Taking Action with Tools and Function Calling
Explore function calling, the technique that transforms an LLM from a passive knower into an active doer by allowing it to use external tools
In our last lesson, we built a powerful RAG pipeline. We successfully provided our model with an “open book” to read from, enabling it to answer questions about specific, static documents. It is now a brilliant expert researcher.
But a researcher’s knowledge is based on the books in their library. What happens when a user asks a question that can’t be answered by a static document?
“What’s the current weather in London?”
“What’s the price of a ticket from NYC to LA next Tuesday?”
“What is the status of my recent order?”
The RAG pipeline is useless in such scenarios. The model’s internal knowledge is outdated or irrelevant. It needs the ability to fetch live information and interact with the outside world. How do we give our LLM “hands”?
From knower to doer: Function calling
The solution is a technique called function calling (also known as “Tool Use”). This is the mechanism that allows us to break the LLM out of its digital confinement and let it interact with our own code. It enables the model to:
Access real-time information: Connect to live APIs for weather, stock prices, or flight information.
Interact with private systems: Query internal company databases or execute actions (like placing an order) on proprietary systems.
Perform complex calculations: Offload precise mathematical calculations to a dedicated calculator tool, overcoming an LLM’s inherent weakness in this area. ...