DataJunction Python Client
This is a short introduction into the Python version of the DataJunction (DJ) client. For a full comprehensive intro into the DJ functionality please check out datajunction.io.
Installation
To install:
pip install datajunction
Intro
We have three top level client classes that help you choose the right path for your DataJunction actions.
DJClientfor basic read only access to metrics, dimensions, SQL and data.DJBuilderfor those who would like to modify their DJ data model, build new nodes and/or modify the existing ones.DJAdminfor the administrators of the system to define the connections to your data catalog and engines.
DJ Client : Basic Access
Here you can see how to access and use the most common DataJunction features.
Examples
To initialize the client:
from datajunction import DJClient
dj = DJClient("http://localhost:8000")
NOTE If you are running in our demo docker environment please change the above URL to "http://dj:8000".
You are now connected to your DJ service and you can start looking around. Let's see what namespaces we have in the system:
dj.list_namespaces()
['default']
Next let's see what metrics and dimensions exist in the default namespace:
dj.list_metrics(namespace="default")
['default.num_repair_orders',
'default.avg_repair_price',
'default.total_repair_cost',
'default.avg_length_of_employment',
'default.total_repair_order_discounts',
'default.avg_repair_order_discounts',
'default.avg_time_to_dispatch']
dj.list_dimensions(namespace="default")
['default.date_dim',
'default.repair_order',
'default.contractor',
'default.hard_hat',
'default.local_hard_hats',
'default.us_state',
'default.dispatcher',
'default.municipality_dim']
Now let's pick two metrics and see what dimensions they have in common:
dj.common_dimensions(
metrics=["default.num_repair_orders", "default.total_repair_order_discounts"],
name_only=True
)
['default.dispatcher.company_name',
'default.dispatcher.dispatcher_id',
'default.dispatcher.phone',
'default.hard_hat.address',
'default.hard_hat.birth_date',
'default.hard_hat.city',
...
And finally let's ask DJ to show us some data for these metrics and some dimensions:
dj.data(
metrics=["default.num_repair_orders", "default.total_repair_order_discounts"],
dimensions=["default.hard_hat.city"]
)
| default_DOT_num_repair_orders | default_DOT_total_repair_order_discounts | city |
| ----------------------------- | ---------------------------------------- | ----------- |
| 4 | 5475.110138 | Jersey City |
| 3 | 11483.300049 | Billerica |
| 5 | 6725.170074 | Southgate |
...
Reference
List of all available DJ client methods:
-
DJClient:
list
-
list_namespaces( prefix: Optional[str])
-
list_dimensions( namespace: Optional[str])
-
list_metrics( namespace: Optional[str])
-
list_cubes( namespace: Optional[str])
-
list_sources( namespace: Optional[str])
-
list_transforms( namespace: Optional[str])
-
list_nodes( namespace: Optional[str], type_: Optional[NodeType])
-
list_nodes_with_tags( tag_names: List[str], node_type: Optional[NodeType])
-
list_catalogs()
-
list_engines()
find
- common_dimensions( metrics: List[str], name_only: bool = False)
- common_metrics( dimensions: List[str], name_only: bool = False)
execute
- sql( metrics: List[str], dimensions: Optional[List[str]], filters: Optional[List[str]], engine_name: Optional[str], engine_version: Optional[str])
- node_sql( node_name: str, dimensions: Optional[List[str]], filters: Optional[List[str]], engine_name: Optional[str], engine_version: Optional[str])
- data( metrics: List[str], dimensions: Optional[List[str]], filters: Optional[List[str]], engine_name: Optional[str], engine_version: Optional[str], async_: bool = True)
- node_data( node_name: str, dimensions: Optional[List[str]], filters: Optional[List[str]], engine_name: Optional[str], engine_version: Optional[str], async_: bool = True)
-
DJ Builder : Data Modelling
In this section we'll show you few examples to modify the DJ data model and its nodes.
Start Here
To initialize the DJ builder:
from datajunction import DJBuilder
djbuilder = DJBuilder("http://localhost:8000")
NOTE If you are running in our demo docker container please change the above URL to "http://dj:8000".
Namespaces
To access a namespace or check if it exists you can use the same simple call:
djbuilder.namespace("default")
Namespace(dj_client=..., namespace='default')
djbuilder.namespace("foo")
[DJClientException]: Namespace `foo` does not exist.
To create a namespace:
djbuilder.create_namespace("foo")
Namespace(dj_client=..., namespace='foo')
To delete (or restore) a namespace:
djbuilder.delete_namespace("foo")
djbuilder.restore_namespace("foo")
NOTE:
The cascade parameter in both of above methods allows for cascading
effect applied to all underlying nodes and namespaces. Use it with caution!
