Add TestContainers for running the Mock server

[jdubois]

As a Java developer, I don't want to run those NPM scripts on my machine, could we have TestContainers support to automate running the Mock server inside a container?

It would be something like:

val prism =
            GenericContainer("stoplight/prism:5.12.1")
                .withCommand(
                    "mock https://storage.googleapis.com/stainless-sdk-openapi-specs/openai-0d64ca9e45f51b4279f87b205eeb3a3576df98407698ce053f2e2302c1c08df1.yml"
                )
                .withExposedPorts(4010)

        prism.start()

[TomerAberbach]

Interesting idea!

I'm curious, are you using the mock Prism server to write tests for your usage of the SDK?

[jdubois]

I wanted to have a look at your SDK and maybe contribute (I work at Microsoft, and I also contribute to @langchain4j ), so first thing I’m doing is building the project and running the tests.

The goal here is to make it easier to contribute.
Here’s a similar integration test from LangChain4j, with TestContainers, to give you an example:
https://github.com/langchain4j/langchain4j/blob/main/document-loaders/langchain4j-document-loader-azure-storage-blob/src/test/java/dev/langchain4j/data/document/loader/azure/storage/blob/LocalAzureBlobStorageDocumentLoaderIT.java

[TomerAberbach]

Cool, I'll take a look at that!

Just to set expectations, this SDK is generated by the Stainless code generator (I work there), and we're actively making changes to the generator. You'll see reflected in releases of the SDK.

So if you want to work on something or have ideas, then I would recommend running the idea buy us before starting, just to make sure you don't waste your time and that whatever custom/manual work you do could conceivably be integrated back into the code generation.

[jdubois]

Ah ah I totally know who you are and what you do at Stainless, I work with the Java SDK team at Microsoft :-)
Also I built @jhipster and I understand how you work with the generator, which is why I posted this ticket and not a PR.

To be transparent: my plan is to add OpenAI Java SDK support to LangChain4j in the next weeks, so as soon as I see anything to update or modify, I'll open up tickets here.

[TomerAberbach]

Ah ah I totally know who you are and what you do at Stainless, I work with the Java SDK team at Microsoft :-) Also I built @jhipster and I understand how you work with the generator, which is why I posted this ticket and not a PR.

Nice :)

To be transparent: my plan is to add OpenAI Java SDK support to LangChain4j in the next weeks, so as soon as I see anything to update or modify, I'll open up tickets here.

Sounds great! In case you didn't know/see, just a heads up that this SDK is in alpha so there will likely be some breaking changes, in case that affects your integration

[TomerAberbach]

@jdubois I'm curious if know what work would be involved to address this issue?

I'm personally fine with npm being required to set up the mock server, so if setting up a container is a big effort, then I might just close this issue.

[jdubois]

I have it working but it’s a bit complex…
I’ll clean it up and do a PR tomorrow.
It’s much easier to use and more reproducible (no npm install, it’s the Docker image made by Prism), but it’s a good number of lines at the moment.

[jdubois]

PR is done at #537

[TomerAberbach]

Gonna close this for now since not sure the slowdown from the PR is worth it, but feel free to reopen if you have any ideas around how to speed up