特徴
ソリューション
ドキュメント
資源
ファームズ
GPTBots ドキュメント > TUTORIAL > FlowBot > Flow Component
GPTBots ドキュメント > TUTORIAL > FlowBot > Flow Component
Flow Components
最新の更新：2024-02-20

Flow Components

LLM

image-20240923163130980

OpenAI ChatGPT

This component uses input information, calls the ChatGPT service provided by OpenAI based on the component configuration.

The configuration items are consistent with the LLM in the Bot, and will not be further elaborated here.

Azure ChatGPT

This component uses input information, calls the ChatGPT service provided by Microsoft Azure based on the component configuration.

The configuration items are consistent with the LLM in the Bot, and will not be further elaborated here.

Baidu ERNIE Bot

This component uses input information, calls the ERNIE Bot service provided by Baidu based on the component configuration.

The configuration items are consistent with the LLM in the Bot, and will not be further elaborated here.

Knowledge Data

image-20240923163211805

This component uses the input information, performs vector retrieval in the selected knowledge document based on the component configuration, and outputs the knowledge recall results.

Within this component, you can select specific knowledge documents from the Bot's knowledge base as the knowledge retrieval range for the current component.

The remaining configuration items are consistent with the knowledge base in the Bot, and will not be further elaborated here.

Flow Rules

Branch

image-20240923163316778

This component uses input information, based on component configuration, through LLM capabilities, based on natural language, to determine the intent of the input content, and outputs through the corresponding branch.

Condition

image-20240923163442056

This component uses input information, based on component configuration, through LLM capabilities, based on natural language, to determine the condition of the input content, and outputs through the corresponding condition.

Preset Response

Cards

image-20240923163551075

This component ignores the input information and directly outputs the preset content.

The preset response content can be configured as the following types:

  • Text: Output as text.
  • Card: Output as card. Rich information can be configured in the card, such as title, picture, video, location, button events, etc.
  • Form: Output as form. The form input fields and submit URL can be configured in the form.
  • JSON: Output in JSON structure, suitable for scenarios where the downstream needs to be submitted to the API, etc.