Laiki
Self-service data science. 100% open source.
Quid?
In Greece a Laiki (λαϊκή αγορά) is a traditional people's market where farmers sell fresh produce directly to people on the street. Similarly Laiki is a platform data scientists can use to bring their work directly to end users while respecting best practices of infrastructure as code. Rolling out reports and data science apps is one Git commit away.
Components
Laiki is a platform that is built on top of time-tested open source products:
| Component | Product | Options |
|---|---|---|
| Workbench | ShinyProxy | VS Code, RStudio, Jupyter notebooks... |
| App publication | ShinyProxy | Shiny, Streamlit, Dash, Gradio, ... |
| Report publication | Crane | HTML, PDF, PowerPoint ... |
| Package management | RDepot | in-house R or Python packages, CRAN / BioC mirror, PyPi ... |
It comes with a Laiki operator that
- manages seamless integration of these products
- adds self-service operations for smooth and compliant data science processes
Design principles
-
make the data scientists life easy:
- abstract away all underlying infrastructure
- fully self-service (no operational reliance on infra teams, no tickets to open)
- declarative approach
- Git based and no 'ClickOps'
- technology agnostic (Python, R, Julia, etc.)
- easy sharing of artifacts (packages, libraries, apps, reports)
-
make the infrastructure engineer's life pleasant:
- infrastructure as code
- managed through GitOps
- scalable to internet scale
- cloud vendor agnostic
- batteries included with (Kubernetes) operators
- no license bureaucracy (fully open source)
- integrated metrics, monitoring and logging stack
-
take into account use in regulated industries (e.g. pharmaceuticals, finance)
- audit trails
- advanced security
- full control
- full transparency
- high test coverage
- commercial support by Open Analytics