ThinkInAI-Hackathon/ble-hrm-server

ble-hrm-server

A MCP server, which serve as a BLE Heart Rate Monitoring to connect with a HRM device.

Build Status

Installation

We recommend using a virtual environment and uv to manage dependencies.

This project uses pyproject.toml for dependency management.

# Create and activate a virtual environment
python3 -m venv .venv
source .venv/bin/activate  # On Windows use `.venv\Scripts\activate`

# Install uv if not already installed
pip install uv

# Install dependencies
uv sync

Environment Configuration

This project supports configuration via a .env file. You can use the provided .env.sample as a template.

cp .env.sample .env

The following environment variables are used:

• QINIU_ACCESS_KEY: Your Qiniu access key.
• QINIU_SECRET_KEY: Your Qiniu secret key.
• QINIU_BUCKET_NAME: The name of the Qiniu bucket.
• QINIU_BUCKET_DOMAIN: The domain associated with the Qiniu bucket.

Usage

Run the server

uv run src/hrm/server.py

After installing from PyPI, you can run the server directly using:

uvx ble-hrm-server

Run the tests

uv run pytest

Run the tests with coverage

uv run coverage run -m pytest
uv run coverage html

MCP Definition

Resources

Bluetooth HRM is based on Bluetooth protocol, we should use bleak to discover the device and connect to it.

• Discover Bluetooth Device & Filter by HRM profile (Heart Rate Service 0x180D)
• resource: discover://hrm

Tools

• Tool: Monitoring Heart Rate monitoring_heart_rate

  • Summary: Start monitoring the heart rate of the device for the given duration, default duration is 30 minutes (1800 sec). The monitoring will be done in the background.
  • Inputs:
    • device_id: str, the device UUID to monitor
    • duration: int, the duration to monitor, default is 1800 seconds (30 minutes)
  • Outputs: None

• Tool: Get Heart Rate get_heart_rate

  • Summary: Get the current HR, use last 10 sec and return the average of HR
  • Inputs: None (assumes baseline reading)
  • Output: Current HR (int), e.g. {"avg_hr": 60}

• Tool: Evaluate Active Heart Rate evaluate_active_heart_rate

  • Summary: Evaluate the maximum heart of the last 60 seconds, the observed maximum during this exercise is recorded as the active heart rate.
  • Inputs: None (assumes baseline reading)
  • Outputs: Max HR: int, e.g. {"max_hr": 100}

• Tool: Get Heart Rate Bucket get_heart_rate_bucket

  • Summary: Get the heart rate bucket of the last 10 seconds, the bucket size is 1 second.
  • Inputs:
    • since_from: float, the start time of the monitoring, default is 10 seconds ago
    • bucket_size: float, the size of the bucket, default is 1 second
  • Outputs: Heart Rate Bucket: list[dict], e.g. [{"time": 1715904000, "value": 60}, {"time": 1715904001, "value": 61}]

• Tool: Build Heart Rate Chart build_heart_rate_chart

  • Summary: Build the heart rate chart of the last 600 seconds, the bucket is a dynamic size, the default size is 1 second. The max bucket count is 60, if the bucket count is more than 60, the bucket size will be increased to duration / 60 seconds.
  • Inputs:
    • since_from: float, the start time of the monitoring, default is 600 seconds ago
  • Outputs: Heart Rate Chart PNG URL: str, e.g. https://example.com/chart.png

MCP Settings

Here is the MCP settings for the server:

{
  "mcpServers": {
    "blue-hrm": {
      "command": "uvx",
      "args": [
        "ble-hrm-server"
      ]
    }
  }
}

Here is a sample prompt for the LLM tested by DeepSeek V3.1:

你是一位健康专家，通过心率来检测测试者的心脏健康程度，现在测试者携带了蓝牙的心率带，

以下为检测步骤
- 你通过发现来获取的设备ID,通过rssi获取最接近的设备，再以此设备开始监控。
- 获取当前测试者的静息心率。
- 当测试结束后，提示测试者开始中高强度运动，比如波比跳，当测试者完成运动后，回复后，立刻测试其最高运动心率。
- 之后提示测试者计时休息2分钟，得到确认后，再次获取当前心率，并与最高心率的差额作为恢复心率。 

最后通过恢复心率的值和以下列表评估测试者的心脏健康程度，并用健康专家的口吻给测试者一些关于心脏健康方面的建议。
------
- <22: 身体年龄略大于 真实年龄
- 22～52: 身体年龄= 真实年龄 
- 53～58: 身体年龄略小于真实年龄 
- 59～65: 身体年龄比真实年龄小 
- >=66: 身体年龄比真实年龄小很多

Code Coverage

Sponsor

Special thanks for Qiniu Cloud for providing the storage service. Qiniu also provides LLM inference services.