Foundation Deployment Pack

Overview

The Foundation Deployment Pack (dp:foundation) is the recommended starting point for any new Cognite Data Fusion project. Instead of hand-building access controls, a data model, extraction pipelines, and CI/CD from scratch on every engagement, you deploy a clean, best-practice foundation as a Toolkit module and configure it through a guided wizard. The result: a solid, production-ready project scaffold stood up in days, not weeks.

Who is this for & when to use? 

Target personas:

• Data Engineers / Solution Architects — you deploy and configure it on a project.
• Field Engineers — you use it to get a customer environment ready fast and demo the data foundation with customer data in a matter of hours.
• Partners — you stand up a best-practice CDF foundation on your own engagements, the same way Cognite delivery teams do.

Use this pack when:

• Use this pack when you are setting up a new CDF project — a QuickStart or a LAND engagement — and you need a standard, best-practice foundation (access groups, data model, extraction and contextualization pipelines) without building it by hand.

When NOT to use this pack:

• This pack is built for new projects. It is not designed to retrofit an existing project that already has established access groups, a data model, and configuration in place. For brownfield projects, talk to the Accelerator team about the right approach before deploying.

Why you need it?

Without this pack, every new project repeats the same setup work: defining persona access groups, extending a CDM, wiring up extraction pipeline configs, and standing up CI/CD. That work is slow, easy to get subtly wrong, and tends to be re-solved differently on each project — which means inconsistency across deliveries and rework when something doesn't follow best practice.

The Foundation Deployment Pack eliminates that. It ships the proven patterns as reusable, tested components, so teams skip the repetitive scaffolding and spend their time on what the customer actually needs. It cuts foundation setup from weeks to days, keeps deliveries consistent with Cognite best practice, and reduces the risk of misconfigured access or data models reaching a live project.
 

What's included

The pack is a set of Toolkit modules. The common module (cdf_project_foundation) is always included; you then select one data model and any source-system and contextualization modules your project needs.
 

Prerequisites

Verify all of the following before starting deployment:

• Cognite Toolkit version 0.8 or later is installed. Follow the setup instructions.
• A cdf.toml file exists in your project root. If missing, run cdf init and select Create toml file (required).
• Authentication is configured and verified — run cdf auth init and cdf auth verify using a local .env file. See the Toolkit authentication docs.
• Your CDF project has Data Modeling enabled.
• The CDF group used by Toolkit has the following capabilities: spaces:write, datamodels:write, transformations:write, workflows:write, datasets:write, rawdb:write.

Deployment & installation

Full step-by-step deployment, including all wizard options and CI/CD setup, lives in the README. The summary below covers the main path.

a. Initialize the project

In an empty directory:

cdf modules init .

b. Select modules.

Pick exactly one data model variant (ISA-95 or CFIHOS), then any source-system and contextualization modules. Always keep cdf_project_foundation; qualitizer is recommended.

• Choose Data models (data models that extend the core data model).
• Use Space to select 
• Press Enter to confirm.

c. Run the setup wizard.

Run setup_project.py. It prompts for project names, site/location, Entra ID group IDs, and owner contacts, then writes all config.<env>.yaml files and .env in one pass. The wizard is idempotent — re-running pre-fills current values.

d. Build and deploy

cdf build
cdf deploy --dry-run
cdf deploy

e. Set up CI/CD (optional).

Run generate_actions.py to create GitHub Actions workflows for automated build, dry-run, and deploy on PR / merge / release.

Verification

After deploying, confirm success with these checks:

1. In CDF, the three persona groups (consumer, producer, admin, suffixed by environment and optional site) are present and bound to the correct Entra ID source IDs.
2. The selected data model (ISA-95 or CFIHOS spaces) is deployed and visible.
3. Any selected source-system extraction pipeline configs appear in CDF.
4. Re-run the wizard with --check for a CI drift check — it should report no unexpected differences.
5. If qualitizer was installed, its dashboards load and begin reporting ingestion/contextualization coverage

Configuration reference

Parameters are set in config.dev.yaml at the project root. Default values work for an initial test deployment.

Access groups deployed

Group naming follows the SOP: <persona>-[site]-<environment>.

Architecture 

The pack is structured as a common foundation module plus selectable model, source, and contextualization modules. The setup wizard configures all of them in one pass, writing per-environment config files and a .env file.

<project>/
├── .github/                  # CI/CD workflows (optional, generated by the wizard)
├── build/                    # Toolkit build output
├── docs/                     # generated docs (e.g. FOUNDATION_CICD.md)
├── modules/
│   ├── common/               # cdf_project_foundation — persona groups + setup wizard
│   ├── contextualization/    # entity matching, file annotation (optional)
│   ├── datamodels/           # ISA-95 or CFIHOS enterprise model (select one)
│   ├── sourcesystem/         # PI / SAP / OPC-UA / DB / files pipeline configs (optional)
│   └── tools/                # qualitizer and other tooling (optional)
├── .env                      # Entra ID group object IDs (written by the wizard)
├── cdf.toml                  # Toolkit project config
├── config.dev.yaml           # per-environment configuration
└── config.prod.yaml

Data flow: source systems → extraction pipeline configs (PI / SAP / OPC-UA / DB / files) → CDF RAW / instances → data model (ISA-95 or CFIHOS) → optional contextualization (entity matching, file annotation) → consumed via persona-scoped access groups.

In scope: project-level foundation — access groups, data model, source pipeline configs, CI/CD scaffolding. Out of scope: the source-system extractors themselves, customer-specific data logic, and brownfield migration.

Troubleshooting

For issues not covered here, contact your Cognite support team or post in #topic-deployment-packs on Slack.

Support & ownership