Flow Bot is particularly suitable for solving complex and dynamic business scenarios and problems. It allows for the coordination and division of labor among multiple LLMs (Large Language Models) by connecting them in series or parallel, enabling the rapid construction of a workflow.

The core element of Flow Bot is the component. Components are the basic building blocks of Flow Bot and are functional modules that encapsulate different capabilities for specific business scenarios. The supported component categories in GPTBots currently include:

  1. LLMs: Components based on the abilities of large language models like ChatGPT. They support AI conversations, identity settings, short-term and long-term memory, plug-in invocation, and allow for flexible arrangement of data submitted to the LLM to achieve the latest AI response quality.

  2. Knowledge Vector: Components based on the ability to search knowledge vectors. Developers have the freedom to choose the scope, number, and scores of knowledge recall, enabling precise utilization of the Bot's knowledge base.

  3. Logical Judgment: Components encapsulating logical judgment based on capabilities provided by large language models such as ChatGPT. Developers can customize conditions and automatically extract relevant content that meets the criteria, outputting it to downstream components.

  4. Preset Responses: Supports developers in customizing predefined content, currently limited to text format but with plans to support more UI interactive formats in the future (such as images, videos, forms, IM, etc.).

Last updated