Skip to main content

atmos ai chat

Use this command to start an interactive chat session with the Atmos AI assistant. Get intelligent help with Atmos concepts, configuration analysis, and best practices.

Experimental

Configure AI

Learn how to configure AI providers, skills, tools, and sessions in your atmos.yaml.

atmos ai chat --help

Description​

The atmos ai chat command opens a terminal-based chat interface where you can have extended conversations with the AI assistant about your Atmos infrastructure.

The AI assistant has access to your current Atmos configuration and can help with:

  • Explaining Atmos concepts and architecture
  • Analyzing your specific components and stacks
  • Suggesting optimizations and best practices
  • Debugging configuration issues
  • Providing step-by-step implementation guidance

Usage​

atmos ai chat [flags]

Flags​

--session <name>
Start or resume a named session. If the session doesn't exist, it will be created. If it exists, the conversation history will be loaded.

Examples​

# Start anonymous session (auto-generated name)
atmos ai chat

# Start or resume named session
atmos ai chat --session vpc-refactor

# Resume existing session
atmos ai chat --session my-session

Examples​

Basic Chat Session​

atmos ai chat

This opens an interactive chat where you can ask questions like:

You: What components are available in my configuration?
AI: Based on your configuration, I can see the following components...

You: How do I validate my stack configuration?
AI: You can validate stack configurations using...

Using with Different Providers​

All configured providers are available concurrently. Press Ctrl+P during a chat session to switch between them.

Interactive Features​

The chat interface supports:

  • Multi-turn conversations: Ask follow-up questions
  • Context awareness: The AI remembers your conversation
  • Markdown rendering: AI responses rendered with rich formatting (bold, italic, lists, tables)
  • Syntax highlighting: Code blocks displayed with language-specific syntax highlighting
  • History navigation: Navigate through previous messages with ↑/↓ arrow keys (like Bash history)
  • Multi-line input: Press Shift+Enter or Alt+Enter to add newlines in your messages
  • Skill switching: Press Ctrl+A to switch between 21+ specialized AI skills (e.g., atmos-terraform, atmos-stacks, atmos-validation)
  • Session management: Press Ctrl+L to switch between sessions without leaving the chat
  • Provider switching: Press Ctrl+P to switch between configured AI providers mid-conversation
  • Session history: See session metadata (name, created date, message count) in the header
  • Exit commands: Type exit, quit, or press Ctrl+C to close the chat

Keyboard Shortcuts​

Enter
Send message.
Shift+Enter or Alt+Enter
Add newline (multi-line input).
↑
Navigate to previous message in history.
↓
Navigate to next message in history.
Ctrl+L
Open session picker.
Ctrl+N
Create new session.
Ctrl+A
Switch AI skill.
Ctrl+P
Switch AI provider.
Ctrl+C
Quit chat.

History Navigation​

Navigate through your previous messages using the arrow keys, just like in Bash or Zsh:

You: What components are available?
AI: [Response about components...]

# Press ↑ to recall your last message
You: What components are available?  [← Previous message restored]

# Edit and resend, or press ↓ to return to empty input

How it works:

  • Press ↑ to recall previous messages, starting from most recent
  • Press ↓ to move forward through history
  • When you reach the end of history (↓ past the last message), your original unsent input is restored
  • Sending a message automatically resets history navigation

This is especially useful for:

  • Reusing similar queries with minor changes
  • Correcting typos in sent messages
  • Iterating on complex questions
  • Avoiding retyping long commands

Markdown Rendering & Syntax Highlighting​

AI responses are automatically rendered with rich markdown formatting for better readability:

Supported Formatting:

  • Bold and italic text
  • Headers and subheadings
  • Bulleted and numbered lists
  • Tables with proper alignment
  • Code blocks with syntax highlighting
  • Blockquotes and horizontal rules

Multi-line Messages​

You can compose multi-line messages using Shift+Enter:

You: I need help with a complex Terraform configuration. Here's what I'm trying to do:
[Shift+Enter]
1. Create a VPC with multiple subnets
[Shift+Enter]
2. Set up route tables for each subnet
[Shift+Enter]
3. Configure NAT gateways
[Enter to send]

This is especially useful for:

  • Pasting code snippets
  • Writing detailed questions
  • Formatting structured information
  • Creating multi-paragraph descriptions

Session Picker​

While chatting, press Ctrl+L to open the session picker:

Session List
↑/↓: Navigate | Enter: Select | Esc/q: Back | Ctrl+C: Quit

→ vpc-refactor (Jan 20, 14:30, 47 messages)
  rds-troubleshooting (Jan 18, 09:15, 23 messages)
  session-20251015-091032 (Jan 15, 09:10, 12 messages)

Session Picker Controls:

  • ↑/↓ or j/k - Navigate sessions
  • Enter - Switch to selected session
  • Esc or q - Return to current chat
  • Ctrl+C - Quit application

This allows you to seamlessly switch between different conversation topics without restarting the chat.

Provider Switching​

While chatting, press Ctrl+P to switch between configured AI providers:

Provider Selection
↑/↓: Navigate | Enter: Select | Esc/q: Back | Ctrl+C: Quit

→ Anthropic Claude - Industry-leading reasoning and coding
  OpenAI GPT - Most popular, widely adopted models
  Google Gemini - Strong multimodal capabilities
  xAI Grok - Real-time data access
  Ollama - Local models for privacy and offline use

Provider Switching Controls:

  • ↑/↓ or j/k - Navigate providers
  • Enter - Switch to selected provider
  • Esc or q - Return to current chat
  • Ctrl+C - Quit application

When you switch providers:

  • Message history is preserved - Continue your conversation with a different AI
  • Session is updated - The new provider is saved to your session
  • Seamless transition - No need to restart the chat or lose context

Example Use Case:

You: [Using Claude] Explain the difference between Terraform and Helmfile
Claude: [Detailed technical explanation...]

[Press Ctrl+P, switch to GPT-4]

You: Can you give me a simpler explanation?
GPT-4: [Alternative explanation with different perspective...]

This feature is useful for:

  • Comparing responses from different AI models
  • Using specialized models for specific tasks
  • Switching to local models (Ollama) for privacy-sensitive questions
  • Trying different providers to find the best fit for your workflow

Note: Only providers configured in your atmos.yaml will appear in the provider picker.

Common Use Cases​

Learning Atmos Concepts​

You: I'm new to Atmos. What should I know about stacks and components?
AI: [Provides detailed explanation of Atmos architecture]

Debugging Configuration Issues​

You: I'm getting a validation error for my vpc component. How do I fix it?
AI: [Analyzes the error and provides specific guidance]

Best Practices​

You: What's the best way to organize my stack configurations?
AI: [Suggests organizational patterns and best practices]