Gemini Pro Quick Start Examples in Python

• Conda 2/3
• Python 3.11

Environment Setup

conda create --name google -c conda-forge python=3.11
source activate google

Inside google conda virtual environment, use pip3 to install all packages we want,

pip3 install google-generativeai streamlit python-dotenv

# Optional. Disable stats collection if you want.

cat <<EOT > $HOME/.streamlit/config.toml
[browser]
gatherUsageStats = false
EOT

How to Run

Setup local .env file first. We need the following key-value pairs setup. See Quick Start on Your Google Account and Setup ADC or API Keys to ramp up some default credential settings to use Google Cloud APIs, this is required if you haven't done so.

Note: for GOOGLE_API_KEY, go to UI https://console.cloud.google.com/.

cat <<EOT > ./.env
GOOGLE_API_KEY="REPLACE_WITH_YOUR_OWN_KEY"
PROJECT_ID="REPLACE_WITH_YOUR_PROJECT_ID"
LOCATION="us-west1"
GOOGLE_APPLICATION_CREDENTIALS="/<WHERE_YOU_STORE_YOUR_GOOGLE_ADC_KEY>/<YOUR_PROJECT_ID_xxxxxx>.json"
EOT

# Gemini Pro Model via generativeai APIs
streamlist run gemini.py

# Gemini 1.5 Pro Model via VertexAI APIs
streamlist run gemini_pro.py

Appendix

Quick Start on Your Google Account and Setup ADC or API Keys

As first time Google Cloud user, see Google official doc. For quick and lazy setup, here are the TL;DR version.

Note: You still need to enable Vertex AI via your Google Console and accept the terms and services, etc. in your Google account.

simply run the script, it will try to fetch your default project, create an IAM, assign some roles, and generate the key JSON files for GOOGLE_APPLICATION_CREDENTIALS. You can also do it on UI https://console.cloud.google.com/.

./setup_gcloud.sh