curl (https://github.com/curl/curl) is one of the most ubiquitous command-line tools when it comes to testing API communications. It allows us to learn the fundamentals of API communication with OpenAI, such as establishing connectivity, handling authentication, and processing responses directly in the command-line interface (CLI).

Getting ready

For many Linux distributions, curl is installed by default. We can check whether that is the case with the curl command:

$ curl --version
curl 8.5.0 (x86_64-pc-linux-gnu) libcurl/8.5.0 OpenSSL/3.0.13 zlib/1.3 brotli/1.1.0 zstd/1.5.5 libidn2/2.3.7 libpsl/0.21.2 (+libidn2/2.3.7) libssh/0.10.6/openssl/zlib nghttp2/1.59.0 librtmp/2.3 OpenLDAP/2.6.7

However, if it is not already installed, you can find the installation instructions at https://curl.se/docs/install.html.

Additionally, make sure you obtain OpenAI keys and have them ready to go (see the Setting up an OpenAI environment...

When it comes to API testing, curl is great at stripping away all the non-essential items, such as the graphical user interface. But once we are familiar with the basic process, we might want to move on to more polished tools with more refined user interfaces. After all, typing in command-line options every single time is probably not your favorite activity in the day.

Postman (https://www.postman.com/) is another popular tool for working with APIs. It elevates your API testing capabilities by introducing a user-friendly graphical interface. Postman provides a professional-grade interface for API testing, development, and collaboration.

This recipe covers the basic features for using Postman with the OpenAI API, such as environmental variable management, automated testing, and team collaboration.

Getting ready

To use Postman, we need to either use the Postman web-based interface or download the Postman application specific to the...

The OpenAI LLM base model, sometimes referred to as the foundation model, is amazingly functional and can handle most, if not all, the tasks we throw at it. As we have seen so far, it can help us with many general network engineering tasks, such as configuring VLANs, assigning IP addresses, and changing interface descriptions.

But one thing it does not currently do, as no out-of-the-box solution can do, is personalize the feedback for our specific organizational needs. For example, if our team wants to always use Loopback 0 for router-id when we configure BGP, the foundation model would not know that.

We can customize our OpenAI model for our specific needs by fine-tuning it. Fine-tuning allows us to teach OpenAI models our organization’s specific network configuration standard, for example, so that we do not just get the generic configurations.

In this recipe, we will learn how to prepare training data, execute the fine-tuning process...

Show me a network engineer and I will show you an engineer who spends countless hours putting together network diagrams, drawing lines to illustrate connection points, and even trying to pick the color scheme that matches their artistic taste. If this sounds like you, you will like this recipe.

In this recipe, we will demonstrate how to automatically generate network topology using OpenAI’s LLM. Traditional topology generation relies heavily on both the engineer’s network knowledge and their familiarity with tools. If the engineer is used to Visio diagrams or draw.io, those are probably the tools they will use. However, we can use OpenAI to automatically generate a network topology diagram by interpreting the topology description and producing code for visualization purposes.

Getting ready

We will use several Python packages to complete this recipe. Please install the following packages in your virtual environment...

Any network engineer working in a medium to large-sized enterprise environment is probably familiar with the MOP document. The MOP is generally a detailed document outlining the steps for executing a specific task or process, often used in network engineering operations. An example would be a MOP to create a new VLAN, find an available IP block to be used with the VLAN, and then assign a gateway IP to the VLAN interface.

The MOP focuses on the technical aspect of the process and is often used in a form of knowledge transfer between senior to junior engineers, or just as a simple reminder of steps when a sleepy engineer is performing the task at 3 A.M. While the MOP is useful, creating it can be pretty boring. In this recipe, we will show you how to leverage OpenAI LLMs to help you create one.

Getting ready

We just need the openai Python package for this recipe.

How to do it…

1. We will use the following...

In this chapter, we provided five comprehensive recipes that progressively built your expertise in leveraging OpenAI’s capabilities for network engineering tasks. Each recipe followed a practical, hands-on approach that you could immediately apply in your environment. More importantly, they were written with a network engineering focus, so they are relevant and applicable to our daily jobs.

In the next chapter, we will move on to one of the most important topics relating to AI LLMs: prompt engineering. See you there!