Skip to content

Latest commit

 

History

History
137 lines (98 loc) · 7.93 KB

README.md

File metadata and controls

137 lines (98 loc) · 7.93 KB

Project description

This project was a part of broader demo. That broader demo analyzed customers movement in a retail store, determined their behaviour (for example: "customer stopped in men's clothes department") and use Machine Learning to model for purchase/product recommendation. The customer location was determined by movement sensors placed in the store.

This service customer behaviour in a retail shop:

  • customer entering the store
  • customer movement
  • customer exiting the store
  • by generating proper MQTT messages.

In this case, we are using a big store as an example, however, for the demo purposes, we predict customers' behaviors only in 5 areas: Women's, Men's, Boy's, Girl's, and Sport.

Functionality

This application visualize the store and customers movement in it. The app also allows creating new scenarios and sending them to the scenario-player.

This is the UI for the demo described above.

This application uses a plan which is 1808px wide and 1315px high. In the plan below there are main areas borders (limit points) described. There are also the entrance and exit points shown.

alt text

Table of contents

Usage

This is a web service (implemented with FastAPI). By default, it works on port 8000. (See instructions for details on configuring and running the service.)

When starting, it does the following:

  • connects to MQTT server
  • exposes http endpoints with the UI
  • exposes REST API (HTTP + WS) for the UI
  • waits for, registers, and pass to the UI users movement

Dependencies

Dependencies of the project are contained in requirements.txt file. All the packages are publicly available.

All the packages can be installed with: pip install -f requirements.txt

This application assumes running MQTT broker.

Service configuration

The service reads the following environment variables:

Variable Description Default
MQTT_HOST MQTT broker host -
MQTT_PORT MQTT broker port 1883
MQTT_USERNAME username for MQTT authentication
MQTT_PASSWORD password for MQTT authentication
MQTT_CLIENT_ID MQTT client ID prefix demoVisClient
ENTER_TOPIC customer/enter
MOVE_TOPIC customer/move
EXIT_TOPIC customer/exit
BROWSING_TOPIC customer/browsing
COUPON_PREDICTION_TOPIC customer/prediction
SCENARIO_PLAYER_SCENARIO_ENDPOINT full address (ex: http://localhost:8004/scenario) to the scenario-player's scenario endpoint -
LOG_LEVEL logging level INFO
LOG_FILENAME log file name ''

(Parameters with - in "Default" column are required.)

Use env variables log_config.py to configure logging behaviour. By default, console is used for logging. File handler is added if LOG_FILENAME is provided.

Running the service

Development

.environment.variables.sh can be used for that purpose. Then, in order to run the service the following commands can be used:

$ . .environment.variables.sh
$ . venv/bin/activate
(venv)$ uvicorn app.main:app --host 0.0.0.0 --reload --reload-dir app

Please, note reload-dir switch. Without it the reloader goes into an infinite loop because it detects log file changes (messages.log).

Production

The service is made to be run in an OpenShift environment. The deployment process is described in (infra)[../infra] path.

App interfaces

Endpoint/topic/address Type Description
/ Web (HTTP) The main UI page
/phone/{customer_id} Web (HTTP) Customer's phone simulation
/health HTTP Service Health
/api/new_scenario HTTP Proxy to the scenario-player
/api/customers HTTP Returns list of a few customers for demo purposes
/ws/movement Websocket Passes customers movements thru when received from MQTT topic
/ws/coupons Websocket Passes coupons info thru when recieved from MQTT topic
{CUSTOMER_ENTER_TOPIC} MQTT Recieves enter events from Phone App or scenario-player
{CUSTOMER_EXIT_TOPIC} MQTT Recieves exit events from Phone App or scenario-player
{CUSTOMER_MOVE_TOPIC} MQTT Recieves move events from Phone App or scenario-player

Using the UI

When service is up and running go to the main path of it to see the UI, for example: http://localhost:8000

There are 2 main tabs - "Store preview" and "Create customer scenario".

1st one allows observing customers' movements in the store. On the right-hand side you can switch between the "Mobile app" and "Events log" view. "Mobile app" lets you choosing the customer to observe. After choosing one its color on the plan changes. You can also see the simulation of the client's phone application. "Events log" allows following the messages sent to the application via MQTT and then passed to the UI via Websockets.

2nd tab - "Create customer scenario" allows creating a new scenario for the selected customer. Select the user, then the pin type, and put some pins on the plan. "Movement" pin lets you create points where the customer occurs for a short while. "Focus" points will generate multiple "Movement" points to simulate the client's focus on the product. Next, select Customer step time (how long does it take to move between "movement" points). Now start the simulation.

To do

  • describe the usage of certificates (MQTT_BROKER_CERT_FILE)
  • explain usage of the CUSTOMERS_LIST_FILE, COUPONS_LIST_FILE parameters (or remove them)
  • add BROWSING_TOPIC, COUPON_PREDICTION_TOPIC to "App interfaces" section (and explain its usage)
  • add env variables descriptions in "Service configuration" section
  • update "Production" section to be more specific (possibly, remove OpenShift references as the service can be run in any environment)