Connector tool
Introduced 2.15
The ConnectorTool
uses a connector to call any REST API function. For example, you can use a ConnectorTool
to call a Lambda function through its REST API interface.
Step 1: Register a connector with an execute action
The ConnectorTool
can only run an execute
action within a connector. Before you can create a ConnectorTool
, you need to configure a connector and provide an execute
action in the actions
array. The execute
action is used to invoke a function at a REST API endpoint. It is similar to the predict
action, which is used to invoke a machine learning (ML) model.
For this example, you’ll create a connector for a simple AWS Lambda function that accepts two integers and returns their sum. This function is hosted on a dedicated endpoint with a specific URL, which you’ll provide in the url
parameter. For more information, see Lambda function URLs.
To create a connector, send the following request:
POST _plugins/_ml/connectors/_create
{
"name": "Lambda connector of simple calculator",
"description": "Demo connector of lambda function",
"version": 1,
"protocol": "aws_sigv4",
"parameters": {
"region": "YOUR AWS REGION",
"service_name": "lambda"
},
"credential": {
"access_key": "YOUR ACCESS KEY",
"secret_key": "YOUR SECRET KEY",
"session_token": "YOUR SESSION TOKEN"
},
"actions": [
{
"action_type": "execute",
"method": "POST",
"url": "YOUR LAMBDA FUNCTION URL",
"headers": {
"content-type": "application/json"
},
"request_body": "{ \"number1\":\"${parameters.number1}\", \"number2\":\"${parameters.number2}\" }"
}
]
}
copy
OpenSearch responds with a connector ID:
{
"connector_id": "Zz1XEJABXWrLmr4mewEF"
}
Step 2: Register a flow agent that will run the ConnectorTool
For this example, the Lambda function adds the two input numbers and returns their sum in the result
field:
{
"result": 5
}
By default, the ConnectorTool
expects the response from the Lambda function to contain a field named response
. However, in this example the Lambda function response doesn’t include a response
field. To retrieve the result from the result
field instead, you need to provide a response_filter
, specifying the JSON path to the result
field ($.result
). Using the response_filter
, the ConnectorTool
will retrieve the result with the specified JSON path and return it in the response
field.
To configure the Lambda function workflow, create a flow agent. A flow agent runs a sequence of tools in order and returns the last tool’s output. To create a flow agent, send the following register agent request, providing the connector ID from the previous step and a response_filter
:
POST /_plugins/_ml/agents/_register
{
"name": "Demo agent of Lambda connector",
"type": "flow",
"description": "This is a demo agent",
"app_type": "demo",
"tools": [
{
"type": "ConnectorTool",
"name": "lambda_function",
"parameters": {
"connector_id": "YOUR CONNECTOR ID",
"response_filter": "$.result"
}
}
]
}
copy
For parameter descriptions, see Register parameters.
OpenSearch responds with an agent ID:
{
"agent_id": "az1XEJABXWrLmr4miAFj"
}
Step 3: Run the agent
Then, run the agent by sending the following request:
POST /_plugins/_ml/agents/9X7xWI0Bpc3sThaJdY9i/_execute
{
"parameters": {
"number1": 2,
"number2": 3
}
}
copy
OpenSearch returns the output of the Lambda function execution. In the output, the field name is response
, and the result
field contains the Lambda function result:
{
"inference_results": [
{
"output": [
{
"name": "response",
"result": 5
}
]
}
]
}
Register parameters
The following table lists all tool parameters that are available when registering an agent.
Parameter | Type | Required/Optional | Description |
---|---|---|---|
connector_id | String | Required | A connector ID of a connector configured with an execute action that invokes an API. |
response_filter | String | Optional | A JSON path to the response field that contains the result of invoking the API. If a response_filter is not specified, then the ConnectorTool expects the API response to be in a field named response . |
Execute parameters
When running the agent, you can define any parameter needed for the API call in the request_body
of your connector’s execute
action. In this example, the parameters are number1
and number2
:
"actions": [
{
"action_type": "execute",
"method": "POST",
"url": "YOUR LAMBDA FUNCTION URL",
"headers": {
"content-type": "application/json"
},
"request_body": "{ \"number1\":\"${parameters.number1}\", \"number2\":\"${parameters.number2}\" }"
}
]