this is amazing. Now we are going to integrate llama 3 in autogen. To do this, we are going to use Gro API. You can also use OL if you want to run this locally on your computer. We are also going to evaluate how the model is capable of converting a text to SQL. I'm going to teach you how to do that. In this, we are going to see how to integrate llama 3 in autogen using Gro and also using OL armor, which means it's completely private and local to your computer. So first, we'll be configuring. Next, we'll be importing data. Third, we are going to create agents to evaluate the model. Then we are going to create custom tools to execute the SQL. And finally, make them work together. That's exactly what we're going to see today. Let's get [Music] started. Hi everyone. I'm really excited to show you about autogen integration with llama 3. I'm going to take you through step by step on how to do this. But before that, I regularly create

videos in regards to Artificial Intelligence on my YouTube channel. So, do subscribe and click the Bell icon to stay tuned. Make sure you click the like button so this video can be helpful for many others. Like here, first step pip install Pi autogen spider hyphen EnV, and then click enter. Spyer EnV is the package which we are going to use to download SQL schema to evaluate the model. Next, let's export Gro API key like this, and then click enter. Now, let's create a file called app.py and let's open it. So, these are the steps which we are going to follow. First is to configure in your file. First step configuration in that we're going to import Json, OS, annotated, and dict spider EnV conversible agent user proxy agents, and config list from Json. Next, we are going to assign autogen user Docker equals false. If you have Docker, you can mention that as true. Next, we are going to define the llm config. This is where we are going to integrate our llama 3 70 billion parameter model, and we going to

use Grok to do that. And here is the end point. The second step is to import data. We're going to use Spider Env. This will automatically load the question or the instruction and the schema for the database. So, with this, we are able to understand if the LUN language model is able to convert a text that is the instruction to SQL query based on the SQL schema. This is a Spider Env package. It's to evaluate text to SQL. And you should be able to import lots of challenges to understand this Spider Env. We are going to run this code in your terminal: `python app.py` and then click enter. And you can see the schema or the database gets loaded. Here you can see the schema of the database that is: create table artist volume music festival. And this is the question which we are going to ask: find the famous titles of artists that do not have any volume. Based on this question and this schema, the autogen agent or the L language model behind the autogen agent should give a SQL query. Now we are going

to extend this code, so step number three is to create agents. We're going to create two agents: one is the SQL writer agent, another one is a user proxy agent. But before that, we are going to write a function to check if there is any function calling happening or tool call. So the way we interact or integrate our evaluation is using tool call. So we are checking if tool call is happening. So first, SQL writer. Here we are providing a normal system message. You are good at writing SQL queries. Always respond with function call to execute SQL. Now, next, the user proxy agent. Human input mode: never, and maximum consecutive auto-reply is five. Now, the fourth step is to create tools or function calling. It is a custom tool to execute the SQL query generated by the large language model or the agent, also called as SQL writer. Then we are going to run that SQL query and check if it is an error or correct one. So the way we add the function calling is by using a decorator. This is a much more easier version to implement a function calling. We're going to register for LLM.

function to SQL writer so that the SQL writer will know there is a function which exists to execute SQL query. Next, we're going to add a user proxy decorator with register for execution function. Next, we're going to create the function. It's execute SQL function here using gym. Step, we are going to run the SQL query generated by the SQL writer agents, and if any error occurs, we are going to return accordingly. Next, step number five, we're going to initiate the chant. We're going to define a prompt template where we are going to provide the schema and the question. So, this is the schema which you saw in the terminal before, and this is the question. So, that will go here. And finally, user proxy do initiate chat SQL writer, and the message is the prompt template. That's it. Now we have completed the code. As a quick summary, we defined the configuration. We integrated Llama 370 billion parameter model using Gro. Next, we imported the data using Spyder ENV package. Third, we created agents SQL writer and user proxy agents. Then we created a custom tool to execute SQL.

added that to SQL writer and user proxy agent. Finally, we are initiating the chat. Now I'm going to run this curve in your terminal Python app and then click enter. Now you can see the user proxy agent is telling the writer, Below is a schema for SQL database, and then it's providing the question, Here generate SQL query to answer the following question: find the famous titles of the artist that do not have any volume. So the SQL writer is going to use a tool called function calling. Here the execute SQL function is being used. Based on the question, it is providing SQL query, and this SQL query is run by the execute SQL function, and finally, the response is being returned. And here is the response, and that is correct. Now we have successfully tested Llama 70 billion parameter model. Now we have successfully tested Llama 370 billion parameter model of its ability to convert from text to SQL query. But this is just one test. You can even loop multiple tests using this Spyre EnV package. This is just a beginner tutorial, so I'm going to leave.

here now you have learned how to integrate llama 3 in autogen. How to integrate Gro in autogen. How to create a custom function using decorator and how to evaluate an agent for a text to SQL query conversion. You can even integrate this with ol Lama. To do that in your terminal, make sure you pull llama 3 from oama pull llama 3 and then click enter considering you've already installed olama from ama.com website. Next, export open model name equals Lama 3 export open a API base equals Local Host 11434 4/ V1 and then click enter. That's it. Just these two changes. Now I'm going to run this code python app. pi and then click enter and you can see it's working as before and SQL writer is giving the SQL query but it's not initiating the function call. That's because llama 7 billion parameter model is comparatively smaller than llama 70 billion parameter model and you can see it's not able to perform the task. This is because llama 3

smaller version is not as powerful in regards to function. C calling so I'm not able to get a correct answer. You might need to try much more advanced version or a model with higher parameter to get better respond. I'm really excited about this. I'm going to create more videos similar to this so stay tuned. I hope you like this video. Do like, share, and subscribe, and thanks for watching.

AutoGen Llama 3: How to Convert Text to SQL Query using AI Agents?

AutoGen Llama 3: How to Convert Text to SQL Query using AI Agents?