中文

Assistants

The Assistant Management module is one of the core functional modules in the intelligent service platform, used for unified management and configuration of various AI assistants (also known as "Agents"). Through this module, administrators can create, edit, classify, deploy, and monitor multiple AI assistant instances to meet automation service needs in different business scenarios.

This module supports multi-dimensional configuration, including name, Prompt, model selection, knowledge base binding, MCP service integration, etc. It also provides search and permission control functions for administrators, facilitating efficient operation and maintenance for enterprise users.

Key Features

  • Add Assistant: Pop-up form to fill in emoji, name, model, and prompt fields.
  • Assistant List Display: Supports displaying all created assistants and their key attributes.
  • Search Assistant: Provides global search function to find assistants by name or other fields.
  • Delete Assistant: Supports deleting assistants that are no longer in use.
  • Edit Assistant: Supports deep configuration such as model settings, knowledge base settings, and MCP settings.

Assistant Editing Details

1. Model Settings

Configure the language model and related parameters used by the assistant, supporting automatic saving after modification. The main configuration items are as follows:

  • Model: Select the base model to use.
  • Temperature: Controls the randomness of the output. Range 0-2, default 1.0. Higher values make the output more random, lower values make it more deterministic.
  • Top P: Controls the diversity of the output. Range 0-1, default 1. Higher values make the output more diverse, lower values make it more concentrated.
  • Context Count: Controls the number of historical messages included in each conversation. Range 1-20, default 5.
  • Max Tokens: Select whether to enable, limiting the maximum number of tokens per reply.
  • Stream Output: Select whether to enable, enabling it will display the model's output content in real-time.
  • MCP Call Method: Select prompt call or function call (function call is more accurate, prompt call is more flexible).
  • Custom Parameters: Select whether to use directly, requires pasting custom parameters in JSON format directly.

2. Knowledge Base Management

This module is mainly used to manage the system's knowledge base, supporting the association of private knowledge documents with assistants, enabling them to answer questions based on internal enterprise data. Administrators can bind exclusive knowledge bases by setting fields such as description, embedding model, reranking model, chunk size, overlap size, and enable status, achieving answering capabilities based on private data.