Heuristic:Marker Inc Korea AutoRAG Warning Deprecated Legacy QA Creation
| Knowledge Sources | |
|---|---|
| Domains | Data_Engineering, QA_Generation, Deprecation |
| Last Updated | 2026-02-08 06:00 GMT |
Overview
Deprecation warning: the legacy QA dataset creation subsystem (autorag/data/legacy/qacreation/) is superseded by the modern QA schema-based pipeline (Corpus.sample() + QA.batch_apply()).
Description
The entire autorag/data/legacy/qacreation/ module tree is located in the legacy/ directory, indicating it is maintained for backward compatibility but is not the recommended path for new projects. The make_qa_with_existing_qa function in base.py explicitly raises DeprecationWarning at line 147. The modern replacement uses the composable QA schema class with batch_apply() for query generation, answer generation, and filtering.
Usage
Apply this warning when encountering code that imports from autorag.data.legacy.qacreation. Prefer migrating to the modern pipeline (Corpus.sample() followed by QA.batch_apply() with factoid_query_gen, make_basic_gen_gt, etc.) unless specific legacy backend features (RAGAS evolution types, guidance-based structured generation) are required and not yet available in the modern API.
The Insight (Rule of Thumb)
- Action: Migrate from legacy QA creation functions to the modern QA schema pipeline.
- Value: Replace make_single_content_qa(corpus_df, ...) with Corpus.sample().batch_apply(factoid_query_gen).batch_apply(make_basic_gen_gt).
- Trade-off: The legacy API supports some backends (RAGAS evolution, guidance structured gen) that may not have direct modern equivalents yet.
Reasoning
The legacy module was the original approach before the QA schema class was introduced. The modern API provides:
- Composable method chaining (QA.batch_apply())
- Built-in filtering (QA.filter() / QA.batch_filter())
- Standardized schema with to_parquet() export
- Better async batching and error handling
The legacy code remains functional but receives no new features. The explicit DeprecationWarning in make_qa_with_existing_qa signals active deprecation of at least part of the API.
Related Pages
- Implementation:Marker_Inc_Korea_AutoRAG_Make_Single_Content_QA
- Implementation:Marker_Inc_Korea_AutoRAG_Generate_QA_LlamaIndex
- Implementation:Marker_Inc_Korea_AutoRAG_Generate_QA_Ragas
- Implementation:Marker_Inc_Korea_AutoRAG_Generate_Simple_QA_Dataset
- Principle:Marker_Inc_Korea_AutoRAG_Legacy_QA_Dataset_Creation