Generative AI: Getting Started

Pre-requisites

• A computer
• Open AI API Key
• Python 3+ installed
• Pip3 (usually included with Python install)
• Git
• Clone this repository
• (Optional, for Lesson 2 samples) Install Ollama to run Llama and some other models locally: https://ollama.com/
  1. In Terminal: ollama pull llama3.2
  2. For vision capabilities: ollama pull llama3.2-vision

Recommended

• Visual Studio Code IDE

Setup

Clone the repo

1. Create a folder somewhere on your computer, e.g. ai-tutorial-py
2. Open the Terminal app (or console on Windows) and use the cd command to navigate to the folder you created
3. git clone https://github.com/bowdoincollege/ai-tutorial-py.git

Managing dependencies in Python

Make sure you are in the ai-tutorial-py directory in the Terminal app.

python3 -m venv venv

source venv/bin/activate

python3 -m pip install -r requirements.txt

• Copy the .env.example file to .env and fill out the environment variables

Lesson 1.1: Getting Started!

Making the very basic 'hello world' call to the OpenAI API.

lesson-1-1.py

Lesson 1.2: Chat

Accepting basic user input.

lesson-1-2.py

Lesson 1.3: Chat memory

Adding context so that the chat bot remembers the conversation.

lesson-1-3.py

Lesson 1.4: Visualizations

Taking advantage of Markdown syntax to improve the output formatting

lesson-1-4.py

Lesson 1.5: Summarization

Taking things up a notch. Adding code to scrape the web and generating a summary of text that exceeds the context window of the LLM (128,000 tokens for OpenAI models). This is often referred to as RAG: Retrieval-Augmented Generation, supplementing data from an external source from the LLM training data.

lesson-1-5.py (US House docs)

lesson-1-5b.py (Supreme Court docs)

Lesson 1.6: Vision

Getting started with OpenAI vision capabilities.

lesson-1-6.py

Lesson 1.6: Generated HTML

Creating a dynamic web application

lesson-1-7.py

Lesson 2: Running local LLMs with Ollama

This sample code shows you how to run an llm locally with Ollama, e.g. llama 3.2, etc. These take up a lot of hard drive space and can be slower than using the commercial APIs, but it's free!

References

• Python Rich Console library: https://realpython.com/python-rich-package/
• OpenAI quickstart: https://platform.openai.com/docs/quickstart?language-preference=python&quickstart-example=completions
• LangChain summarization: https://python.langchain.com/docs/tutorials/summarization/#map-reduce
• Ollama quickstart: https://github.com/ollama/ollama/blob/main/README.md#quickstart
• LangChain Python Ollama: https://python.langchain.com/docs/integrations/llms/ollama/
• OpenAI Python library: https://github.com/openai/openai-python