Dave Hulbert's Today I Learned (TIL)

Making API Specifications with Wiretap and AI

I work with APIs most days and occasionally come across one without an accurate API specification. This happened recently and I thought it would be worth creating a spec before starting to use the API: that way I could make use of existing tools and libraries, rather than manually piecing together curl commands.

My journey involved using Wiretap, an OpenAPI spec tool, in conjunction with ChatGPT and Vacuum.

OpenAPI Specs

OpenAPI specifications are the backbone of clear, consistent API development. They serve as a contract between the API and its consumers, ensuring reliability and consistency. Benefits include automated client generation, AI integration, maintaining backward compatibility, and providing clear documentation for users. Following the Open API spec also allows you to make use of all sorts of existing tools, which is what I was after.

Starting with ChatGPT

Initially, I used ChatGPT to generate a basic OpenAPI spec from a handful of sample API responses. ChatGPT was great at this, writing all the standard boiler plate. After tweaking some assumptions that ChatGPT made and adding necessary details, I needed a robust way to validate and refine this spec. That's where Wiretap came into the picture.

Setting Up Wiretap

Installing Wiretap was straightforward:

brew install pb33f/taps/wiretap

Running Wiretap was just as simple:

wiretap -s path/to/spec.yaml -u https://api.example.com

This set up a local monitor at http://localhost:9091/ and a proxy at http://localhost:9090 for API requests.

Then, all you need to do is fire requests to the localhost proxy and watch Wiretap report on any issues.

Discovering and Fixing Issues with Wiretap and Vacuum

Wiretap proved invaluable by identifying things like an invalid response type and optional parameters in my spec. This prompted me to validate the spec further using the Vacuum linter, another useful tool in the API development toolkit.

Once the spec was in place, it then only takes seconds to paste it into Swagger Editor (or a tool like Postman) and get a working API client, making it easy to explore the API. No more messing around with curl and manually escaped JSON!

The ease and efficiency of using tools like Wiretap and AI for documentation are game-changing - gone are the days of manual labor-intensive documentation.