easydispatch

This is a real time field service dispatching planner. It automatically dispatch jobs to different workers under business restrictions like working hour, skills, etc.

Some of the key feature highlights:

Real Time Dispatching based on three types of Dispatching Algorithms: Heuristic, Optimization, Reinforcement Learning
Vehicle Routing Problem capability embedded
Flexible Business Rule Plugins, specifies what solution is allowed but not how to reach a solution
Explainable AI to do drag and drop what-if analysis.
Full set of REST API and API Documents

Quick Start

planner_ui

Local Installation

EasyDispatch relies on Postgres DB, Redis and Kafka. Those three components can be started by docker and compose or provisioned seperately. You also should have npm and node for frontend development.

To run easydispatch locally, first install it by:

git clone https://github.com/alibaba/easydispatch.git && cd easydispatch
pip install -e .

Then copy and modify env file from $e a s y d i s a p t c h / e t c / d e v_{e} n v_{s} a m p l e t o$ easydisaptch/dev.env . Start the database, redis and kafka by docker composer.

docker-compose -f kafka-redis-postgres-compose.yml -p easy up

Open another terminal, populate some sample data and run the frontend:

python -m dispatch.cli database init
python -m dispatch.cli server start --port 8000 dispatch.main:app

Visit the page at : http://localhost:8000/login

Other Information

Those scripts can be used for testing purpose:

clear off all existing workers and jobs
generate sample data for a period
use batch optimizer to dispatch all jobs in the planning window

cd $EASHDISPATCH_HOME
python  tests/kandbox_clear_data.py 
python -m dispatch.cli util generate --dataset veo --team_code london_t1 --start_day 20210503 --end_day 20210505 --username demo --password demo
python -m dispatch.cli job dispatch  --team_code london_t1  --start_day 20210503 --dispatch_days 2

OS and Environements

We have tested it on Ubuntu 20.04 and MacOS, Python 3.7 / 3.8 / 3.9

Reference

The frontend and server technology stack (vue + python) were adapted from Netflix Dispatch. Data structures are not compatible.