AI Agent with Gemini Function Calling

Unlocking Gemini Power by Tool Box

Said Rasidin
8 min readMay 18, 2024
Photo by Jeremy Straub on Unsplash

This AI race reminds me of spaceships in sci-fi movies that can operate themselves and talk to humans. For example, if there is an oxygen leak, they start a self-healing process or locate a nearby habitable planet for humans

Today, we can interact with LLMs(Large Language Models) that are trained using (almost) all human literature ever written over centuries.

Introduction

In this blog, I want to share my learning process about creating an AI Agent powered by Gemini and function calling, it might be a little late as it was already hyped last year, but better late than never ya.

LLM is a large language model that can understand natural language just like humans, training with massive datasets on the internet. It works by predicting the next world from the previous word, using probability and attention mechanisms (only focusing on important things)

However, LLMs don’t think like humans often just generate the most make-sense sentences. To tackle this, different techniques are applied from Retrieval-augmented generation (RAG) which feeds the model with relevant information before answering, and also grounding it with search engine

--

--