Using the Chat

The chat interface is where you talk to the chorus. Here’s everything you need to know.

The Layout

┌─────────────┬──────────────────────────────────┬──────────────┐
│   SIDEBAR   │            CHAT PANEL            │   CONTEXT    │
│             │                                  │    PANEL     │
│  Projects   │  ┌────────────────────────────┐  │   (toggle)   │
│  Convos     │  │  Messages appear here      │  │              │
│             │  │                            │  │   Memory     │
│             │  │                            │  │   injected   │
│             │  └────────────────────────────┘  │              │
│             │                                  │              │
│             │  ┌────────────────────────────┐  │              │
│             │  │  Type your message...      │  │              │
│             │  └────────────────────────────┘  │              │
└─────────────┴──────────────────────────────────┴──────────────┘

Three-Panel Design

• Sidebar (left): Projects, conversations, settings
• Chat Panel (center): Where the conversation happens
• Context Panel (right): Shows what memory is being used (toggle with button)

All panels are resizable. Drag the borders.

Sending Messages

Type your message in the input box at the bottom and hit Enter (or click Send).

Attaching Images

Click the 📎 icon or drag-and-drop images into the chat. Vision-capable models (GPT-4V, Claude 3) can analyze them.

Keyboard Shortcuts

The Provider Pill

Look at the header of the chat panel. You’ll see something like:

╭─────────────────────────────────────╮
│ 🤖 Claude Sonnet  │  $0.002  │  ⚙️ │
╰─────────────────────────────────────╯

This tells you:

• Which AI is currently responding
• Cost so far for this session
• Settings access

Changing Providers

Click the provider name to open the selector. You can:

• Let Chorum decide (Auto mode) — routes based on task type
• Force a specific provider — Claude, GPT, Gemini, or local models

Why force? Maybe you want Claude’s writing style, or GPT’s code generation, or you want to use a free local model.

Messages

Each message shows:

• Who sent it (you or the AI)
• The content (with markdown rendering)
• Cost (for AI responses)

Message Actions

Hover over an AI response to see:

• 📋 Copy — Copy to clipboard
• 🔄 Regenerate — Try again with fresh output
• 🤝 Get Second Opinion — Trigger peer review

Cost Tracking

Bottom-right of Each Message shows the cost:

$0.0012 (43 in / 127 out)

This means:

• $0.0012 spent on this response
• 43 tokens sent to the AI
• 127 tokens received back

Session Total

The header shows your running total for this session. Resets when you close the browser.

Don’t Want to See Costs?

Some people find it distracting. You can hide it in Settings → Preferences (coming soon).

The Context Panel

Click the toggle button in the header to show/hide the context panel.

This shows you what memory was injected into the AI’s context:

• Patterns from this project
• Relevant decisions
• Invariants being enforced

It’s useful for debugging why the AI knows (or doesn’t know) something.

Conversations

Each chat session is a conversation. You can:

• Start a new one: Click “New Chat” in sidebar
• Continue an old one: Click any conversation in the sidebar
• Name them: Click the title to edit

Conversations belong to projects. Switch projects to see different conversation lists.

Tips

Use Projects for Context Switch

Don’t mix topics in one project. Create separate projects for:

• Work stuff
• Learning projects
• Side projects
• Personal AI experiments

Each project has its own memory, so the AI won’t mix up contexts.

Let Auto Mode Work

Unless you have a specific reason, let Chorum pick the provider. The router considers:

• What you’re asking
• Cost per token
• Provider capabilities
• Your budget limits

Use Agents for Specialized Tasks

Instead of just chatting, pick a specific agent for structured tasks:

• Need analysis? Use Analyst
• Need code reviewed? Use Code Reviewer
• Need to learn something? Use Tutor

→ Next: Understanding Providers