Messages

What Are Messages?

Every interaction with a NovaAI chat model is represented as a sequence of messages. Rather than sending isolated prompts, you provide the conversation history, allowing the model to understand context and generate more relevant responses.

Each message includes a role and content, enabling the model to distinguish between system instructions, user input, and assistant responses.

Message Structure

Each message is represented as a simple object.

{
  "role": "user",
  "content": "Explain how vector databases work."
}

Defines who authored the message.

The text, instruction, or information sent to the model.

Message Roles

NovaAI supports three primary message roles.

Using these roles correctly helps the model maintain consistency throughout a conversation.

Conversation Flow

Conversation Example

1[
2  {
3    "role": "system",
4    "content": "You are a helpful technical assistant."
5  },
6  {
7    "role": "user",
8    "content": "What is an API?"
9  },
10  {
11    "role": "assistant",
12    "content": "An API allows different software applications to communicate with each other."
13  },
14  {
15    "role": "user",
16    "content": "Can you give me a real-world example?"
17  }
18]

Notice how the model receives the previous assistant response along with the latest user message. This allows it to continue the conversation naturally.

Message Lifecycle

Conversation History

A conversation grows over time as more messages are exchanged.

Keeping relevant history improves contextual understanding while removing unnecessary messages helps reduce token usage.

Best Practices

• Start every conversation with a clear system instruction.

• Keep user messages focused on a single objective.

• Include only relevant conversation history.

• Remove outdated messages to reduce token usage.

• Preserve important context for long-running conversations.

• Separate unrelated topics into different conversations whenever possible.

Common Mistakes

Frequently Asked Questions

Next Step

Now that you understand how conversations are structured, continue to Tokens to learn how message length, conversation history, and model context affect token usage and overall API cost.