Skip to content

tsdive

Start from your question

question run read
Is my data trustworthy? tsdive profile Ingest a PI, IP.21 or OPC export, then reading the profile
Did the process change? tsdive screen, spc, compare Choose a clean baseline window, then compare's three tables
Did my change work? tsdive switchback plan and analyze Test a controller tuning

New to tsdive? Getting started takes you from pip install to a profile of your own CSV export in about 15 minutes, with demo data that tsdive demo writes.

What it is

tsdive is a Python library and command-line tool that checks process time series (historian and sensor archives) for gaps, bad quality codes, clipped ranges, timestamp faults and unit mismatches, then runs baseline, control-chart and PCA monitoring on the same window. It reads single-tag parquet archives, built from CSV or parquet exports with ingest or through a source adapter (docs/SOURCES.md), and writes text, JSON, Python objects or a static HTML page. A check with no answer raises a typed error naming the check. It also runs randomized two-setting trials on one unit (switchback).

Install

tsdive needs Python 3.12 or newer. Install it, then write the demo archives the examples read into data/:

pip install "git+https://github.com/NorthernLightx/tsdive.git"
tsdive demo data

Guides and reference: https://northernlightx.github.io/tsdive/. Wheels are on the releases page. Add [ml] to the install for the IsolationForest detector (scikit-learn) or [mcp] for the MCP server. The package is not on PyPI.

The library and CLI make no network calls. The dataset fetch scripts under scripts/ download on request and print their size first.

Contents