> ## Documentation Index
> Fetch the complete documentation index at: https://docs.athina.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# OpenAI Assistant

To log your OpenAI Assistant chats, you can log the messages using the [API logging](/api-reference/logging/log-via-api-request) method.

When you log to Athina, each chat message should be logged using a separate POST request.

You can group messages into a single conversation by using the `session_id` field. This is equivalent to the thread ID from OpenAI.

* **Method**: `POST`
* **Endpoint**: `https://log.athina.ai/api/v1/log/inference`
* **Headers**:

  * `athina-api-key`: YOUR\_ATHINA\_API\_KEY
  * `Content-Type`: `application/json`

#### Request Body[](#request-body)

Here's a simple logging request:

```cpp
{
    // Required fields for logging
    "language_model_id": "gpt-4",
    "prompt": [
        { "role": "system", "content": "System Instructions for the assistant" },
        { "role": "user", "content": "Individual message level instructions" },
    ],
    "response": "Response message from the assistant (as a string)",
    "session_id": "session-123",

    // Required fields for most evaluations
    "user_query": "The message from the user you are responding to",
    "context": "Your retrieved documents as a string or object", // Can be a string or object

    // Token usage logs (optional, but recommended). These can be found in the "Run" object from OpenAI Assistants API.
    "prompt_tokens": 123,
    "completion": 100,
    "total_tokens": 223,
    "cost": 0.12314, // if you provide tokens, we will calculate the cost for you automatically.

    // Optional fields for better segmentation and analytics (recommended)
    "prompt_slug": "customer-support/v1", // any string
    "environment": "production", // any string
    "customer_id": "disney-usa-2314832", // any string, this is your internal customer identifier
    "customer_user_id": "michaelmouse@gmail.com", // any string, this is the end user identifier,
    "external_reference_id": "239ej92-r24-4838-45663", // any string, this is your internal reference ID for the message,

    // Tool calls
    "tool_call": tool_calls, // can be any JSON object
    "tools": tools, // can be any JSON object

    // Ground truth data - used for some evaluations
    "expected_response": "If you have a ground truth response, you can send this here",

}
```

**Note:** *OpenAI Assistant's API doesn't provide the actual prompt that was sent.* *So, the prompt you are logging might not be the complete prompt that was actually sent to the model.*

*However, the prompt field is not currently used for evaluations so it will not affect the ability to run evals*