Skip to main content

Contextual Chat

Contextual Chat allows your agent to maintain the conversation history — enabling multi-turn interactions where each new message builds on prior context.

This is useful for:

  • Building natural, back-and-forth conversations
  • Implementing agents with a short-term memory
  • Applications where follow-up questions or clarifications are expected

🔗 Endpoint​

POST /chat

Required Headers​

  • X-AGENT-API-KEY: Your API key

  • Model can be specified via:

    • X-Model header
    • model query param
    • model in request body

See Chat Overview for examples.

Scenario Example​

Customer Support Conversation​

An agent is helping a user troubleshoot an issue, across multiple back-and-forth messages:

Request:

{
  "messages": [
    {
      "role": "user",
      "content": "My internet is down."
    },
    {
      "role": "assistant",
      "content": "I'm sorry to hear that! Can you tell me what lights are showing on your router?"
    },
    {
      "role": "user",
      "content": "The power and DSL lights are on, but Internet is blinking."
    }
  ],
  "stream": false
}

Response:

{
  "status": "success",
  "status_code": 200,
  "message": "chat completed successfully",
  "data": {
    "Messages": {
      "role": "assistant",
      "content": "Thanks for the details! A blinking Internet light usually indicates that your router is trying to establish a connection. Try restarting the router and let me know if the light changes."
    }
  }
}

Usage Notes​

  • Include the full conversation history in each request
  • Messages array should alternate user / assistant roles as needed
  • Context length is limited by model capabilities (see Model Discovery)
  • Use when agents need to appear "conversational"

Next​