fix(settings): Add fallback in case models httpRequest returns a JsonArray instead of JsonObject

[ETdoFresh]

I wanted to have support for https://www.together.ai/. However, AI Commits did not work with this endpoint out the box. I found out that the results from the "models" call differed.

From OpenAI, a call to https://api.openai.com/v1/models looked something like this:

{
    "object": "list",
    "data": [
        {
            "id": "dall-e-3",
            "object": "model",
            "created": 1698785189,
            "owned_by": "system"
        },
        ...
    ]
}

Where as Together.ai, a call to https://api.together.xyz/v1/models looked something like this:

[
    {
        "id": "Austism/chronos-hermes-13b",
        "object": "model",
        "created": 1692896905
    },
    ...
]

So, I ended up trying openai.models() first, and if it failed, fall back to parsing the jsonarray manually. Which leads me to another pain point, which was that owned_by is also not present in together.ai's model. So fallback only requires id and created.

If this is helpful to others, feel free to use this. Here's a video of it in action...

java_zuqwE6Wfgr.mp4

[Blarc]

Hey @ETdoFresh,

thanks for opening this PR. I was thinking about supporting different providers anyway, so I was planning to implement this a bit more generically via interface class, so new providers can easily be added. (Since I already had requests for other providers #132)

I'll look into this ASAP.

[ETdoFresh]

Excellent! Yeah, feel free to close this PR if you already have plans for this. I found out I had to make another tweak because created was required on openAI models call as well, but when I use LLM Studio, it did not have a created field when calling models. Sooo... yeah!

And I went ahead and ran "trim" on the commit message as well, because sometimes I had leading spaces.

My changes will be on my personal modified version of main branch... but yeah, let me know if there is anything I can do to help.

[Blarc]

Hey @ETdoFresh,

I have refactored the code. I now use langchain4j for creating LLM API clients. Even though together.ai is currently not supported by langchain4j, adding it to the plugin should now be easier.