🧠 CoT Reasoning Agent

A structured Chain-of-Thought (CoT) reasoning agent built using LLMs with step-by-step planning, tool calling, and JSON-based structured outputs.

---

🚀 Project Overview

The agent:

• ✅ Breaks down user queries step-by-step (START → PLAN → TOOL → OUTPUT)
• ✅ Uses structured JSON responses
• ✅ Calls external tools when required
• ✅ Retrieves real-time weather information
• ✅ Demonstrates tool-augmented reasoning

It acts as both:

• 🧠 A Chain-of-Thought reasoning agent
• 🌦️ A Current Weather Retrieval Agent

---

🚀 Features

• Structured reasoning using START, PLAN, TOOL, and OUTPUT steps
• Tool-calling capability
• Real-time weather retrieval via wttr.in
• JSON schema validation using Pydantic
• Secure API key management using .env

---

📂 Project Structure

cot_reasoning_agent/ │ ├── aiAgent_pydantic.py ├── requirements.txt └── README.md

---

⚙️ Setup Instructions

1️⃣ Clone the repository

git clone https://github.com/your-username/cot-weather-agent.git

---

2️⃣ Create Virtual Environment

python -m venv .venv

Activate Environment

Windows (PowerShell):

.venv\Scripts\activate

Mac/Linux:

source .venv/bin/activate

---

3️⃣ Install Dependencies

pip install -r requirements.txt

---

4️⃣ Add Your API Key (Important 🔐)

Create a file named:

.env

Add your Google API key:

GOOGLE_API_KEY=your_api_key_here

---

▶️ Run the Agent

python cot_weather_agent.py

---

Example Queries

👉 What is the weather in Delhi? 👉 Solve 45*12/3 👉 What is the weather in Bangalore?

---

🧠 How It Works

1. The user enters a query.
2. The model reasons step-by-step using Chain-of-Thought.
3. If required, it calls the get_weather(city) tool.
4. The tool response is injected back into the reasoning loop.
5. The agent generates the final structured OUTPUT.

---

🔧 Tool Used

get_weather(city: str)

Fetches current weather using:

https://wttr.in/{city}?format=%C+%t

Returns formatted weather information for the given city.

---

🎯 Learning Objectives

This project demonstrates:

• How Chain-of-Thought improves reasoning transparency
• How to build tool-augmented LLM agents
• How to structure LLM outputs using JSON schemas
• How to securely manage API keys