[BUG] Bedrock multimodal model interface can't work with neural search

[ylwu-amzn]

What is the bug?
Model interface for Bedrock multi-modal model can't work with neural search

How can one reproduce the bug?
Steps to reproduce the behavior:

1. Create connector and model following this blueprint
2. Create ingest pipeline and try to ingest doc following this doc , then can see such error when ingest doc

PUT /my-nlp-index/_doc/1
{
 "image_description": "Orange table"
}

error

{
  "error": {
    "root_cause": [
      {
        "type": "status_exception",
        "reason": "Error validating input schema: Validation failed: [$: required property 'parameters' not found] for instance: {\"algorithm\":\"REMOTE\",\"text_docs\":[\"Orange table\"],\"return_bytes\":false,\"return_number\":true,\"target_response\":[\"sentence_embedding\"]} with schema: {\n    \"type\": \"object\",\n    \"properties\": {\n        \"parameters\": {\n            \"type\": \"object\",\n            \"properties\": {\n                \"inputText\": {\n                    \"type\": \"string\"\n                },\n                \"inputImage\": {\n                    \"type\": \"string\"\n                }\n            }\n        }\n    },\n    \"required\": [\n        \"parameters\"\n    ]\n}"
      }
    ],
    "type": "status_exception",
    "reason": "Error validating input schema: Validation failed: [$: required property 'parameters' not found] for instance: {\"algorithm\":\"REMOTE\",\"text_docs\":[\"Orange table\"],\"return_bytes\":false,\"return_number\":true,\"target_response\":[\"sentence_embedding\"]} with schema: {\n    \"type\": \"object\",\n    \"properties\": {\n        \"parameters\": {\n            \"type\": \"object\",\n            \"properties\": {\n                \"inputText\": {\n                    \"type\": \"string\"\n                },\n                \"inputImage\": {\n                    \"type\": \"string\"\n                }\n            }\n        }\n    },\n    \"required\": [\n        \"parameters\"\n    ]\n}"
  },
  "status": 400
}

What is the expected behavior?
This is a break change. We should not break BWC by introducing model interface.

What is your host/environment?

• OS: Linux
• Version 2.17.1
• Plugins

Do you have any screenshots?
If applicable, add screenshots to help explain your problem.

Do you have any additional context?
Add any other context about the problem.

[maltelorbach]

Same behaviour in 2.18.0.

Digging into the code, I think the issue is the algorithm type used by the text_image_embedding processor. It uses a hardcoded value FunctionName.TEXT_EMBEDDING when instantiating the MLInput here:
https://github.com/opensearch-project/neural-search/blob/7feacd67b3c7694ff4a1c1c2b430f2447a1ed4ab/src/main/java/org/opensearch/neuralsearch/ml/MLCommonsClientAccessor.java#L293

but for Bedrock (and possibly others), the algorithm type should be FunctionName.REMOTE such that the inputs are parsed from parameters:

ml-commons/common/src/main/java/org/opensearch/ml/common/input/MLInput.java

Lines 203 to 206 in c2a40c1

	case REMOTE:
	RemoteInferenceInputDataSet remoteInferenceInputDataSet = (RemoteInferenceInputDataSet) this.inputDataset;
	Map<String, String> parameters = remoteInferenceInputDataSet.getParameters();
	builder.field(PARAMETERS_FIELD, parameters);