AI Chat Interface

The left panel of the Workspace Editor is an AI-powered chat assistant specialized for infrastructure-as-code. It can write Terraform code, explain resources, troubleshoot errors, and execute Terraform operations through tool calls.

1. 1

   Sending Messages

   • Type your message in the text input at the bottom of the left panel
   • Press Enter or click the Send button to submit
   • Use Shift+Enter for multi-line messages
2. 2

   Voice Input (Speech to Text)

   • Click the microphone button (Dictate button) next to the message input
   • Allow browser microphone access when prompted
   • Speak your message clearly
   • The speech is transcribed using Microsoft Cognitive Services
   • The transcribed text appears in the input field
   • Edit as needed before sending
   • Click the microphone button again to stop recording
3. 3

   Model Selection

   • Click the model name displayed above the chat input
   • A dropdown opens showing all available AI models
   • Models are grouped by provider:
   • Anthropic: Claude Opus, Claude Sonnet, Claude Haiku
   • OpenAI: GPT-4, GPT-4o, GPT-4o Mini
   • Azure OpenAI: Azure-hosted GPT models
   • Google: Gemini Pro, Gemini Flash
   • Ollama: Locally hosted models (if configured)
   • Select a model to switch — subsequent messages use the new model
   • Different models have different strengths, costs, and context window sizes
4. 4

   Token Usage Display

   • Token usage is shown for each AI response
   • Displayed metrics:
   • Input Tokens: Tokens sent to the model (your messages + context)
   • Output Tokens: Tokens generated by the model (AI response)
   • Cached Tokens: Tokens served from the model's prompt cache (lower cost)
   • Use this to monitor costs and optimize prompt length
5. 5

   Message History

   • Previous messages in the workspace are loaded automatically on open
   • Scroll up to see older messages
   • Full conversation context is maintained between sessions
6. 6

   AI Capabilities

   • Write Terraform code for any resource type
   • Explain existing code and resource configurations
   • Troubleshoot Terraform errors
   • Suggest best practices and improvements
   • Perform infrastructure research (AWS, Azure, GCP documentation)
   • Execute Terraform commands via tool calls (requires approval)
7. 7

   Tool Approval / Confirmation Dialogs

   • When the AI wants to execute a Terraform command or modify files, a confirmation dialog appears
   • The dialog shows exactly what action will be taken
   • Click 'Approve' to allow the action
   • Click 'Deny' to reject the action
   • This ensures you remain in control of all infrastructure changes
8. 8

   Thinking Indicator

   • For complex requests, the AI shows a 'thinking' animation while processing
   • This indicates the AI is reasoning through your request before responding
   • Do not send additional messages while the AI is thinking
9. 9

   Streaming Responses

   • AI responses are streamed in real-time as they are generated
   • You can read the response as it appears without waiting for completion
   • A stop button appears during streaming if you want to halt the response
10. 10

    Code Blocks

    • Terraform code in AI responses is displayed in syntax-highlighted code blocks
    • Copy button: Copies the code to your clipboard with one click
    • Code blocks support HCL, JSON, YAML, Bash, and other formats
11. 11

    Chat Landing Page

    • When starting a new chat session, a landing page shows suggested prompts
    • Click any suggested prompt to pre-fill the input
    • Examples: 'Create an S3 bucket', 'Plan my infrastructure', 'Explain this error'