Tags
You can read existing tags as well as create new ones.
djbuilder.tag(name="deprecated", description="This node has been deprecated.", tag_type="standard", tag_metadata={"contact": "Foo Bar"})
Tag(dj_client=..., name='deprecated', description='This node has been deprecated.', tag_type='standard', tag_metadata={"contact": "Foo Bar"})
djbuilder.tag("official")
[DJClientException]: Tag `official` does not exist.
To create a tag:
djbuilder.create_tag(name="deprecated", description="This node has been deprecated.", tag_type="standard", tag_metadata={"contact": "Foo Bar"})
Tag(dj_client=..., name="deprecated", description="This node has been deprecated.", tag_type="standard", tag_metadata={"contact": "Foo Bar"})
To add a tag to a node:
repair_orders = djbuilder.source("default.repair_orders")
repair_orders.tags.append(djbuilder.tag("deprecated"))
repair_orders.save()
And to list the node names with a specific tag (or set of tags):
djbuilder.list_nodes_with_tags(tag_names=["deprecated"]) # works with DJClient() as well
["default.repair_orders"]
Nodes
To learn what Node means in the context of DJ, please check out this datajuntion.io page.
To list all (or some) nodes in the system you can use the list_<node-type>() methods described
in the DJ Client : Basic Access section or you can use the namespace based method:
All nodes for a given namespace can be found with:
djbuilder.namespace("default").nodes()
Specific node types can be retrieved with:
djbuilder.namespace("default").sources()
djbuilder.namespace("default").dimensions()
djbuilder.namespace("default").metrics()
djbuilder.namespace("default").transforms()
djbuilder.namespace("default").cubes()
To create a source node:
repair_orders = djbuilder.create_source(
name="repair_orders",
display_name="Repair Orders",
description="Repair orders",
catalog="dj",
schema_="roads",
table="repair_orders",
)
Nodes can also be created in draft mode:
repair_orders = djbuilder.create_source(
...,
mode=NodeMode.DRAFT
)
To create a dimension node:
repair_order = djbuilder.create_dimension(
name="default.repair_order_dim",
query="""
SELECT
repair_order_id,
municipality_id,
hard_hat_id,
dispatcher_id
FROM default.repair_orders
""",
description="Repair order dimension",
primary_key=["repair_order_id"],
)
To create a transform node:
large_revenue_payments_only = djbuilder.create_transform(
name="default.large_revenue_payments_only",
query="""
SELECT
payment_id,
payment_amount,
customer_id,
account_type
FROM default.revenue
WHERE payment_amount > 1000000
""",
description="Only large revenue payments",
)
To create a metric:
num_repair_orders = djbuilder.create_metric(
name="default.num_repair_orders",
query="""
SELECT
count(repair_order_id)
FROM repair_orders
""",
description="Number of repair orders",
)
Reference
List of all available DJ builder methods:
-
DJBuilder:
namespaces
- namespace( namespace: str)
- create_namespace( namespace: str)
- delete_namespace(self, namespace: str, cascade: bool = False)
- restore_namespace(self, namespace: str, cascade: bool = False)
nodes
- delete_node(self, node_name: str)
- restore_node(self, node_name: str)
nodes: source
- source(self, node_name: str)
- create_source( ..., mode: Optional[NodeMode] = NodeMode.PUBLISHED)
- register_table( catalog: str, schema: str, table: str)
- register_view( catalog: str, schema: str, view: str, query: str, replace: bool = False)
nodes: transform
- transform(self, node_name: str)
- create_transform( ..., mode: Optional[NodeMode] = NodeMode.PUBLISHED)
nodes: dimension
- dimension(self, node_name: str)
- create_dimension( ..., mode: Optional[NodeMode] = NodeMode.PUBLISHED)
nodes: metric
- metric(self, node_name: str)
- create_metric( ..., mode: Optional[NodeMode] = NodeMode.PUBLISHED)
nodes: cube
- cube(self, node_name: str)
- create_cube( ..., mode: Optional[NodeMode] = NodeMode.PUBLISHED)
DJ System Administration
In this section we'll describe how to manage your catalog and engines.
Start Here
To initialize the DJ admin:
from datajunction import DJAdmin
djadmin = DJAdmin("http://localhost:8000")
NOTE If you are running in our demo docker container please change the above URL to "http://dj:8000".
