# Chat API

Our Chat API allows a user to have a conversation with your AI agent via a simple API call. 

The **messages** array should contain the conversation between the user and the AI agent. You can send the entire conversation within the API request. 

Each message object has a **role** (user or assistant) and **content**. 

In the example below you can see a short conversation between a user and an AI agent. 

The message from the user has the **role: user** and all replies from the My AskAI have the **role: assistant**.

{% openapi src="<https://content.gitbook.com/content/hu0WRsaCQQRKP3KeuIOQ/blobs/BfcMHrCjypJvVnFki8RA/chat.yaml>" path="/ask-ai-chat" method="post" expanded="true" %}
[chat.yaml](https://content.gitbook.com/content/hu0WRsaCQQRKP3KeuIOQ/blobs/BfcMHrCjypJvVnFki8RA/chat.yaml)
{% endopenapi %}

{% hint style="warning" %}
If your AI agent doesn't know the answer, the API will also return an additional field to confirm an answer wasn't found as well as 3 suggested questions the user can retry (these have a very high likelihood of being answered correctly):&#x20;

`"unknown_answer": "yes",`

`"suggestedQuestions": ["Example question", ...]`
{% endhint %}

{% hint style="info" %}
You can also pass in `"insights": true` to generate insights in your Dashboard from any API usage.&#x20;

When Insights is enabled in the API a `conversation_id` is also returned with each request. This can be included in subsequent API requests to keep the requests link to 1 conversation. However, you still need to always pass in the message history as the API is stateless.
{% endhint %}

{% hint style="info" %}
If human handover [guidance](https://support.myaskai.com/features/improve/guidance) is triggered the API response will include: `"human_handover": true`
{% endhint %}