ReAct AI Agent from Scratch using DeepSeek: Handling Memory & Tools without Frameworks

ReAct AI Agent from Scratch using DeepSeek 🐳

This repository contains a ReAct (Reasoning + Acting) AI agent built from scratch using Python and DeepSeek.

For an in-depth explanation of this agent implementation read my medium article:

ReAct AI Agent from Scratch using DeepSeek: Handling Memory & Tools without Frameworks

📌 Features

✅ Reasoning Loop built from scratch with prompt-based control
✅ Memory Management using 🤗Transformers library to count tokens in formatted messages
✅ Tool integration with a scalable tool environment. Not using function-calling API features
✅ Streamlit UI to interact with the agent and display its responses and reasoning process
✅ FastAPI backend for handling chat requests from Streamlit

🔧 How to Set Up

1️⃣ Clone the repository

git clone https://github.com/Wencho8/ReAct-AI-Agent-from-Scratch-using-DeepSeek.git
cd ReAct-AI-Agent-from-Scratch-using-DeepSeek

2️⃣ Set up environment variables

Copy the example environment file and configure your API keys:

cp .env.example .env

3️⃣ Install dependencies

pip install -r requirements.txt

🚀 How to Run

🖥️ Running from the Terminal

If you'd like to run the agent directly from the terminal without using the UI:

cd Agent
python agent.py

This will start the agent, allowing you to interact with it via the command line.

🌐 Using the UI

1️⃣ Start the FastAPI Backend

cd Agent
python -m uvicorn app:app --reload

2️⃣ Start the Streamlit UI

python -m streamlit run chat_ui.py