speakeasy VS taxilang

I’ve built conviction that code generation only gets useful in the long term when it is entirely deterministic, or filtered through humans. Otherwise it is almost always technical debt. Hence LLM code generation products are a cool toy, but no sensible teams will use them without an amazing “Day 2” workflow.

As an example, in my day job (https://speakeasyapi.dev), we sell code generation products using the OpenAPI specification to generate downstream artefacts (language SDKs, terraform providers, markdown documentation). The determinism makes it useful — API updates propagate continuously from server code, to specifications, then to the SDKs / providers / docs site. There are no breaking changes because the pipeline is deterministic and humans are in control of the API at the start. The code generation itself is just a means to an end : removing boilerplate effort and language differences by driving it from a source of truth (server api routes/types). Continuously generated, it is not debt.

We’ve put a lot of effort into trying to make an LLM agent useful in this context. However giving them control of generated code directly means it’s hard to keep the “no breaking changes”, and “consistency” restrictions that’s needed to make code generation useful.

The trick we’ve landed on to get utility out of an LLM in a code generation task, is to restrict it to manipulating a strictly typed interface document, such that it can only do non-breaking things to code (e.g. adjust comments / descriptions / examples) by making changes through this interface.

In my mind the analagous behaviour would be if the golang checksum database added in license terms that stated "you need to abide by a BSL to use data from this service". What that actually would mean is so nebulous that it feels threatening.

[0] Source: https://registry.terraform.io/v1/providers/airbytehq/airbyte...

[1] Source: https://github.com/airbytehq/terraform-provider-airbyte/tree... gzipped : ~300 resources, ~300 data sources

(NB: in airbyte's case the TF Provider was generated from a ~150Kb OpenAPI spec via https://speakeasyapi.dev: implying docs could be compressed even more)

I'm working on a company https://speakeasyapi.dev/ with the goal helping companies in this ecosystem get great production quality client sdks, terraform providers, cli(s) and all the developer surfaces you may want supported for our API. We also manage the spec and publishing workflow for you so all you have to do is build your API and we'll do the rest.

Feel free to email me at [email protected] or join our slack (https://join.slack.com/t/speakeasy-dev/shared_invite/zt-1cwb...) . We're in open beta and working with a few great companies already and we'd be happy for you to try out the platform for free!

Hi all I am a founding engineer for a API Experience company called Speakeasy - speakeasyapi.dev and we have recently released a Client SDK Generator for APIs using OpenAPI 3.0.X documents (soon to support 3.1). The generator will generate idiomatic Golang SDKs (along with other languages) that feel natural to use, easy to mock, and just work. The generator is free to use and can be run via a standalone golang built CLI with no external dependencies that can be easily installed as a binary or via homebrew (mac & linux). Check it out here https://github.com/speakeasy-api/speakeasy. If you have any questions or want to get in touch to see how Speakeasy can help you improve your APIs, just let me know!

The generator has been battle tested on thousands of APIs and we are sharing the results in our github repo. If you want to try it out on your own, download the CLI or brew install and get started in minutes:

Under the hood, the tags (and associated query language) are actually Taxi - an OSS meta-language and toolchain we build (and have shared previously).

Taxi is a relatively new entrant in the API space. It's goal is to let developers add simple, (but type-safe) tags into their APIs, so software can understand how different APIs relate to one another.

One of the gripes I have about OpenAPI is that it has a very low signal-to-noise ratio. It was bad in JSON, it's just as bad in YAML, with an added whitespace pedantry.

It's great to see a number of alternatives listed in this thread - there's much more active development in this space then I was aware of, and I hope that some of it gets upstreamed back into OpenAPI.

I'll shamelessly plug our tool in this space - Taxi (https://github.com/taxilang/taxilang), which has a dedicated DSL (not YAML) you can either use standalone, or embeddedd within OpenAPI.

I also happen to think that (for internal teams at least), generating clients on ${apiSpec} is a form of tight coupling, where producer and consumer become tied together. If you can avoid it, you should, as it allows producers and consumers to stay loosely coupled and evolve independetly without the gymnastics of avoiding breaking changes.

I've talked about this before, with proposed solutions.[0]

[0]https://orbitalhq.com/blog/2023-01-16-using-semantic-metadat...