ML inference search response processor

Introduced 2.16

The ml_inference search response processor is used to invoke registered machine learning (ML) models in order to incorporate their outputs as new fields in documents within search results.

PREREQUISITE
Before using the ml_inference search response processor, you must have either a local ML model hosted on your OpenSearch cluster or an externally hosted model connected to your OpenSearch cluster through the ML Commons plugin. For more information about local models, see Using ML models within OpenSearch. For more information about externally hosted models, see Connecting to externally hosted models.

Syntax

The following is the syntax for the ml-inference search response processor:

{
  "ml_inference": {
    "model_id": "<model_id>",
    "function_name": "<function_name>",
    "full_response_path": "<full_response_path>",
    "model_config":{
      "<model_config_field>": "<config_value>"
    },
    "model_input": "<model_input>",
    "input_map": [
      {
        "<model_input_field>": "<document_field>"
      }
    ],
    "output_map": [
      {
        "<new_document_field>": "<model_output_field>"
      }
    ],
    "override": "<override>",
    "one_to_one": false
  }
}

Request body fields

The following table lists the required and optional parameters for the ml-inference search response processor.

The input_map and output_map mappings support standard JSON path notation for specifying complex data structures.

Setup

Create an index named my_index and index one document to explain the mappings:

POST /my_index/_doc/1
{
  "passage_text": "hello world"
}

Using the processor

Follow these steps to use the processor in a pipeline. You must provide a model ID when creating the processor. Before testing a pipeline using the processor, make sure that the model is successfully deployed. You can check the model state using the Get Model API.

For local models, you must provide a model_input field that specifies the model input format. Add any input fields in model_config to model_input.

For remote models, the model_input field is optional, and its default value is "{ \"parameters\": ${ml_inference.parameters} }.

Example: Externally hosted model

The following example shows you how to configure an ml_inference search response processor with an externally hosted model.

Step 1: Create a pipeline

The following example shows you how to create a search pipeline for an externally hosted text embedding model. The model requires an input field and generates results in a data field. It converts the text in the passage_text field into text embeddings and stores the embeddings in the passage_embedding field. The function_name is not explicitly specified in the processor configuration, so it defaults to remote, signifying an externally hosted model:

PUT /_search/pipeline/ml_inference_pipeline
{
  "description": "Generate passage_embedding when search documents",
  "processors": [
    {
      "ml_inference": {
        "model_id": "<your model id>",
        "input_map": [
          {
            "input": "passage_text"
          }
        ],
        "output_map": [
          {
            "passage_embedding": "data"
          }
        ]
      }
    }
  ]
}

When making a Predict API request to an externally hosted model, all necessary fields and parameters are usually contained within a parameters object:

POST /_plugins/_ml/models/cleMb4kBJ1eYAeTMFFg4/_predict
{
  "parameters": {
    "input": [
      {
        ...
      }
    ]
  }
}

When specifying the input_map for an externally hosted model, you can directly reference the input field instead of providing its dot path parameters.input:

"input_map": [
  {
    "input": "passage_text"
  }
]

Step 2: Run the pipeline

Run the following query, providing the pipeline name in the request:

GET /my_index/_search?search_pipeline=ml_inference_pipeline_local
{
  "query": {
    "match_all": {
    }
  }
}

The response confirms that the processor has generated text embeddings in the passage_embedding field. The document within _source now contains both the passage_text and passage_embedding fields:

{
  "took": 288,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 1,
      "relation": "eq"
    },
    "max_score": 0.00009405752,
    "hits": [
      {
        "_index": "my_index",
        "_id": "1",
        "_score": 0.00009405752,
        "_source": {
          "passage_text": "hello world",
          "passage_embedding": [
            0.017304314,
            -0.021530833,
            0.050184276,
            0.08962978,
            ...]
        }
      }
      }
    ]
  }
}

Example: Local model

The following example shows you how to configure an ml_inference search response processor with a local model.

Step 1: Create a pipeline

The following example shows you how to create a search pipeline for the huggingface/sentence-transformers/all-distilroberta-v1 local model. The model is a pretrained sentence transformer model hosted in your OpenSearch cluster.

If you invoke the model using the Predict API, then the request appears as follows:

POST /_plugins/_ml/_predict/text_embedding/cleMb4kBJ1eYAeTMFFg4
{
  "text_docs":[ "today is sunny"],
  "return_number": true,
  "target_response": ["sentence_embedding"]
}

Using this schema, specify the model_input as follows:

 "model_input": "{ \"text_docs\": ${input_map.text_docs}, \"return_number\": ${model_config.return_number}, \"target_response\": ${model_config.target_response} }"

In the input_map, map the passage_text document field to the text_docs field expected by the model:

"input_map": [
  {
    "text_docs": "passage_text"
  }
]

Because you specified the field to be converted into embeddings as a JSON path, you need to set the full_response_path to true. Then the full JSON document is parsed in order to obtain the input field:

"full_response_path": true

The text in the passage_text field will be used to generate embeddings:

{
  "passage_text": "hello world"
}

The Predict API request returns the following response:

{
  "inference_results" : [
    {
      "output" : [
        {
          "name" : "sentence_embedding",
          "data_type" : "FLOAT32",
          "shape" : [
            768
          ],
          "data" : [
            0.25517133,
            -0.28009856,
            0.48519906,
            ...
          ]
        }
      ]
    }
  ]
}

The model generates embeddings in the $.inference_results.*.output.*.data field. The output_map maps this field to the newly created passage_embedding field in the search response document:

"output_map": [
  {
    "passage_embedding": "$.inference_results.*.output.*.data"
  }
]

To configure an ml_inference search response processor with a local model, specify the function_name explicitly. In this example, the function_name is text_embedding. For information about valid function_name values, see Request fields.

The following is the final configuration of the ml_inference search response processor with the local model:

PUT /_search/pipeline/ml_inference_pipeline_local
{
  "description": "search passage and generates embeddings",
  "processors": [
    {
      "ml_inference": {
        "function_name": "text_embedding",
        "full_response_path": true,
        "model_id": "<your model id>",
        "model_config": {
          "return_number": true,
          "target_response": ["sentence_embedding"]
        },
        "model_input": "{ \"text_docs\": ${input_map.text_docs}, \"return_number\": ${model_config.return_number}, \"target_response\": ${model_config.target_response} }",
        "input_map": [
          {
            "text_docs": "passage_text"
          }
        ],
        "output_map": [
          {
            "passage_embedding": "$.inference_results.*.output.*.data"
          }
        ],
        "ignore_missing": true,
        "ignore_failure": true
      }
    }
  ]
}

Step 2: Run the pipeline

Run the following query, providing the pipeline name in the request:

GET /my_index/_search?search_pipeline=ml_inference_pipeline_local
{
"query": {
  "term": {
    "passage_text": {
      "value": "hello"
      }
    }
  }
}

Response

The response confirms that the processor has generated text embeddings in the passage_embedding field:

{
  "took": 288,
  "timed_out": false,
  "_shards": {
    "total": 1,
    "successful": 1,
    "skipped": 0,
    "failed": 0
  },
  "hits": {
    "total": {
      "value": 1,
      "relation": "eq"
    },
    "max_score": 0.00009405752,
    "hits": [
      {
        "_index": "my_index",
        "_id": "1",
        "_score": 0.00009405752,
        "_source": {
          "passage_text": "hello world",
          "passage_embedding": [
            0.017304314,
            -0.021530833,
            0.050184276,
            0.08962978,
            ...]
        }
      }
    ]
  }
}