nemo_retriever package#

Subpackages#

Submodules#

nemo_retriever.branch_extraction module#

nemo_retriever.config_utils module#

nemo_retriever.export module#

nemo_retriever.graph_ingestor module#

nemo_retriever.ingest_config module#

nemo_retriever.ingest_manifest module#

nemo_retriever.ingest_plans module#

nemo_retriever.ingest_results module#

nemo_retriever.ingestor module#

nemo_retriever.retriever module#

nemo_retriever.retriever_graph_utils module#

Helpers for Retriever graph execution.

nemo_retriever.retriever_graph_utils.filter_retrieval_kwargs(
kwargs: dict[str, Any],
) → dict[str, Any][source]#

Drop keys reserved for graph coordination (not forwarded to VDB.retrieval).

nemo_retriever.retriever_graph_utils.hits_lists_to_rerank_dataframe(
query_texts: list[str],
hits_per_query: list[list[dict[str, Any]]],
) → DataFrame[source]#

One row per (query, hit) with payload to rebuild hits after reranking.

nemo_retriever.retriever_graph_utils.rerank_long_dataframe_to_hits(
df: DataFrame,
*,
query_texts: list[str],
top_k: int,
score_column: str = 'rerank_score',
) → list[list[dict[str, Any]]][source]#

Group long rerank output by query (preserving query_texts order), take top_k per query.

nemo_retriever.service_ingestor module#

nemo_retriever.version module#

Module contents#