Skip to content

shaikmaricar/AgenticAI

Repository files navigation

AgenticAI

A hands-on project to build your first AI agent using LangChain, Claude Sonnet 4, and Open-Meteo — with every decision explained step by step.

What This Agent Does

  • Answers weather questions in natural language ("What's the weather in Chennai?")
  • Fetches live weather data from Open-Meteo API (free, no API key needed)
  • Remembers conversation context ("Which city was hotter?" works across turns)
  • Validates input with Pydantic (format) + LLM (is it a real city?)
  • Personalizes responses per user (Celsius vs. Fahrenheit) via ToolRuntime

Quick Start

# Clone
git clone https://github.com/shaikmaricar/AgenticAI.git
cd AgenticAI

# Install dependencies
uv sync

# Set your Anthropic API key
export ANTHROPIC_API_KEY="your-key-here"          # Linux/Mac
$env:ANTHROPIC_API_KEY = "your-key-here"           # Windows PowerShell

# Run
uv run python main.py

Sample Session

Weather Agent (type 'quit' to exit)
==================================================

You: What's the weather in Chennai?
Agent: The current weather in Chennai is 27.3°C, Partly cloudy,
       Humidity: 90%, Wind: 4.2 km/h

You: How about Tokyo?
Agent: The current weather in Tokyo is 13.3°C, Mainly clear,
       Humidity: 94%, Wind: 2.3 km/h

You: Which city is hotter?
Agent: Chennai is hotter at 27.3°C compared to Tokyo's 13.3°C —
       about 14°C warmer!

Tech Stack

Library Purpose
LangChain Agent framework
LangGraph Agent runtime (ReAct loop, state, checkpointing)
Claude Sonnet 4 LLM brain
Open-Meteo Free weather API (no key required)
Pydantic Input validation
httpx HTTP client

Key Concepts Covered

User Input
  │
  ├─ Pydantic Validation (format: length, type)
  │
  ├─ LLM Validation (semantic: is it a real city?)
  │
  ├─ ReAct Loop (reason → act → observe → respond)
  │     │
  │     ├─ Tools with ToolRuntime (per-user context)
  │     ├─ recursion_limit (infinite loop safety)
  │     └─ InMemorySaver (conversation memory)
  │
  └─ Response to User

Detailed Article

For a complete walkthrough explaining why each piece exists, every gotcha we hit, and when to use one approach over another:

How to Create Your First AI Agent? An Easy to Understand Detailed Article

Prerequisites

License

MIT

About

Agentic AI resources

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages