ml-cube / ml3-platform-docs

Documentation repository for ML cube Platform. It contains informations about the Product in general and the SDK to use it.

0 stars 0 forks source link

Scrittura documentazione #5

Open alelavml3 opened 1 week ago

alelavml3 commented 1 week ago

Chapter 1 Company and subscriptions

[ ] COMPLETATO
What is a Company?
Which info are required?
How to add users in a Company?
What is a subscription?
What are the subscriptions feature modules?

Chapter 2 Project

[ ] COMPLETATO
What is a Project?
What entities and modules are at Project level? (kpi, integrations, user roles)
What is storage policy?
What are demo projects?

Chapter 2.1 KPI monitoring

What is a KPI?
Why are KPIs at Project level?
Why are they monitored?

Chapter 2.2 Integrations

What is an integration?
When are they used?

Chapter 2.3 Job status

What is a job?
How can I monitor a job?

Chapter 3 Task

[ ] COMPLETATO
What is a Task?
What is a Task type? Explain each attribute specific for Task Type
What is Data Structure?
When should I use optional target?
What are cost info?

Chapter 3.1 Data Schema

What is data schema?
What are mandatory columns?
What are Role and Subrole?
What are mandatory attributes for each data type?
Which are optional columns? (like additional embedding)
How can I create a data schema?

Chapter 3.2 Task Home

Description of every card in the home

Chapter 3 Models

[ ] COMPLETATO
What is a Model?
What is additional probabilistic output?
What is metric?
What is suggestion type?
What is retraining cost?
What is retraining trigger?

Chapter 4 Data

[ ] COMPLETATO
How are data represented in Platform? (input, target, prediction)
Explain historical, reference, production
How to upload data? Explain SDK concepts like Data, DataSource and all the attributes
What is production boundary?
What is add_target call?

Chapter 4.1 Data Batches

What is a data batch?
What are consistency rules inside a data batch?

Chapter 5 Monitoring

[ ] COMPLETATO
What is Monitoring module?
What is a monitoring target?
What is a monitoring metric?
How is monitoring done? set reference, add production

Chapter 5.1 Detection page

what is a detection event?
what is severity?
explain different types

Chapter 5.2 Automation rules

what is an automation rule?
what are the conditions?
what are the actions?

Chapter 5.3 Drift explainability

what is an explainability report?
how and when is it computed?

Chapter 6 Retraining

[ ] COMPLETATO
what is retraining report?
how is it computed? new reference as last concept
explain every card in the report
explain how to get the suggestion via sdk
explain suggestion types

Chapter 7 RAG Evaluation report

[ ] COMPLETATO
what is rag evaluation?
what are the computed metrics?
what are the required data?

Chapter 8 Topic modelling

[ ] COMPLETATO
what is topic modelling?
when can I create a report?
what is a topic?

Chapter 9 Analytics

[ ] COMPLETATO
what are timeseries?
what are plots?
what are representations?
what is a view?
what is a configuration?

Chapter 10 RBAC

[ ] COMPLETATO
what is a role?
role tables

Chapter 11 Integrations setup

[ ] COMPLETATO
creating credentials
retrain triggers

Chapter 12 API

[ ] COMPLETATO
Python SDK
Notebook Examples

GiovanniGiacometti commented 1 day ago

Dovremmo trovare un modo (banalmente anche un excel) dove segnare tutte le cose della documentazione "facilmente soggette a modifiche", così che chi sviluppa una nuova feature possa andare a controllare la lista e verificare se debba anche aggiornare la documentazione (senza lasciare il fardello a chi effettivamente fa il deploy).

Per cose "facilmente soggette a modifiche" intendo cose come screenshot di pagine della platform, lista di monitoring targets e metriche, lista di possibili plot da fare in analytics (quindi non cose come la descrizione del monitoring).