Excited to share this Conversational Weather Agent with LangChain Tool Calling project I worked on. I implemented tool/function calling through LangChain agents and built a conversational weather interface using the OpenWeather API.
This AI assistant answers your natural language questions about the current weather and 5-day forecast for any city or zip code. Ask it anything - “Is it raining in London today?” or “What’s the forecast for Paris tomorrow?” - and get the info you need in a chat-like format.
This project combines LangChain’s agent capabilities with a powerful LLM for a user-friendly weather experience.
I have just defined some functions with a well-defined DocString and a proper prompt for the LLM. LLM automatically decides if it’s a forecast or a if it’s for a current weather(since they are two different APIs) and calls the function automatically with the required parameters. It also asks for the required details when it’s not given. 𝙁𝙪𝙣𝙘𝙩𝙞𝙤𝙣 𝙨𝙘𝙧𝙚𝙚𝙣𝙨𝙝𝙤𝙩𝙨 attached or visit the github
One another jaw-dropping phenomenon was that it calls a function which I defined to take the local timestamp only when forecast is asked for. It decides that by itself with just the docstring. How cool is that?
This is one of the amazing achievements that have happened recently! Just with a function and a natural language prompt we can make LLM to call the function by itself to do an operation. This saves so much effort in time and man power. !
Still Amazed !