# Google Cloud Cortex Framework
[**Cortex Framework v7**](https://docs.cloud.google.com/cortex/docs/overview) introduces a modular deployment architecture, simple data orchestration via [Dataform](https://cloud.google.com/dataform), and AI-ready data products with [BigQuery](https://cloud.google.com/bigquery) and [Knowledge Catalog](https://docs.cloud.google.com/dataplex/docs/introduction) integration \- enabling enterprises to quickly build, extend, and deploy data assets and pipelines for advanced analytics and agentic use cases with less risk, complexity, and cost.
## Key features
* **Modular deployment and smart dependency resolution**: Deploy exactly what you need. Simply select the desired data products, and the framework will automatically identify, retrieve, and transform the necessary tables to the data foundation layer, ensuring no unnecessary data is processed. Easily add custom fields or logic without breaking standard models.
* **Native runtime dependency graph generation:** Automatically orchestrates the runtime execution order for complex data models using Dataform, ensuring prerequisite tables are processed before downstream data foundations and data products.
* **Knowledge Catalog integration:** Provides an option to automatically synchronize deployed Cortex Framework data products, complete with enriched business descriptions and metadata, directly into Knowledge Catalog for discovery and governance.
* **Skills for an agentic data product builder:** Accelerate development by creating new data products simply via natural language. Cortex Framework provides packaged agentic skills to translate your business needs into tangible data assets.
* **Extensibility framework:** Maintain a clean separation between your custom data products and Cortex Framework delivered data products using namespaces. This ensures you can benefit from the latest updates without impacting your custom work.
* **Bring your own CDC (External data foundation):** A flexible architecture allows you to bypass built-in change data capture processing and connect your own existing CDC pipelines directly to the foundation layer.
* **Multi-system SAP support:** Built-in dynamic dependency resolution and logic differentiation allows for parallel deployment of SAP ECC and SAP S/4HANA source systems. Bring in data from multiple SAP ERP systems.
* **High data fidelity & semantics:** Features core platform capabilities like dynamic discovery of custom fields, AI-ready metadata generation, and robust semantic mapping. In SAP-specific data products, this framework handles advanced logic, such as translating complex SAP terminology into business-friendly terms and integrating the SAP TCURX table for exact currency decimal shifts.
* **Serverless BigQuery-native execution:** Orchestration relies entirely on Google Cloud Dataform, which natively executes the version-controlled SQL generated by Cortex Framework for seamless data transformation and processing.
* **Incremental loading:** Cortex Framework data products are built with native incremental loading configurations to ensure highly efficient processing of large enterprise datasets in BigQuery. This reduces processing time and costs by processing only new or changed data since the last execution.
* **Environment management and operations:** Gain deeper control over deployments with the option to provide [compilation overrides](https://docs.cloud.google.com/dataform/docs/manage-repository#create-compilation-variables) that integrate cleanly with Dataform release configurations.
Please refer to the [public documentation](https://docs.cloud.google.com/cortex/docs/overview) for more information.
## Supported data sources
Cortex Framework v7 supports the following data sources:
* [SAP ECC](https://docs.cloud.google.com/cortex/docs/source-system-integration/sap-erp)
* [SAP S/4HANA](https://docs.cloud.google.com/cortex/docs/source-system-integration/sap-erp)
* [SAP Business Data Cloud](https://docs.cloud.google.com/cortex/docs/source-system-integration/sap-bdc)
Find all available data products on the [public documentation](https://docs.cloud.google.com/cortex/docs/data-product) site.
## Getting started
### Demo deployment
To quickly spin up a demo deployment with raw sample data for evaluations:
`uv run cortex-demo --project_id=<PROJECT_ID> --region=<REGION> --location=<LOCATION>`
For detailed instructions on deployment and configuration, please refer to our [public documentation](https://docs.cloud.google.com/cortex/docs/overview).
## Agentic data product builder
Cortex Framework provides modular developer skills located in `.agents/skills/` that AI coding agents follow to scaffold, update, validate, and deploy data products.
Invoke these workflows directly in your AI assistant or agent chat.
### Customizing skills
Each agentic data product builder skill directory under `.agents/skills/<skill_name>/` contains a `custom/` subdirectory designed to extend or override default instructions without modifying core skill files. If custom instructions or files are added to this folder, the AI agent automatically reads, merges, and adheres to them; if the `custom/` directory contains only a `.keep` placeholder, the agent follows the standard base skill instructions.
### Adding new skills
To add a new skill, create a dedicated directory under `.agents/skills/<new_skill_name>/` containing a `SKILL.md` file with YAML frontmatter (`name`, `description`), workflow guidelines, and verification quality gates. You can also include supporting directories like `assets/`, `references/`, `scripts/`, and `custom/` (with a `.keep` file), and register any associated slash commands in `.agents/AGENTS.md`.
## Telemetry
Cortex Framework gathers anonymized usage statistics during deployment to help us analyze and improve the overall framework experience.
When telemetry is active, a specialized metadata usage header is transmitted with [Dataform](https://cloud.google.com/dataform) and [BigQuery](https://cloud.google.com/bigquery?e=48754805) API requests. This metadata is managed within Google Cloud’s internal analytics platform in strict accordance with established data privacy and governance standards, and contains the following information:
* Cloud project number
* Deployment region
* Framework version
* Specific module configurations
For more information on how data is handled, please consult the [Google Privacy Policy](https://policies.google.com/privacy).
#### Opting Out
Telemetry is \*\*active by default\*\*. You can disable this feature by including the \`--disable-telemetry\` flag when running the framework's command-line tools. Applying this setting prevents the framework from including usage headers in any API communication.
\*\*Example:\*\*
```shell
bash
uv run cortex-build-and-deploy --config "config/sample.config.yaml" --disable-telemetry
To opt out globally, run:
bash
uv run cortex-config telemetry disable
```
For detailed information on Telemetry, please refer to our [public documentation](https://docs.cloud.google.com/cortex/docs/overview).