LLM工具交互服务

通过MCP服务器配置与大型语言模型交互的工具服务,支持多模态输入和文档查询功能。

  • by rectalogic
  • Python
  • 本地部署
  • 更新于
在 GitHub 查看 rectalogic/toolchat

使用说明

工具聊天

使用与LLMs聊天 MCP服务器.

在YAML文件中配置MCP工具服务器,并提供路径 --tool-servers, 启用特定工具 --enable-tool-server. 看 样品-甲苯.yml 对于格式。

如果模型是多模式的,您还可以在提示中包含图像、文档等。 请参阅 /help 命令以获取详细信息。

示例

使用带有Gemini Pro 2.5的AWS文档工具搜索AWS文档:

$ uv run toolchat --model google-gla:gemini-2.5-pro-exp-03-25 --tool-servers sample-tools.yml --enable-tool-server awslabs.aws-documentation-mcp-server
ToolChat - Ctrl-D or /quit to quit
Enter /multi to enter and exit multiline mode, /help for more commands
> look up documentation on S3 bucket naming rule. cite your sources
Okay, I can help with that. I will search for the S3 bucket naming rules
documentation and provide a summary along with the source URL.

Tool search_documentation {'search_phrase': 'S3 bucket naming rules', 'limit': 1}

Okay, I found a relevant document. I will now read the content to summarize the
S3 bucket naming rules.

Tool read_documentation {'url': 'https://docs.aws.amazon.com/AmazonS3/latest/userguide/bucketnamingrules.html'}

Okay, here is a summary of the naming rules for Amazon S3 general purpose
buckets, based on the official AWS documentation:

S3 General Purpose Bucket Naming Rules:

 • Length: Must be between 3 and 63 characters long.
 • Characters: Can only contain lowercase letters (a-z), numbers (0-9), periods
   (.), and hyphens (-).
 • Start/End: Must begin and end with a letter or number.
 ...[elided]