Skip to content

elcolumbio/mlrepricer

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Status

This has a character of an MVP pet project.

Things to do:

  • introduce pydantic for type checking and metadata
  • get a basic reinforcement algorithm to work

mlrepricer

Pricing on platforms is essential. Most of the technical parts close to the platform are similiar for merchants. That's why we like to introduce a library for all merchants.

What is the philosophy behind pricing?

There is a rich world of concepts, strategies and tactics about how to set a price at any given point in time. While we are open to all of them. We believe that it is not a zero sum game for all buyers and sellers.

We want to share knowledge, provide the data infrastracture where you can deploy your strategies on top. There is nothing which couldnt help us to move forward.

We provide you some modules you can use in your existing infrastructure. It should be very easy to do so.

Lately we switched to redis. We still have schema validations for sql, need some work to integrate the missing parts.

This is not a repricer. It should be used for statistical and ml analysis of your data. It would be nice if we share statistic analysis and discuss them and validate them. The data on it's own is not valueable.

Where are we?

Right now you can use existing simple rules and easy add your own rules. There is no framework for a multi agent problem yet. Right now it is only build for redis, you should be able to replace it in most parts easy. How long do you need for setup for the first time?

Where are we going?

There won't be a service ever. You can see that by how flexible we build it, for datastorage, modularity. We rather want to take some of the heavy lifting when you handle price data. That's not only mws related. I like to start implementing some eBay prices too. Also we should be able to validate flexible metadata from your ERP. Really our strength is to offer building blocks for things everyone would spend hours to solve problems. Where hundreds of people already solved this. Instead you maybe want make this framework better:)

price volatility of buyboxwinners

volatility

How to get started?

For more uptodate guides please check out the notebooks.

Setup Boto3 credentials for aws sqs: ~/.aws/credentials:

aws_access_key_id = YOUR_ACCESS_KEY
aws_secret_access_key = YOUR_SECRET_KEY

We define all configs in a yaml file, if you fill out everything you are fine. So copy mlrepricer/mlrepricer/configs.yaml to the default location ~/.config/mlrepricer/configs.yaml . Or change the path_to_config in the setup package.

Installation

It should work on Linux and Windows.

Database choice: We should make it easy to use whatever you want. For now i go with redis (version '3.0.1' +), on windows i run it manually under WSL.

Package: You can clone it locally and cd into the top level folder. and install it with: pip install .

Dependencies are in requirements.txt, except of mws i use: I use my version of mws: https://github.com/innotrade24/python-amazon-mws1

Usage

See the notebooks. Download this repo and unpack it, or render them in github.

Contribution

Everyone is welcome to contribute. A tipp for cleaning all notebooks: before pushing your code, in the commandline cd into jupyter and run this command:

mlrepricer_private\jupyter> jupyter nbconvert --ClearOutputPreprocessor.enabled=True --inplace **.ipynb