Agentic RAG/

...

/

The AI Research Assistant: Implementing the Agent’s Core Tools

The AI Research Assistant: Implementing the Agent’s Core Tools

Learn to translate the agent’s design into functional, independently tested Python code for each core tool (RAG, arXiv search, and definition lookup).

We'll cover the following...

In our last lesson, we put on our architect hats and drew the complete blueprint for our AI research assistant. We have a clear mission, a defined set of requirements, and a plan for the tools our agent will need.

Now, with our blueprint in hand, we step into the workshop. This lesson marks the beginning of our three-part implementation phase. We will translate our design into functional, working code. Our focus in this lesson is on building the foundational components, the individual tools our agent will use, and testing each one in isolation to ensure it is reliable before we hand it over to our agent.

A well-defined project structure is the cornerstone of professional software development. Before writing our implementation logic, we will establish a modular directory and file structure. This practice is essential for maintaining clean code, enabling independent testing of components (unit testing), and ensuring our application is scalable and easy to maintain as it grows in complexity.

The project file structure

We will not put all our code in one giant file. Instead, we will organize our tools into their own dedicated files. Create the following folder and file structure in your project directory:

/ai-research-assistant
|-- /data
|   |-- paper1.pdf
|   |-- paper2.pdf
|-- /tools
|   |-- __init__.py
|   |-- rag_tool.py
|   |-- api_tools.py
|-- main.py
|-- .env
|-- requirements.txt

  • data/: This folder will hold the PDF research papers for our local RAG tool.

  • tools/: This Python package will contain our modular tool implementations.

  • main.py: This will be our main application file where we assemble and run the agent in the next lesson.

Environment and dependencies

Next, let’s ensure our environment has all the necessary libraries. Create a file named requirements.txt and add the following dependencies:

llama-index
llama-index-llms-groq
llama-index-embeddings-ollama
python-dotenv
arxiv
wikipedia

Install these by running the following command in your terminal: pip install -r requirements.txt.

Finally, create a .env file to securely store your Groq API key:

GROQ_API_KEY="your-groq-api-key-here"

Reminder: Before proceeding, ensure your local Ollama instance is running and you have an embedding model like nomic-embed-text available.

Note for learners: These instructions are meant for local setup. You don’t need to perform these installations here; everything has already been set up for you in this environment.

We will now build our first and most critical component: the RAG tool. This module will be responsible for all interactions with our local document knowledge base.

Building tool 1: The local RAG engine

Our objective in this module is to create a self-contained, modular function that builds our entire RAG pipeline and wraps it as a LlamaIndex Tool ...