Creating Better SDKs with Generative AI

Course details

Programmers use software development kits (SDKs) to leverage their web application programming interfaces (APIs) across dozens of projects while a much smaller core group creates those SDKs. Despite often being an afterthought, SDKs serve as the front door for many APIs giving developers their first impression and understanding of a product and can make or break a project. Instead of treating SDKs as a chore, this course explores how to use generative AI tools to create better, more useful, and well-structured SDKs for your API. Get hands-on experience as instructor Keith Casey lays out the basics of multiple generative AI models and shows you how to use them, how to evaluate their output, and how to incorporate AI into your workflow.

Learning objectives
Explore how to map programming structures to an SDK project.
Set up code generation for small pieces that link together into larger coherent structures.
Convert abstract specifications to concrete code using GitHub Copilot and Claude from Anthropic.
Create and tune LLM models specifically for SDK projects.

Concepts

Introduction

From theory to practice - Building SDKs with generative AI
What you should know

Getting Started with SDKs

What is an SDK
What makes a great SDK - Succinct, purposeful, open source
What makes a great SDK - Idiomatic and logical
Examples of great SDKs
Risks of using SDKs
Challenges of building SDKs

Generating Your SDK

What is code generation
Where does code generation fail
OpenAPI 101
Setting up your build environment
Generating your first SDK
Validating your SDK
Reviewing your SDK
Improving your OpenAPI spec
Challenge - Improving your OpenAPI spec
Solution - Improving your OpenAPI spec

Generating Your SDK with Copilot

What is Copilot
Generating your SDK with Copilot
Validating your Copilot-generated SDK
Reviewing your Copilot-generated SDK
Challenge - Using Copilot on your OpenAPI spec
Solution - Using Copilot on your OpenAPI spec

Generating Your SDK with a Custom Model

What is an LLM
How do we choose code for our model
Generating your SDK with an LLM
Validating your generated SDK
Reviewing your generated SDK
Challenge - Using an LLM to improve your LLM-generated SDK
Solution - Using an LLM to improve your LLM-generated SDK

Improving Your SDK Generation

Tuning your LLM for better code
Building a feedback loop for code quality
Model tuning - The good, the bad, and the ugly
Prompt tuning - Tell me what you want

Next Steps

Choosing your licenses
How do we handle support and distribution
Improving your OpenAPI spec
Learn more