts based on the current processing state — thinking, generating, idle.

You can zoom and pan the canvas with your mouse (scroll to zoom, click-drag to pan). On touch devices, pinch and drag work as expected.

Canvas controls

The toolbar in the top-left corner of the canvas provides:

• Zoom to fit — reframes the view to show all active nodes.
• Toggle labels — show or hide node names.
• Pause animation — freezes the physics simulation if you want a static view.

Configuring your AI’s name and personality

Chvor lets you customize who your AI is — not just what model it uses.

Through the UI

1. Click the Settings icon (gear) in the top-right corner of the page.
2. Navigate to the Identity tab.
3. Set the following fields:

Field	Description	Example
Name	The name your AI uses to refer to itself	Atlas
Personality	A system prompt snippet that shapes tone and behavior	You are a direct, technical assistant. Avoid filler words. Use code examples freely.
Avatar	An image URL or uploaded file for the AI’s profile picture	—

4. Click Save. Changes take effect on the next message — no restart needed.

Through the config file

You can also set these in your .env file or ~/.chvor/config.json:

{
  "identity": {
    "name": "Atlas",
    "personality": "You are a direct, technical assistant. Avoid filler words. Use code examples freely.",
    "avatar": "/path/to/avatar.png"
  }
}

The UI settings take precedence over the config file. If you set a name in both places, the UI value wins.

Personality tips

The personality field is injected into the system prompt, so standard prompt-engineering techniques apply:

• Be specific. “Respond in two sentences or fewer” works better than “Be brief.”
• Define boundaries. “Never provide medical advice” is clearer than “Be careful about health topics.”
• Set tone with examples. “When greeting the user, say ‘Hey’ not ‘Hello, how can I assist you today?’” directly shapes output.

The personality applies across all channels — Web, Telegram, Discord, and Slack conversations all use the same identity unless you override per-channel in the configuration.

---

Next steps

Your AI is talking and the canvas is alive. Next, head to Configuration to set up additional LLM providers, connect messaging channels, and fine-tune your instance.