# Paragon Automation - Ask Paragon

## Overview

In this lab, we introduce a groundbreaking context aware LLM connector (Also known as Ask Paragon), allowing users to bring the power of LLM to Paragon Automation.

Ask anything using NLP (Natural Language Processing) and this AI system can comprehend and respond to a vast array of queries providing detailed answers. Users only need a high-level understanding of the network and do not need to know Junos CLI commands or Paragon automation API’s . The LLM connector will provide answers to questions that will aid not only in troubleshooting network problems but also enabling observability of network in a simplified manner. 

We will also introduce the participant to usage of prompt engineering with Paragon Automation and examples on how to use the llm optimally.

<div class="alert alert-info">
    <strong>Note:</strong> Name of the feature is subject to change
</div>

## Starting Lab

The objective of this Blueprint is to use Ask Paragon to answer questions, which is an NLP based AI assistant that uses a LLM to get answers to questions. showcase the power of LLM in Network Automation. 

Use cases are as follows

-	Use of natural language to converse with the LLM and get answers, high level knowledge of networking is good enough.
-	Prompt Engineering for Troubleshooting and root causing
-	Tracking config differentials
-	Language specific Personalization

### Lab Blueprint Topology

The blueprint multiple windows VM’s that each participant will be allocated. 

We also have a Paragon Automation 2.0 on-prem instance onboarded with devices. 

![Topology](./images/Topology.png)

### Connect to Jumpbox

1.  Using RDP Client, connect to your Labuser assigned IP address/Port

    ![windows_rdp](./images/windows_rdp.png)

## Login to Paragon

1.  Open **Chrome**, and navigate to https://100.123.42.100, you shoud see the login page to Paragon Automation UI

    ![paragon_login_page](./images/paragon_login_page.png)

2.  Login using the following credentials:

    -   **Username** - jcluser@juniper.net
    -   **Password** - Juniper!1

## Invoke ASK-Paragon

1. In the bottom right-hand side of the Paragon UI, you will see the **Ask Paragon** icon. Click the Icon.

    ![ask_paragon](./images/ask_paragon.png)

2.  You should see the Ask-Paragon UI ready for use.

    <div class="alert alert-info">
    <strong>Note:</strong> Note that this is context aware , you can ask questions and follow up questions as well.
    </div>

    ![ask_paragon_prompt](./images/ask_paragon_prompt.png)

## Prompt Engineering Zero Shot Prompting

We provide no information to the LLM, just a straight question.

Start with a simple question without any context with basic instructions. Please note answers may be different for the same question, this is the behavior of LLM. You can type the questions inside the window as shown and press enter.

1.  Ask, List the devices in the network.

    ![ask_paragon_question1](./images/ask_paragon_question1.png)

2.  Review the Answer and the question is also displayed under the Message section. Note down one of the device names will be handy for next set of questions.

    ![ask_paragon_question1_answer](./images/ask_paragon_question1_answer.png)

3.  Let’s get to the next question as to what the solution can do (as of today). ASk, What can you help me with today?

    ![ask_paragon_question2](./images/ask_paragon_question2.png)

## Prompt Engineering One Shot Prompting

We provide some information to the LLM to help get a better answer. Be specific, like mention the name of the device, if you don’t remember the device names ask the question again -> list devices in this network. From that copy the name of one of the devices.

1.  ASk, There were some issues listed in the device *device name* (Put the name from the device earlier listed in Task 2), can you explain the issues.

    ![ask_paragon_question3](./images/ask_paragon_question3.png)

## Prompt Engineering Chain of Thought Prompting

Guide the LLM through a logical reasoning process and ask a question. Please note there can be LLM hallucinations as this is a complex question and you may not receive a satisfactory answer.  In that case break down your question and make sure not to close the window so that there is context awareness.

1.  Ask, Imagine you are the network operations manager of a large enterprise; your priority is to isolate critical issues in your network today and fix them. Can you list the critical issues on the device *device name* and explain how to fix them.

    ![ask_paragon_question4](./images/ask_paragon_question4.png)

    ![ask_paragon_question4_answer](./images/ask_paragon_question4_answer.png)

## Language Personalization

Ask a simple question in the language of your choice (subject to keyboard or you can use google translate and then copy paste)

1.  Ask in English, Explain the issue PEM offline from a Junos perspective?

2.  Ask in Simplified Chinese, 从junos角度解释PEM离线问题?

    ![ask_paragon_question6_answer](./images/ask_paragon_question6_answer.png)

## Ask Anything

The Floor is yours!!! Ask anything, but DON’T not do WRITE operations please, as it may impact the lab itself.