AI and Home assistant -homeassistant-ai/ha-mcp-integration

我们使用自然语言跟大模型交流，大模型选择一个合适工具函数，把函数名称还有参数返回给我，再使用Python具体执行这个函数，这个就是大模型Function Call的一个实现思路。

你是一个智能家居AI，你的根据用户输入返回对函数，必须从函数列表里面选，#后面备注了函数说明，如果能查找到功能的函数，你只需要输出函数加参数，参数就加引号写在函数的括号里，如果没有参数则不加。如果查寻不到任何函数，就输出"对不起，办不到"

原文链接：https://blog.csdn.net/techshrimp/article/details/136186191

how homeassistant-ai/ha-mcp-integration create automation

ha-mcp-integration does not "think up" automations by itself. Instead, it exposes Home Assistant capabilities as MCP tools, allowing an AI (ChatGPT, Claude, Cursor, etc.) to inspect your Home Assistant configuration and then create or modify automations.

workflow:

User
  │
  ▼
LLM (ChatGPT / Claude)
  │
  ▼
MCP Server's Tool Calls
  │
  ▼
Home Assistant

For example:

Create an automation that turns on the bedroom AC when the temperature exceeds 28°C..

The AI MCP client perform these steps.

1. Discover available entities

list_entities()

2. Build the automation

The AI generates standard Home Assistant automation YAML:

alias: Bedroom AC Auto

trigger:
  - platform: numeric_state
    entity_id: sensor.bedroom_temperature
    above: 28

condition: []

action:
  - service: climate.set_hvac_mode
    target:
      entity_id: climate.bedroom_ac
    data:
      hvac_mode: cool

mode: single

4. Save the automation

Depending on the tools exposed by ha-mcp-integration, the AI:

automation.create(...)