Library for data quality process

nefertem is an exetensible framework for monitoring and managing data quality processes. With nefertem you can define your own data quality process, run them and get the results. You can also create specific plugins that enable the use of your favourite data quality frameworks.

Overview

nefertem adopt a run execution model. A user defines an execution run with a client that handles also the I/O storages. Every run is executed under an experiment, an organizational unit.

Running nefertem produces in-memory objects, deriving from the execution frameworks plugged-in (e.g. frictionless, ydata_profiling, etc.), a bunch of process descriptive metadata and a series of artifacts that can be persisted on various backend storage.

The typical workflow involves the configuration of the resources, of the input storages in which the resources are saved (local or remote filesystems, databases and datalakes) and the configuration of the run itself, where the user specifies the desired operations and the frameworks to be used.

Out-of-the-box nefertem supports the following data quality operation:

• Validation
• Inference
• Profiling
• Metrics

Example

import nefertem

# Set configurations
output_path = "./nefertem_run"
store = {"name": "local", "store_type": "local"}
data_resource = {
    "name": "resource_name",
    "path": "path/to/resource",
    "store": "local",
}
run_config = {
    "operation": "inference",
    "exec_config": [{"framework": "frictionless"}]
}

# Create a client and run
client = nefertem.create_client(output_path=output_path, store=[store])
with client.create_run([data_resource], run_config) as nt_run:
    nt_run.infer()
    nt_run.log_schema()
    nt_run.persist_schema()

Documentation

• Installation and requirements
• Configuration
• Modules