Gemini Workshop

This repository contains the code for a workshop on Generative AI using Gemini.

Setup

1. Install Python: Ensure you have Python 3.10 or later installed.
2. Install Dependencies: Install the required packages using pip, or poetry. You can find the required packages in the pyproject.toml file.
3. Set up Vertex AI and Google Cloud Credentials:
   • Create a new Google Cloud Project following the instructions.
   • Enable the Vertex AI API.
   • Create a service account with the necessary permissions.
   • Download the service account key file.
   • Set the GOOGLE_APPLICATION_CREDENTIALS environment variable to the path of the key file.
4. Set up AI-Studio API Key:
   • Obtain a Google Cloud API key from the Google AI-Studio.
   • Set the GOOGLE_API_KEY environment variable to the API key.

Code Structure

The repository is organized as follows:

• utils.py: Contains utility functions for the demos and examples. Such as:
  • Downloading files.
  • Normalizing URLs.
  • Getting images from URLs or paths.
  • Getting image bytes in base64 encoding.
  • Plotting embeddings.
• ML-Foundations:
  • This folder contains the code for the ML Foundations section of the workshop.
• Embeddings and VAE demos:
  • embeddings.py: Generates embeddings for images and their captions.
  • vae.py: Implements a Variational Autoencoder (VAE) for dimensionality reduction.
  • embeddings_vertexai.py: Client for getting multimodal embeddings using Vertex AI.
  • embeddings_image_bind.py: Client for getting multimodal embeddings using ImageBind.
• GeminiAPI usage Examples:
  • reasoning_engine.py: Demonstrates using Gemini as a reasoning engine with Vertex AI.
  • standalone.py: Simple example of using Gemini standalone for text generation.
  • getting_started.py: Basic introduction to using Gemini.
  • function_calling.py: Examples of using function calling with Gemini.
  • context.py: Demonstrates using Gemini with memory and retrieval-augmented generation (RAG).
  • callbacks.py: Shows how to use callbacks with Gemini.
  • multimodality.py: Examples of using Gemini with multimodal inputs (audio, image, video).
• Applications Examples:
  • LLM_app.py: Creates a simple chat app using Gemini and Vertex AI for image generation.
  • RAG_app.py: Creates a RAG app using Gemini and Vertex AI for document retrieval and question answering.