This repository contains all components that make up the CMS DQM-DC DIALS service.

During a Run, the CMS Experiment collects particle collision data in Lumisection time-frame and the experiment subsystem health is monitored by shifters using DQMGUI. Multiple shifters for many subsystems monitor the most recent lumisection of the current run plots to check a subsystem's health, but given the number of plots and the limited manpower it is not possible to closely pay attention to all plots during the lumisection time-frame. On the other hand, using certified data it is possible to extract knowledge from old runs (at lumisection-level) to help shifters get a glimpse of detector's health on past runs.

Data Inspector for Anomalous Lumi-Sections (DIALS) is an application designed to be an access point to DQMIO per-LS monitoring elements. It is firstly responsible for indexing, storing pre-processed DQMIO data and serving via a WEB UI and REST Api, so that it could be used by any CMS sub-group for exploratory analysis, statistical learning and machine learning. Secondly we plan to expanding its capabilities adding Machine Learning HUB-like abilities (storing, training and evaluating models).

This application was born in CMS Tracker ML as a prototype under the name "MLPlayground" and was migrated to CMS DQM-DC as an effort of centralizing anomaly detection techniques for every subsystem.

This project is a full-stack application written in Python (^3.10) and JavaScript (Node.js ^20.11.0). The backend is built using Django (specifically Django Rest Framework), the job queue is managed by Celery and the frontend is built with plain React. It uses a PostgreSQL database instance for storing all data and Redis in-memory database acting as the message broker for the job queue.

• Backend
• Frontend
• Authentication
• Deployment

Check development planning here and TODO tracker here.

Check development guidelines here and local development instructions here.