We present KGpipe, a modular open-source framework for constructing and executing pipelines for integrating RDF, JSON, and
text sources into knowledge graphs (KG). KGpipe supports heterogeneous execution backends, intermediate exchange formats, static pipeline validation, and can reuse existing implementations for integration tasks such as information extraction or ontology and entity matching, including LLM-based components. To demonstrate the framework’s flexibility, we implement and comparatively evaluate multiple pipeline variants for the different source formats. The results show that structured RDF pipelines currently provide the most stable integration behavior, whereas JSON and text pipelines remain more sensitive to errors in mapping, extraction, and linking.