Add Connectors and ML updates for 2.9

[Naarcha-AWS]

Fixes #3063

Checklist

• By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license and subject to the Developers Certificate of Origin.
  For more information on following Developer Certificate of Origin and signing off your commits, please check here.

[kolchfa-aws]

LGTM except for some comments.

[natebower]

@Naarcha-AWS A few more changes. Let me know if I reviewed what you needed or if there's other content that needs my review.

[ylwu-amzn]

Add this to settings page

[ylwu-amzn]

remove this?

[ylwu-amzn]

Most connectors require the use of dedicated ML nodes -> By default, connectors require the use of dedicated ML nodes. Actually this is the default setting: "plugins.ml_commons.only_run_on_ml_node": true

For remote connector, it consumes much less resource. So should be ok if user prefer to run on data node. If they don't have dedicate ML node, and prefer to run on data node, they can set

PUT /_cluster/settings
{
    "persistent": {
        "plugins.ml_commons.only_run_on_ml_node": false
    }
}

[ylwu-amzn]

parameter -> parameters

Type is : Map<String, ?>

including endpointandmodel. -> for example endpointandmodel.

[ylwu-amzn]

From security team , we should call out all parameters in this block will be overridable in predict request. User can provide parameter with same in predict request to override the default parameter value defined in connector.

[ylwu-amzn]

Type: Map<string, string>

[ylwu-amzn]

ML Commons uses **AES/GCM/NoPadding** symmetric encryption with a key length of 32 bytes. When a connection cluster first starts, the key persists in OpenSearch. Therefore, you do not need to manually encrypt the key.
->
ML Commons uses **AES/GCM/NoPadding** symmetric encryption to encrypt your credential. When a connection cluster first starts, ml-commons will create a random 32 bytes key and persist in OpenSearch system index. Therefore, you do not need to manually set the encryption key.

[ylwu-amzn]

Tells the connector what actions to run after a connection to ML Commons has been established.
->
Define what actions can run within this connector.

[ylwu-amzn]

Type: Map<String, String>

[ylwu-amzn]

Default is application/json
->
Default "Content-Type" is "application/json"

[ylwu-amzn]

For example, to connect to a ChatGPT completion model,

->

For example, to connect to a ChatGPT chat model,

[ylwu-amzn]

Remote this

and status

[ylwu-amzn]

The following example registers a model named openAI-GPT-3.5 completions:

->

The following example registers a model named openAI-gpt-3.5-turbo:

[ylwu-amzn]

Suggest add the example for creating this model group

POST /_plugins/_ml/model_groups/_register
{
  "name": "remote_model_group",
  "description": "This is an example description"
}

Sample response
{
  "model_group_id": "wlcnb4kBJ1eYAeTMHlV6",
  "status": "CREATED"
}

[ylwu-amzn]

We can remove this line, as model already defined in connector

[ylwu-amzn]

This is not chat model example. Let's keep consistent with the standalone connector.

[ylwu-amzn]

After creating the connector, you can retrieve the task_id, deploy the model, and use the Predict API, similar to a standalone connector.

->

After creating the connector, you can use the connector id to register model, deploy and predict.

[ylwu-amzn]

As all items defined in parameters will be overridable and visible (user can see parameters with get connector API). If user don't want this, they can move these two reserved parameters for aws_sigv4 to credential.

[ylwu-amzn]

aws-sigv4
->
aws_sigv4

Please also check all other places

[ylwu-amzn]

We should add some security warning for remote connector. As user need to configure credential in connector, for example AWS credential, openAI API key, user should always use security enabled cluster. Otherwise their credentials will not be protected which is risky.

[dylan-tong-aws]

@natebower any mention of "SageMaker" should be "Amazon SageMaker". I saw some instances of "AWS SageMaker" that needs to be corrected. The same for ChatGPT. I believe we our standard is to use the full name, "OpenAI ChatGPT".

[natebower]

@natebower any mention of "SageMaker" should be "Amazon SageMaker". I saw some instances of "AWS SageMaker" that needs to be corrected. The same for ChatGPT. I believe we our standard is to use the full name, "OpenAI ChatGPT".

@dylan-tong-aws Re: "Amazon SageMaker", that's correct, thanks for the callout. @Naarcha-AWS Can you please global find and replace? Re: ChatGPT, we don't actually have a standard for this, but given that OpenAI simply uses "ChatGPT", I'd prefer that we follow suit. Thanks!