Avatar Avatar

Get ready to get started using LLMs
Explore Generative AI capabilities
Learn how to use LLMs in your projects
Try the new DevNet LLM Learning Lab

Introducing the LLM

The buzz around generative artificial intelligence is at an all-time high as companies race to find innovative ways to apply it to their ventures. And while artificial intelligence (AI) is not a new concept, the advent of generative AI (also referred to as GenAI) opens new doors not just for pattern recognition… but pattern creation.

Many products leveraging AI are responding in pre-programmed ways to expected inputs. GenAI, on the other hand, trains Large Language Models (LLMs) to take inputs and create something brand new. LLMs can be thought of as the engine of GenAI. They are machine learning models that have been trained on vast amounts of data, and as a result, can understand and generate human language. These LLMs use deep learning to “self-train” (also known as unsupervised training) by using interconnected nodes not unlike the human brain. These LLMs can be proprietary to a company, open-source, or offered as a managed service.

How LLMs Are Used by the Cisco DevNet Community

Cisco DevNet recently conducted a survey on X (Twitter) and LinkedIn asking our community how they are currently utilizing LLMs. Survey takers were given the choices of code generation, summarizing the doc/task, image generation, and as a knowledge base. The results are shown below.

LLMs Results from survey presented via X (Twitter)

LLMs Results from survey presented on LinkedIn

According to the survey results in X/Twitter, the “As knowledge base” option took first place. For LinkedIn, the “Code Generation/Troubleshooting” option took first place; the second one was “As knowledge base.” As we can see, users also utilize Generative AI to Summarize the document/task. AI Assistants and LLMs help solve the tasks mentioned in the survey; in some cases, LLMs can show results faster and help discover topics compared to surfing the web and working with multiple resources.

Applying LLMs to IT and Network Engineering Use Cases

Knowing that our community has a vested interest in GenAI, how can we begin using LLMs for use cases more specific to IT and network engineering? Below, we present four examples of how LLMs are currently being used to transform the field.

Using LLMs with APIs

Developers and engineers work a lot with documentation, especially with API documentation. Different models allow you specific numbers of input tokens (the size of the prompt). For example, the open-source model Mistral 8x7B can operate with 32k input tokens; Gemini 1.5 Pro will enable you to create a prompt 1M tokens long. Developers can use API documentation, Postman collections, and OpenAPI specification documentation as a part of the prompt and ask LLM to create related scripts and instructions on how to complete associated tasks. For example, how to utilize the new Umbrella Tagging API.

Local Packet Whisperer

This open-source project lets you chat locally with PCAP/PCAG NG files. Local Packet Whisperer utilize open-source LLMs using Ollama, Streamlit and PyShark, allowing python packet parsing using wireshark dissectors.

Network Support Chatbots

How much time could you free up for your support team if they no longer had to answer repetitive questions? And how many more customers could you assist if you provided 24/7 support? Network support chatbots can be used internally by IT and support engineers to assist with troubleshooting, as well as externally by customers to get answers to common questions or to guide them through basic troubleshooting. For example, T-Mobile has implemented a chatbot that can help customers troubleshoot common issues even if human support engineers aren’t yet online.

AI Assistants

AI Assistant capabilities vary, and may include:

  • task automation
  • performance optimization
  • configuration management
  • health and security monitoring
  • troubleshooting
  • streamlining user interactions

AI Assistants in network engineering exceed the capabilities of traditional chatbots and can become a powerful tool for a variety of tasks, which is why Cisco is currently in development of an AI Assistant for Security.

Get Started Using LLMs

Ready to get started using LLMs? Explore Generative AI capabilities and learn how to use LLMs in your projects with this Cisco DevNet Learning Lab.

Keeping Up With GenAI



Authors

Erika Dietrick

Developer Advocate

Cisco DevNet

Oleksii Borysenko

Developer Advocate

DevNet