fix: update assistants components and add integrations tests

[phact]

A couple of the assistants components were broken due to the way the OpenAI sdk client was being set up so I fixed and added integration tests.

[aws-amplify-sa-east-1]

This pull request is automatically being deployed by Amplify Hosting (learn more).

Access this pull request here: https://pr-3887.dmtpw4p5recq1.amplifyapp.com

[jordanrfrazier]

Are there potential issues that may arise from using a class variable for the client? I see we're creating a new assistant for each invocation, so this seems low risk, but just wondering.

[phact]

Previously we were doing one per invocation so I actually think a client per class is less problematic. If they were async clients we could have races but these are synchronous.

One potential future improvement would be to do a shared client for all the assistants. Not sure what a good pattern for this would be, just a singleton?

[jordanrfrazier]

I looked over at the Dotenv component, and curious how that works.

Q) Why does it need an output method?
Q) If I have a .env file I've loaded up during my langflow run --env .env, and I use this component with a separate env file, will it replace that entire set, or union the two?

I see why this is needed though - if I can't do langflow run myself (in Astra), we need an easy way to pass env vars.

class Dotenv(Component):
    display_name = "Dotenv"
    description = "Load .env file into env vars"

    inputs = [
        MultilineSecretInput(
            name="dotenv_file_content",
            display_name="Dotenv file content",
            info="Paste the content of your .env file directly, since contents are sensitive, using a Global variable set as 'password' is recommended",
        )
    ]

    outputs = [
        Output(display_name="env_set", name="env_set", method="process_inputs"),
    ]

    def process_inputs(self) -> Message:
        fake_file = io.StringIO(self.dotenv_file_content)
        result = load_dotenv(stream=fake_file, override=True)

        message = Message(text="No variables found in .env")
        if result:
            message = Message(text="Loaded .env")
        return message

[phact]

Yeah assistants supports loads of LLM providers so it's very convenient to allow the user to add them all in one go. If there are existing env vars that are set they do not get unset but they could get overwritten.