Node - Dify Docs

A plugin can reverse invoke the capabilities of certain nodes within a Dify Chatflow/Workflow application.

Plugins can call the ParameterExtractor and QuestionClassifier nodes. Both encapsulate complex prompt and code logic, using LLMs to handle tasks that are difficult to solve with hardcoded rules.

Call the Parameter Extractor Node#

Entry Point#

    self.session.workflow_node.parameter_extractor

Interface#

    def invoke(        self,        parameters: list[ParameterConfig],        model: ModelConfig,        query: str,        instruction: str = "",    ) -> NodeResponse        pass

parameters: The list of parameters to extract.
model: Conforms to the LLMModelConfig specification.
query: The source text for parameter extraction.
instruction: Any additional instructions the LLM might need.

For the structure of NodeResponse, see the General Specifications Definition.

Use Case#

This example extracts a person’s name from a conversation:

from collections.abc import Generatorfrom dify_plugin import Toolfrom dify_plugin.entities.tool import ToolInvokeMessagefrom dify_plugin.entities.workflow_node import ModelConfig, NodeResponse, ParameterConfigclass ParameterExtractorTool(Tool):    def _invoke(        self, tool_parameters: dict    ) -> Generator[ToolInvokeMessage, None, None]:        response: NodeResponse = self.session.workflow_node.parameter_extractor.invoke(            parameters=[                ParameterConfig(                    name="name",                    description="name of the person",                    required=True,                    type="string",                )            ],            model=ModelConfig(                provider="langgenius/openai/openai",                name="gpt-4o-mini",                completion_params={},            ),            query="My name is John Doe",            instruction="Extract the name of the person",        )        extracted_name = response.outputs.get("name", "Name not found")        yield self.create_text_message(extracted_name)

NodeResponse is a Pydantic model defined in dify_plugin.entities.workflow_node with three dictionary fields: process_data, inputs, and outputs. Extracted values live under response.outputs.

Call the Question Classifier Node#

Entry Point#

    self.session.workflow_node.question_classifier

Interface#

    def invoke(        self,        classes: list[ClassConfig],        model: ModelConfig,        query: str,        instruction: str = "",    ) -> NodeResponse:        pass

ClassConfig is also exported from dify_plugin.entities.workflow_node. The interface parameters match those of ParameterExtractor, and the final result is stored in response.outputs["class_name"].