Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.atthene.com/llms.txt

Use this file to discover all available pages before exploring further.

The Session Runtime API allows you to execute messages in sessions, retrieve conversation history, and monitor session status. This is where the actual AI interactions happen.

​
Key Features

  • Message Execution: Send messages and receive AI responses
  • Multimodal Support: Text, images, and other content types
  • Conversation History: Retrieve message threads
  • Session Monitoring: Check session status and health
  • Correlation Tracking: Link related events and messages

​
Authentication

All endpoints require authentication using your API key:
  • API Key: x-api-key: <key>

​
Available Endpoints

Execute Message

Send a message and get AI response

Get Messages

Retrieve conversation history

Session Status

Check session health and status

​
Message Types

​
Text Messages

Simple string content for basic conversations: 
{
  "content": "What is the weather like today?"
}

​
Multimodal Messages

Rich content with images, documents, and other media: 
{
  "content": [
    {
      "type": "text",
      "text": "What's in this image?"
    },
    {
      "type": "image_url",
      "image_url": {
        "url": "https://example.com/image.jpg"
      }
    }
  ]
}

​
Response Handling

Runtime responses include:
  • AI-generated content: The agent’s response
  • Tool usage: Information about tools used
  • Metadata: Execution details and timing
  • Error handling: Clear error messages for issues

​
Correlation IDs

Use correlation IDs to:
  • Track related messages across sessions
  • Link events in analytics
  • Debug conversation flows
  • Implement custom logging
Always include correlation IDs in production applications for better observability.

​
Session State

Sessions maintain state between messages:
  • Conversation history: Previous messages and responses
  • Context: Accumulated knowledge from the conversation
  • Tool state: Persistent tool configurations
  • Memory: Agent’s working memory
Sessions have memory limits. Long conversations may need to be summarized or split into new sessions.

​
Error Handling

Common error scenarios:
  • Session not found: Invalid session ID
  • Rate limiting: Too many requests
  • Content policy: Inappropriate content
  • Tool failures: External service issues

​
Performance Considerations

  • Response times: Vary based on model and complexity
  • Concurrent requests: Limited per session
  • Message size: Limits on content length
  • Rate limits: Per-user and per-session limits
For high-volume applications, consider implementing request queuing and retry logic.