Explore DAGs

A Matchbox collection can store several runs of a DAG. Each run is a server-side snapshot of the sources, models, and resolvers that define one entity view for that collection.

This guide shows how to list collections, download a stored run, and inspect the pipeline it contains.

Listing collections

Example

from matchbox.client.dags import DAG

DAG.list_all()

You can also do lightweight exploration from the CLI.

CLI

mbx collections

Run mbx collections --help to inspect the available collection commands and options.

Downloading a DAG

Load the default or pending run for a collection.

load_default() reconstructs the published run for that collection. This is the run other users and services query by default.

Default run

from matchbox.client.dags import DAG

dag = DAG(name="companies").load_default()

load_pending() reconstructs the latest non-default run, which is usually the draft currently being worked on before publication.

Pending run

from matchbox.client.dags import DAG

dag = DAG(name="companies").load_pending()

The downloaded DAG includes the serialisable definitions of every source, model, and resolver in that run.

Inspecting the pipeline

Use draw() to inspect the dependency graph.

Tree viewExecution order

print(dag.draw())

print(dag.draw(mode="list"))
print(dag.sequence)

The default resolver is the single final resolver in a complete published DAG. It’s used in functions like DAG.get_matches() if no resolver is supplied. To make a DAG run the default, a final_resolver must be present.

Inspecting individual steps

You can retrieve sources, models, and resolvers by name.

Example

source = dag.get_source("companies_house")
model = dag.get_model("dedupe_companies_house")
resolver = dag.get_resolver("companies_resolver")

Once you have the DAG locally, you can attach warehouse clients, query resolved entities, or start a new run from the same structure.