Step-by-Step Guide to Building Domain-Specific AI Agents with Phidata SDK

Simplified solution for AI Agent development: Phidata SDK

Nowadays, AI Agent has attracted much attention, and many companies are competing to build their own Agent. While some companies choose to build from scratch, the biggest headache for developers is often the time required to build and test different AI tools (for function calls), which can lead to extended development cycles.

However, over time, new solutions are emerging that help developers simplify the process of building AI Agents. Phidata is one of the solutions.

What is Phidata?

Phidata SDK makes it easy to build AI Agent with just a few lines of Python code. Best of all, it provides many useful AI tools directly, without having to write them from scratch.

Some of the out-of-the-box AI tools provided by Phidata:

• Google Search, Exa (for web search)
• Resend (for sending email)
• Crawl4AI and Firecrawl (for web crawling)
• DuckDB (for data analysis)
• Python Agent (for writing and running Python code)
• Files (used to read files to build RAG Agent)
• GitHub (for interacting with GitHub)
• CalCom Agent (for scheduling meetings using Cal.com)

And more! You can view the full list of tools provided by Phidata here.

Using the above tools or their combination, we can build very complex and exciting AI Agents, such as:

• Data Analysis Agent
• Research Agent
• Sales Agent
• Shopping Agent

Just to name a few.

Use "Team Agent" to create advanced agents

A cool feature of the

Phidata SDK is that you can combine multiple tools into a team, called a "Team Agent". For example, you could create a team that includes the tools "DuckDuckGo" and "Yahoo Finance" that pulls data from two sources, Web and Yahoo Finance.

This team’s Agent works as follows:

• The user issues the following instructions to the Agent: "Summary analysts' recommendations and share the latest news from NVDA"
• Our Agent searches the web for the latest news about NVIDIA
• The agent also searched for the stock symbol "NVDA" on Yahoo Finance to obtain the company's financial data
• Finally, our Agent presents data from two sources (Web and Yahoo Finance) in a beautiful table.

Cool, right?

Now that we have understood the functions that can be achieved using the Phidata SDK, let us continue to see how to use it to create a simple financial analysis Agent.

Get started with Phidata SDK

Let’s start building our financial analysis Agent. It'll be perfect for beginners, so don't worry about keeping up.

We will create our Agent in a Jupyter Notebook on Google Colab which is very interactive and easy to share.

Step 1 - Create a new Notebook on Google Colab

Click here to enter Google Colab, you should see the following interface:

Now click the "New Notebook" button:

It may take a while to load, after which you should be in your newly created notebook, which will look like this:

Great, let’s move on to the next step.

Step 2 - Install all required libraries in Notebook

Before we start creating the AI ​​Agent, we need to ensure that we have the necessary dependencies in the Notebook. Please note that Google Colab pre-installs some commonly used libraries in Notebook, but to ensure that we have all the libraries we need, we will still install all libraries.

We will install the following libraries:

• yfinance - Get a company’s financial data
• openai - Allows Phidata to use OpenAI's LLM to run AI Agents
• duckduckgo-search - Search the web with DuckDuckGo
• Phidata - Loads pre-written AI tools to make function calls and create team agents.

To install these libraries, copy and paste the command below into the first block of cells:

<code>pip install openai yfinance duckduckgo-search phidata</code>

It should look like this:

Next, click on the play icon on the left as shown below:

Now let it install all dependencies for a while. Once the installation is complete, you should see a small green checkmark to the left of the Run button, like this:

Let’s hide the cell’s output since it takes up too much space in the Notebook. Click the button below the Run button and then click Show/Hide Output.

Step 3 - Add OPENAI_API_KEY environment variable

Now we need to add the OpenAI API key to our environment. Continue adding a new cell to the Notebook by clicking the button shown below:

Now paste the following code in this new cell and run it. Replace the your_api_key value with the actual OpenAI API key you obtained from https://www.php.cn/link/9e4aef142346875a7f13f4a42526a69f.

<code>pip install openai yfinance duckduckgo-search phidata</code>

It should look like this:

Step 4 - Write code for Agent

In the last step, we will write the actual code for the Agent. Since this is an "Agent Team" (which means it is an AI Agent composed of multiple Agents), we will first create two Agents using the Phidata SDK, namely web_agent and finance_agent. The web agent will search the web for news about the company, while the finance agent will search Yahoo Finance for the company's financial data. Finally, we will create a third Agent by passing these two Agents to this Agent's "teams" array, which will result in the creation of the "Agent Team". This third Agent will be the one we ultimately use to obtain company data from the web and Yahoo Finance.

<code>import os

os.environ['OPENAI_API_KEY'] = "your_api_key"</code>

Add a new cell to your Notebook. Then copy the above code and paste it into a new cell.

That’s it! Continue running the cell. It will take some time to finish running, once it does you will be able to see the output like this (you need to scroll to the bottom):

The output text looks small because I've scaled it down to fit the entire output into a single screenshot.

Thus, we were able to build this financial analysis Agent in a very short time. Obviously the report is a bit basic and could be a bit more detailed, but we can always improve our Agents by adding more data from different sources by adding new Agents to the team (or building our own functional tools from scratch).

You can follow my account on LinkedIn to learn more about AI Agent!

The above is the detailed content of Step-by-Step Guide to Building Domain-Specific AI Agents with Phidata SDK. For more information, please follow other related articles on the PHP Chinese website!