OpenAPIServiceConnector

Dynamically invoke OpenAPI service methods using information from a ChatMessage. OpenAPIServiceConnector acts as an interface between deepset and OpenAPI services.

Basic Information

• Type: haystack_integrations.connectors.openapi_service.OpenAPIServiceConnector
• Components it can connect with:
  • ChatGenerators: It receives messages from a ChatGenerator.
  • OutputAdapter: It sends the service response to an OutputAdapter that converts it to a format subsequent components can use.

Inputs

Parameter	Type	Default	Description
messages	List[ChatMessage]		A list of ChatMessage objects containing the messages to be processed. The last message is expected to contain parameter invocation payload.
service_openapi_spec	Dict[str, Any]		The OpenAPI JSON specification object of the service to be invoked. All ref values must be resolved.
service_credentials	Optional[Union[dict, str]]	None	The credentials to be used for authentication with the service. Currently, only the http and apiKey OpenAPI security schemes are supported. Use http for Basic, Bearer, or other HTTP authentication schemes. Use apiKey for API keys and cookie authentication.

Outputs

Parameter	Type	Default	Description
service_response	Dict[str, Any]		A list of ChatMessage objects, each containing the response from the service. The response is in JSON format, and the content attribute of the ChatMessage contains the JSON string.

Overview

OpenAPIServiceConnector connects the deepset to OpenAPI services, making it possible to call operations as defined in the OpenAPI specification of the service.

It works with the ChatMessage data class, using the message payload to identify the method to call and the parameters to pass. The payload must be a JSON-formatted function call string that includes the method name and parameters. The connector then invokes the method on the OpenAPI service and returns the response as a ChatMessage.

Before using this component, you typically resolve service endpoint parameters with the OpenAPIServiceToFunctions component.

Keep in mind that OpenAPIServiceConnector is usually not used on its own. It’s designed to be part of a pipeline with the OpenAPIServiceToFunctions component and an OpenAIChatGenerator component that supports function calling. The payload is usually generated by the OpenAIChatGenerator.

Usage Example

Here's a complete YAML pipeline that demonstrates how to use OpenAPIServiceConnector with other components to create a function calling pipeline that:

1. Converts OpenAPI specifications to function schemas using OpenAPIServiceToFunctions
2. Uses an OpenAIChatGenerator to generate function calls
3. Connects to the OpenAPI service using OpenAPIServiceConnector
4. Processes the response through output adapters
5. Generates the final answer using another OpenAIChatGenerator

components:
  spec_to_functions:
    type: haystack.components.converters.openapi_functions.OpenAPIServiceToFunctions
    init_parameters: {}
  functions_llm:
    type: haystack.components.generators.chat.openai.OpenAIChatGenerator
    init_parameters:
      api_key:
        type: env_var
        env_vars:
        - OPENAI_API_KEY
        strict: false
      model: gpt-3.5-turbo-0613

  openapi_container:
    type: haystack.components.connectors.openapi_service.OpenAPIServiceConnector

    init_parameters: {}
  a1:
    type: haystack.components.converters.output_adapter.OutputAdapter
    init_parameters:
      template: |
        {
          "tools": [{
            "type": "function",
            "function": {{ functions[0] }}
          }],
          "tool_choice": {
            "type": "function",
            "function": {"name": {{ functions[0].name }}}
          }
        }
      output_type: Dict[str, Any]
      custom_filters: {}
  a2:
    type: haystack.components.converters.output_adapter.OutputAdapter
    init_parameters:
      template: "{{specs[0]}}"
      output_type: Dict[str, Any]

  a3:
    type: haystack.components.converters.output_adapter.OutputAdapter
    init_parameters:
      template: "{{system_message + service_response}}"
      output_type: typing.List[haystack.dataclasses.ChatMessage]

  llm:
    type: haystack.components.generators.chat.openai.OpenAIChatGenerator
    init_parameters:
      api_key:
        type: env_var
        env_vars:
        - OPENAI_API_KEY
        strict: false
      model: gpt-4-1106-preview

connections:
- sender: spec_to_functions.functions
  receiver: a1.functions
- sender: spec_to_functions.openapi_specs
  receiver: a2.specs
- sender: functions_llm.replies
  receiver: openapi_container.messages
- sender: openapi_container.service_response
  receiver: a3.service_response
- sender: a1.output
  receiver: functions_llm.generation_kwargs
- sender: a2.output
  receiver: openapi_container.service_openapi_spec
- sender: a3.output
  receiver: llm.messages

inputs:
  query:  # These components will receive the query as input
  - "functions_llm.messages"
  - "spec_to_functions.sources"
  filters:  # These components will receive filters as input
  - "openapi_container.service_credentials"
  - "a3.system_message"

outputs:  # Defines the output of your pipeline
  answers: "llm.replies"  # The output of the pipeline is the generated answer
  service_response: "openapi_container.service_response"  # The output from the OpenAPI service

max_runs_per_component: 100

metadata: {}

The pipeline fetches the OpenAPI specification and system prompt from the provided URLs and processes the user query about Sam Altman's ouster from OpenAI.

Parameters

Init Parameters

These are the parameters you can configure in Pipeline Builder:

Parameter	Type	Default	Description
ssl_verify	Optional[Union[bool, str]]	None	Decide whether to use SSL verification for the requests. If a string is passed, itis used as the CA.

Run Method Parameters

These are the parameters you can configure for the component's run() method. This means you can pass these parameters at query time through the API, in Playground, or when running a job. For details, see Modify Pipeline Parameters at Query Time.

Parameter	Type	Default	Description
messages	List[ChatMessage]		A list of ChatMessage objects containing the messages to be processed. The last message should contain the tool calls.
service_openapi_spec	Dict[str, Any]		The OpenAPI JSON specification object of the service to be invoked. All the refs should already be resolved.
service_credentials	Optional[Union[dict, str]]	None	The credentials to be used for authentication with the service. Currently, only the http and apiKey OpenAPI security schemes are supported.