rag.core.passage¶

Passage selection helpers shared by RAG backends.

Attributes¶

`AUTHOR_QUESTION_CUES`
`ACKNOWLEDGEMENT_CUES`
`PROCEEDINGS_BOILERPLATE_CUES`
`QA_GROUNDING_RULES`

`is_author_question`(→ bool)	Return True when the question is asking who wrote a paper.
`is_acknowledgement_text`(→ bool)	Return True when text looks like thanks / manuscript revision notes.
`is_proceedings_boilerplate_text`(→ bool)	Return True when text is SPIE copyright / volume-editor footer.
`page_from_mapping`(→ int \| None)	Read a 1-based page number from chunk metadata.

rag.core.passage.AUTHOR_QUESTION_CUES: Final = ('who wrote', 'who is the writer', 'who is the author', 'who are the authors', 'who is the main...[source]¶

rag.core.passage.ACKNOWLEDGEMENT_CUES: Final = ('acknowledg', 'thanks to', 'thank ', 'grateful to', 'careful revision', 'revising the...[source]¶

rag.core.passage.PROCEEDINGS_BOILERPLATE_CUES: Final = ('edited by', 'proc. of spie', 'ccc code', 'spiedigitallibrary', 'terms of use')[source]¶

rag.core.passage.QA_GROUNDING_RULES: Final = "People thanked for revising a manuscript are not authors. SPIE 'edited by' names are volume...[source]¶

rag.core.passage.is_author_question(*, question: str) → bool[source]¶

Return True when the question is asking who wrote a paper.

rag.core.passage.is_acknowledgement_text(*, text: str) → bool[source]¶

Return True when text looks like thanks / manuscript revision notes.

rag.core.passage.is_proceedings_boilerplate_text(*, text: str) → bool[source]¶

Return True when text is SPIE copyright / volume-editor footer.

rag.core.passage.page_from_mapping(*, metadata: dict[str, Any]) → int | None[source]¶

Read a 1-based page number from chunk metadata.