ASCII Research space

A single monorepo to store all our research. Every person has a folder

development infrastructure (local setup)

• a computer with internet access to install python packages
• an installed pixi https://pixi.sh/latest/ for your operating system
  • ensure the pixicommand is in your path and working correctly
• an installation of make command in your path
• git
• docker
• an SSH key readily set up to connect to github
• permissions to view this (https://github.com/ascii-supply-networks/research-space) repository
• available infrastructure
  • server
  • openAI API key

git clone git@github.com:ascii-supply-networks/research-space.git
cd research-space   

# Activate environment. If the environment is not installed it will be installed automaticially. 

# Default env:
pixi shell -e research-cpu 
# If you need gpu acceleration use:
pixi shell -e research-cuda
# For the pipeline:
pixi run create_environment_pipeline 

# To start a jupyter notebook session
make notebook

# Set up pre-commit hook for automatic formatting. Execute this in reasearch-space
pixi global install pre-commit
export PATH="$HOME/.pixi/bin:$PATH" # If you are on the server execute this and reload terminal
pre-commit install

notes

adding new reference data

Ensure to set DEV_REFERENCE_DATA_OVERRIDE to the location where the reference data repository is checked out.

Ideally, you download the reference data from https://github.com/ascii-supply-networks/reference-data relative to/next to the research space; however, only by setting a softlink to make it appear there.

using R

• talk to georg

  • enable renv
  • integrate Rstudio

• in a shell: conda activate ascii_rs

• then start an R shell by typing R in the command line window

  • then inside R execute: renv::restore()
  • instally any additionally required packages
  • use them in your scripts. Only the used ones are recorded
  • commit these to version control by renv::snapshot() and then commiting the renv

• when you use the commandline you can use the new environment straight away

updating the main deployment

Ensure the quality is high by testing the code beforehand.

Execute the following linting and testing procedures and fix any issues:

make fmt
make lint
make test

make fmt-sql
make lint-sql

prepare secrets. Decrypt the secrets - you must have a valid key file to perform this step

make secrets-decrypt

Ensure ROOT_DOMAIN is pointing to the right setting.

Test one more time with the containerized pipeline:

make start-pipeline-staging

Run and build the containers and expose via proxy:

make start-pipeline-production

Open the URL to see the dashboard: https://orchestration.ascii.ac.at/

WARNING a click will trigger the pipeline - some pipelines (for commoncrawl) cost 1k€/execution - know what you do.