Skip to content


What is pipestat?

Pipestat standardizes reporting of pipeline results. It provides 1) a standard specification for how pipeline outputs should be stored; and 2) an implementation to easily write results to that format from within Python or from the command line.

How does pipestat work?

A pipeline author defines all the outputs produced by a pipeline by writing a JSON-schema. The pipeline then uses pipestat to report pipeline outputs as the pipeline runs, either via the Python API or command line interface. The user configures results to be stored either in a YAML-formatted file or a PostgreSQL database. The results are recorded according to the pipestat specification, in a standard, pipeline-agnostic way. This way, downstream software can use this specification to create universal tools for analyzing, monitoring, and visualizing pipeline results that will work with any pipeline or workflow.

Installing pipestat

Minimal install for file backend

Install pipestat from PyPI with pip:

pip install pipestat

Confirm installation by calling pipestat -h on the command line. If the pipestat executable is not in your $PATH, append this to your .bashrc or .profile (or .bash_profile on macOS):

export PATH=~/.local/bin:$PATH

Optional dependencies for database backend

Pipestat can use either a file or a database as the backend for recording results. The default installation only provides file backend. To install dependencies required for the database backend:

pip install pipestat['dbbackend']

Optional dependencies for pipestat reader

To install dependencies for the included pipestatreader submodule:

pip install pipestat['pipestatreader']

Set environment variables

export PIPESTAT_RESULTS_SCHEMA=output_schema.yaml
export PIPESTAT_RESULTS_FILE=results_file.yaml

When setting environment variables like this, you will need to provide an output_schema.yaml file in your current working directory with the following example data:

title: An example Pipestat output schema
description: A pipeline using pipestat to report sample and project results.
type: object
  pipeline_name: "default_pipeline_name"
    type: object
          type: string
          description: "ResultName"

Pipeline results reporting and retrieval

These examples assume the above environment variables are set.

Command-line usage

# Report a result:
pipestat report -i result_name -v 1.1

# Retrieve the result:
pipestat retrieve -r my_record

Python usage

import pipestat

# Report a result
psm = pipestat.PipestatManager(){"result_name": 1.1})

# Retrieve a result
psm = pipestat.PipestatManager()

Pipeline status management

From command line:

# Set status
pipestat status set running

# Get status
pipestat status get

Python usage

import pipestat

# Set status
psm = pipestat.PipestatManager()

# Get status
psm = pipestat.PipestatManager()