🧠 AI Agents Playground

This project explores how AI agents can be implemented using three different frameworks. Each framework has its own folder with independent logic and tools, allowing for clear comparisons and modular experimentation.

📂 Project Structure

.
├── smolagents/     # Agents built with Smol Agents framework
├── llamaindex/     # Agents built with LlamaIndex for document interaction
├── langgraph/      # Agents built with LangGraph's node/edge system
├── requirements.txt  # dependencies
└── README.md       

🧠 LLMs Used

This project uses the following local LLMs served through Ollama:

1. Qwen2.5 7B – used in Smol Agents
2. LLaMA 3.1 8B – used in LlamaIndex-based agents
3. Mistral – used in LangGraph-based email workflow

These models are configured via wrappers like LiteLLMModel, Ollama, or ChatOllama depending on the framework.

🚀 Overview

This playground serves as a learning and experimentation space to:

• Compare agent frameworks
• Understand how tools, memory, context, and flow are handled
• Build reusable patterns for future projects

🔧 Frameworks Used

1. Smol Agents (smolagents/)

• Lightweight, tool-based agent system
• Focuses on modularity and simplicity

2. LlamaIndex (llamaindex/)

• Index and query over documents
• Useful for RAG and data-aware agents

3. LangGraph (langgraph/)

• Graph-based agent flow with custom edges
• Great for visualizing and managing complex agent logic

📦 Setup Instructions

Create a Virtual Environment

python -m venv .venv

Activate the Environment

Windows:

.venv\Scripts\activate

macOS/Linux:

source .venv/bin/activate

Install Requirements

pip install -r requirements.txt

Run Ollama Server (keep this running in another terminal)

ollama serve

✨ Goals & Features

• Unified design for testing and comparing agents
• Simple folder-based separation
• Extendable and beginner-friendly codebase

📝 Notes

• More examples and benchmarking might be added later
• Each folder is standalone and self-documented

---

🙏 Acknowledgments

A big thank you to Hugging Face for their incredible AI Agents course, which served as the inspiration and learning foundation for this project.

The concepts, workflows, and code explored here were directly influenced by the hands-on lessons and guidance provided throughout the course. Massive respect to the instructors and contributors for making advanced agent tooling approachable and practical.

---

Check out the individual README.md in each folder for detailed documentation and usage examples.