Examples
To list available catalogs:
djadmin.list_catalogs()
['warehouse']
To list available engines:
djadmin.list_engines()
[{'name': 'duckdb', 'version': '0.7.1'}]
To create a catalog:
djadmin.add_catalog(name="my-new-catalog")
To create a new engine:
djadmin.add_engine(
name="Spark",
version="3.2.1",
uri="http:/foo",
dialect="spark"
)
To linke an engine to a catalog:
djadmin.link_engine_to_catalog(
engine="Spark", version="3.2.1", catalog="my-new-catalog"
)
Reference
List of all available DJ builder methods:
-
DJAdmin:
Catalogs
- list_catalogs() # in DJClient
- get_catalog( name: str)
- add_catalog( name: str)
Engines
- list_engines() # in DJClient
- get_engine( name: str)
- add_engine( name: str,version: str, uri: Optional[str], dialect: Optional[str])
Together
- link_engine_to_catalog( engine_name: str, engine_version: str, catalog: str)
Claude Code Integration
DataJunction provides comprehensive Claude Code integration through two components:
- MCP Tools - Live connectivity to your DJ instance for querying metrics, discovering dimensions, and visualizing data
- Skill - Passive knowledge about DataJunction concepts, patterns, and workflows
Both components are bundled with the Python client and can be installed with a single command.
What's Included
MCP Tools provide:
- Query metrics and generate SQL
- Discover available metrics and dimensions
- Find common dimensions across metrics
- Visualize data with inline charts
The DataJunction skill provides:
- Core concepts - Star schema, dimension links, node types, and DJ fundamentals
- Building the semantic layer - Creating metrics, dimensions, cubes, and dimension links
- Repo-backed workflow - YAML node definitions, git workflow, and branch-based development
Installation
As a Claude Code plugin
If you use Claude Code, install the DJ plugin — it bundles the skills, the MCP server config, and the DJ subagent, and Claude Code manages it for you:
/plugin marketplace add DataJunction/dj
/plugin install datajunction@datajunction
The MCP tools run through the dj-mcp command from this package, so also
pip install datajunction[mcp] and set DJ_API_URL to point at your instance
if you want Claude to query live data.
With the DJ CLI
To copy the skills into your home directory and configure Claude Code (also the path for Claude Desktop):
dj setup-claude
This will:
- Copy the bundled skill to
~/.claude/skills/datajunction/ - Configure the DJ MCP server in your Claude config
- Make DataJunction expertise available to Claude in all your conversations
Options:
# Install only the skill (skip MCP server setup)
dj setup-claude --no-mcp
# Install only the MCP server (skip skill installation)
dj setup-claude --no-skills
After installation, restart Claude Code to load the changes.
Usage
Once installed, Claude Code will automatically use both MCP tools and the skill for DataJunction tasks:
MCP tools in action:
- "Show me the revenue metric" → Queries your live DJ instance
- "What dimensions are available for these metrics?" → Discovers common dimensions
- "Visualize revenue by city" → Generates and displays inline charts
Skill in action:
- "How do dimension links work in DataJunction?" → Explains concepts
- "How do I create a metric in YAML?" → Shows YAML examples and patterns
- "Explain the repo-backed workflow" → Details git-based development
The MCP tools provide live data access while the skill provides conceptual knowledge and best practices.
Customizing for Your Organization
If you need organization-specific skill content:
Option A: Fork the client
- Fork
datajunction-clients/python - Modify
datajunction/skills/datajunction.mdwith your custom content - Publish your custom client package
Option B: Override after install
- Run
dj setup-claudeto get the base skill - Manually edit
~/.claude/skills/datajunction/SKILL.mdwith your customizations
Skill Location
The skill file is bundled at:
datajunction/skills/datajunction.md
And installed to:
~/.claude/skills/datajunction/SKILL.md
Release files for datajunction 0.0.229
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| datajunction-0.0.229.tar.gz | 152.9 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| datajunction-0.0.229-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 260.6 kB
Release files / datajunction-0.0.229.tar.gz
| Download URL | datajunction-0.0.229.tar.gz |
|---|---|
| Size | 152.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
90cd0a52ac5a0781577ad0a01e2e1c67cf65b0c8e8941522682e781792dfc6b7
|
|
BLAKE2b-256 checksum How to use checksums |
c2c6e3ed326b200349818c4886151085527c872833dd26a3ee263b7aac28880b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 2, 2026.
Transparency logRelease files / datajunction-0.0.229-py3-none-any.whl
| Download URL | datajunction-0.0.229-py3-none-any.whl |
|---|---|
| Size | 107.7 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
99dd6c00e1f1d3674680f9a974d767408f96a1501aa6631b13dcf9841f355412
|
|
BLAKE2b-256 checksum How to use checksums |
fe8b91933478b24aea5cd2a23f9a8416b8e2325d54011d84357bbfb7395ba7f8
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Sep 2, 2026.
Transparency log
