lblend / spuristo

A CLI application for tracking the number of people in the vicinity of a device using bluetooth

spuristo

spuristo is a CLI application that utilizes bluetooth and linear regression to guess the number of people in the vicinity of a device.

This project has a lot of work that remains to be done at the time of writing this.

Get started

todo™️

Options

--adapter		Specify the name of the bluetooth adapter that will be used.
			Default is set to "hci0". Do not touch if you don't know what you're doing.

--threshold		Specify the minimum connection strength in dBm required to "count" a device.
			Default is set to -75 dBm

--training		Run the application in training mode. This will enter all datapoints into a
			separate database table used for training the model when running in production.

A note about CLI arguments

There's a known issue about the --threshold argument where inputing negative numbers are not properly supported. In order to mitigate this, you need to use the following syntax

spuristo --threshold=-75

This is detailed in issue #3

Contribute

This project is meant to be a fun hobby project allowing me to solve the problem of not knowing how many people are at the place I hang out at school, while also giving me an excuse to learn Rust. Therefore I wish to keep contributions to a minimal in order to force myself to learn and figure out the ways of Rust. The exceptions to this are issues marked with the Help wanted tag.

spuristo's Issues

Create independent output sections/widgets

Inspired by htop, pin the number of connected devices to its own section of the stdout while outputting the event log in it's own section

tui-rs can be used to achieve this

add logging

Add ability to sepcify API URI

Specify request URL to the API for inserting and fetching datapoints

add ability to save trained classifiers between runs

Installation documentation

add a "no threshold" flag

create an option to run without any signal threshold

Create docker-compose for spuristo + spuristo-api

implement training data collection mode

negative numbered CLI args cannot be parsed properly

since - is a reserved character for specifying args, the program cannot parse the threshold CLI arg properly when given a negative number

The way to mitigate this issue is to use the --argname=argvalue syntax, i.e --threshold=-75

There should be an instruction telling you to use this syntax or be implemented a solution that parses it properly when used in this manner --threshold -75

create linear regression classifier

Signal strength threshold not being properly enforced

The way the signal strength threshold is enforced may be problematic and cause the application to count the wrong number of devices.

This will happen all of the following happens:

1. A device enters the vicinity of the threshold and triggers a DeviceAdded event
2. The device goes outside the threshold but stays within the vicinity of the host device, not causing a DeviceRemoved event.

This is a critical bug because it may cause the application to gather bad data during training.

The bluer library which is used doesn't seem to provide any way to continuously check the signal strength of connected devices. Therefore it is likely that this library will need to be replaced. I don't know any other way of fixing it.