FlyModel

Overview

Flymodel is a Machine Learning model Version Control & Content Management System with an emphasis on immutability and state validation. It is built in rust, and offers type-validated client libraries in Python and Node.

Features

• Artifact storage and retrieval via s3-compatible services
• Metadata storage via standard SQL
  • Postgresql
  • Sqlite / Memory
  • Mariadb
• Graphql query layer

Development

Pre-Requisites

1. Cargo / Rust
2. Docker
3. Hurl
4. Tasks (opt)

Dev Session

One-time / on changes

Regarding test data, there are two seed envionment cases which must be satisfied:

• single region / basic
• multi region / multi_region

Supplement the below $TEST_DATA with the desired test case

Note

You can run task reset and skip the below steps

1. Start background services

docker-compose up

2. Setup database with test data

cargo migrate-up --test-data $TEST_DATA

Serve

cargo serve

There will now be a local api & graphql service running on http://localhost:9009.