zvec-db 0.3.0__py3-none-any.whl

zvec_db/__init__.py

@@ -0,0 +1,104 @@
+"""zvec-db: Suite d'utilitaires pour la vectorisation sparse/dense et le re-ranking.
+
+This library provides embedders for sparse and dense vectorization, as well as
+rerankers for post-processing and combining search results. It is designed to
+work with zvec (https://github.com/ccdv-ai/zvec).
+
+Example Usage
+-------------
+>>> from zvec_db import BM25Embedder, SentenceTransformersEmbedder
+>>> from zvec_db import NormalizedWeightedReRanker
+>>> from zvec_db import NormalizationConfig
+
+# Sparse embedding
+bm25 = BM25Embedder(max_features=4096)
+bm25.fit(documents)
+sparse_vector = bm25("query")  # __call__ alias for embed()
+
+# Dense embedding
+dense = SentenceTransformersEmbedder(model_name="all-MiniLM-L6-v2")
+dense.fit(documents)
+dense_vector = dense("query")
+
+# Re-ranking
+reranker = NormalizedWeightedReRanker(
+    weights={"sparse": 0.4, "dense": 0.6},
+    normalizer_configs={"sparse": {"method": "bayes"}},
+)
+"""
+
+from .embedders import (  # Sparse; Dense
+    BM25Embedder,
+    BM25LEmbedder,
+    BM25PlusEmbedder,
+    CountEmbedder,
+    DisMaxEmbedder,
+    OpenAIEmbedder,
+    SentenceTransformersEmbedder,
+    TfidfEmbedder,
+)
+from .preprocessing import (
+    NormalizationConfig,
+    normalize_text,
+    remove_stopwords,
+    stem_word,
+)
+from .rerankers import (  # Fusion; Diversification; Cross-encoder; Utils; Defaults
+    BaseCrossEncoderReranker,
+    ClassificationReranker,
+    DefaultHybridReranker,
+    DefaultRrfReranker,
+    DefaultWeightedReranker,
+    HybridFusionReranker,
+    MultiFieldWeightedReranker,
+    NormalizedRrfReRanker,
+    NormalizedWeightedReRanker,
+    OpenAIDecoderReranker,
+    OpenAIEncoderReranker,
+    OpenAIReranker,
+    PipelineReranker,
+    RrfReRanker,
+    SentenceTransformerReranker,
+    SubmodularReranker,
+    WeightedReRanker,
+)
+
+__all__ = [
+    # Sparse Embedders
+    "BM25Embedder",
+    "BM25LEmbedder",
+    "BM25PlusEmbedder",
+    "TfidfEmbedder",
+    "CountEmbedder",
+    "DisMaxEmbedder",
+    # Dense Embedders
+    "SentenceTransformersEmbedder",
+    "OpenAIEmbedder",
+    # Fusion Rerankers
+    "RrfReRanker",
+    "NormalizedRrfReRanker",
+    "WeightedReRanker",
+    "NormalizedWeightedReRanker",
+    "HybridFusionReranker",
+    "MultiFieldWeightedReranker",
+    # Diversification
+    "SubmodularReranker",
+    # Cross-encoder Rerankers
+    "BaseCrossEncoderReranker",
+    "SentenceTransformerReranker",
+    "ClassificationReranker",
+    "OpenAIReranker",
+    "OpenAIEncoderReranker",
+    "OpenAIDecoderReranker",
+    # Utils
+    "PipelineReranker",
+    # Defaults
+    "DefaultRrfReranker",
+    "DefaultWeightedReranker",
+    "DefaultHybridReranker",
+    # Preprocessing
+    "NormalizationConfig",
+    "normalize_text",
+    "stem_word",
+    "remove_stopwords",
+]

zvec_db/embedders/__init__.py

@@ -0,0 +1,36 @@
+"""Embedders for dense and sparse representations."""
+
+from .base import BaseSparseEmbedder
+from .defaults import DefaultDenseEmbedder, DefaultSparseEmbedder
+from .dense import (
+    BaseDenseEmbedder,
+    OpenAIEmbedder,
+    SentenceTransformersEmbedder,
+)
+from .sparse import (
+    BM25Embedder,
+    BM25LEmbedder,
+    BM25PlusEmbedder,
+    CountEmbedder,
+    DisMaxEmbedder,
+    TfidfEmbedder,
+)
+
+__all__ = [
+    # Base
+    "BaseSparseEmbedder",
+    # Defaults
+    "DefaultSparseEmbedder",
+    "DefaultDenseEmbedder",
+    # Sparse
+    "BM25Embedder",
+    "BM25LEmbedder",
+    "BM25PlusEmbedder",
+    "TfidfEmbedder",
+    "CountEmbedder",
+    "DisMaxEmbedder",
+    # Dense
+    "BaseDenseEmbedder",
+    "SentenceTransformersEmbedder",
+    "OpenAIEmbedder",
+]