infrahub-managing-transforms

Subclass InfrahubTransform from infrahub_sdk.transforms, set the query class attribute to a named query in .infrahub.yml, and implement transform() to return a dict or str. Register the class under python_transforms with its file_path and class_name.

Use Jinja2 when the output is text formatting with minimal logic, such as device configs or topology files. Use Python when you need computation like IP math, data reshaping, or platform-based template selection.

The query class attribute likely does not match any query name registered in .infrahub.yml, so collect_data() skips silently and transform() receives None. For Jinja2 transforms, a missing top-level query field causes the template to render against empty data.

Infrahub enforces a fixed enum of eight values including text/plain, text/csv, application/json, application/yaml, application/xml, application/hcl, text/markdown, and image/svg+xml. Note that text/yaml is rejected; use application/yaml instead.

Run infrahubctl transform for Python transforms or infrahubctl render for Jinja2 templates against a live server. You can also define YAML-driven smoke, unit, and integration tests with JSON fixtures using the Resources Testing Framework.

The Infrahub SDK builds its Jinja2 environment from an explicit AVAILABLE_FILTERS allowlist plus netutils filters, so Ansible filters like regex_replace or to_nice_yaml fail with 'No filter named' errors. Use the hybrid pattern with Python preprocessing instead.