fastembed-haystack 0.0.5__tar.gz → 0.1.0__tar.gz

{fastembed_haystack-0.0.5 → fastembed_haystack-0.1.0}/.gitignore

@@ -134,3 +134,4 @@ dmypy.json
 
 # Docs generation artifacts
 _readme_*.md
+.idea

{fastembed_haystack-0.0.5 → fastembed_haystack-0.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: fastembed-haystack
-Version: 0.0.5
+Version: 0.1.0
 Summary: Haystack 2.x component to embed strings and Documents using fastembed embedding model
 Project-URL: Source, https://github.com/deepset-ai/haystack-core-integrations
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/blob/main/integrations/fastembed/README.md
@@ -18,8 +18,8 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: Implementation :: CPython
 Classifier: Programming Language :: Python :: Implementation :: PyPy
 Requires-Python: >=3.8
-Requires-Dist: fastembed>=0.2
-Requires-Dist: haystack-ai
+Requires-Dist: fastembed>=0.2.5
+Requires-Dist: haystack-ai>=2.0.1
 Description-Content-Type: text/markdown
 
 # fastembed-haystack
@@ -67,6 +67,31 @@ doc = Document(content="fastembed is supported by and maintained by Qdrant.", me
 result = embedder.run(documents=[doc])
 ```
 
+You can use `FastembedSparseTextEmbedder` and `FastembedSparseDocumentEmbedder` by importing as:
+
+```python
+from haystack_integrations.components.embedders.fastembed import FastembedSparseTextEmbedder
+
+text = "fastembed is supported by and maintained by Qdrant."
+text_embedder = FastembedSparseTextEmbedder(
+    model="prithvida/Splade_PP_en_v1"
+)
+text_embedder.warm_up()
+embedding = text_embedder.run(text)["embedding"]
+```
+
+```python
+from haystack_integrations.components.embedders.fastembed import FastembedSparseDocumentEmbedder
+from haystack.dataclasses import Document
+
+embedder = FastembedSparseDocumentEmbedder(
+    model="prithvida/Splade_PP_en_v1",
+)
+embedder.warm_up()
+doc = Document(content="fastembed is supported by and maintained by Qdrant.", meta={"long_answer": "no",})
+result = embedder.run(documents=[doc])
+```
+
 ## License
 
 `fastembed-haystack` is distributed under the terms of the [Apache-2.0](https://spdx.org/licenses/Apache-2.0.html) license.

{fastembed_haystack-0.0.5 → fastembed_haystack-0.1.0}/README.md

@@ -43,6 +43,31 @@ doc = Document(content="fastembed is supported by and maintained by Qdrant.", me
 result = embedder.run(documents=[doc])
 ```
 
+You can use `FastembedSparseTextEmbedder` and `FastembedSparseDocumentEmbedder` by importing as:
+
+```python
+from haystack_integrations.components.embedders.fastembed import FastembedSparseTextEmbedder
+
+text = "fastembed is supported by and maintained by Qdrant."
+text_embedder = FastembedSparseTextEmbedder(
+    model="prithvida/Splade_PP_en_v1"
+)
+text_embedder.warm_up()
+embedding = text_embedder.run(text)["embedding"]
+```
+
+```python
+from haystack_integrations.components.embedders.fastembed import FastembedSparseDocumentEmbedder
+from haystack.dataclasses import Document
+
+embedder = FastembedSparseDocumentEmbedder(
+    model="prithvida/Splade_PP_en_v1",
+)
+embedder.warm_up()
+doc = Document(content="fastembed is supported by and maintained by Qdrant.", meta={"long_answer": "no",})
+result = embedder.run(documents=[doc])
+```
+
 ## License
 
 `fastembed-haystack` is distributed under the terms of the [Apache-2.0](https://spdx.org/licenses/Apache-2.0.html) license.

{fastembed_haystack-0.0.5/example → fastembed_haystack-0.1.0/examples}/example.py

@@ -26,7 +26,7 @@ query = "Who supports fastembed?"
 
 result = query_pipeline.run({"text_embedder": {"text": query}})
 
-print(result["retriever"]["documents"][0])  # noqa: T201
+print(result["retriever"]["documents"][0])
 
 # Document(id=...,
 #  content: 'fastembed is supported by and maintained by Qdrant.',

fastembed_haystack-0.1.0/examples/sparse_example.py

@@ -0,0 +1,32 @@
+# Currently, this example shows how to use the FastembedSparseDocumentEmbedder component to embed a list of documents.
+
+# TODO: Once we have a proper SparseEmbeddingRetriever, we should replace this naive example with a more realistic one,
+# involving indexing and retrieval of documents.
+
+from haystack import Document
+from haystack_integrations.components.embedders.fastembed import FastembedSparseDocumentEmbedder
+
+document_list = [
+    Document(
+        content="Oxidative stress generated within inflammatory joints can produce autoimmune phenomena and joint destruction. Radical species with oxidative activity, including reactive nitrogen species, represent mediators of inflammation and cartilage damage.",
+        meta={
+            "pubid": "25,445,628",
+            "long_answer": "yes",
+        },
+    ),
+    Document(
+        content="Plasma levels of pancreatic polypeptide (PP) rise upon food intake. Although other pancreatic islet hormones, such as insulin and glucagon, have been extensively investigated, PP secretion and actions are still poorly understood.",
+        meta={
+            "pubid": "25,445,712",
+            "long_answer": "yes",
+        },
+    ),
+]
+
+document_embedder = FastembedSparseDocumentEmbedder()
+document_embedder.warm_up()
+documents_with_embeddings = document_embedder.run(document_list)["documents"]
+
+for doc in documents_with_embeddings:
+    print(f"Document Text: {doc.content}")
+    print(f"Document Sparse Embedding: {doc.sparse_embedding.to_dict()}")

{fastembed_haystack-0.0.5 → fastembed_haystack-0.1.0}/pydoc/config.yml

@@ -5,6 +5,8 @@ loaders:
       [
         "haystack_integrations.components.embedders.fastembed.fastembed_document_embedder",
         "haystack_integrations.components.embedders.fastembed.fastembed_text_embedder",
+        "haystack_integrations.components.embedders.fastembed.fastembed_sparse_document_embedder",
+        "haystack_integrations.components.embedders.fastembed.fastembed_sparse_text_embedder"
       ]
     ignore_when_discovered: ["__init__"]
 processors:
@@ -24,6 +26,7 @@ renderer:
   order: 80
   markdown:
     descriptive_class_title: false
+    classdef_code_block: false
     descriptive_module_title: true
     add_method_class_prefix: true
     add_member_class_prefix: false

{fastembed_haystack-0.0.5 → fastembed_haystack-0.1.0}/pyproject.toml

@@ -25,8 +25,8 @@ classifiers = [
   "Programming Language :: Python :: Implementation :: PyPy",
 ]
 dependencies = [
-"haystack-ai",
-"fastembed>=0.2",
+"haystack-ai>=2.0.1",
+"fastembed>=0.2.5",
 ]
 
 [project.urls]
@@ -95,12 +95,12 @@ all = [
 ]
 
 [tool.black]
-target-version = ["py37"]
+target-version = ["py38"]
 line-length = 120
 skip-string-normalization = true
 
 [tool.ruff]
-target-version = "py37"
+target-version = "py38"
 line-length = 120
 select = [
   "A",
@@ -155,19 +155,16 @@ ban-relative-imports = "parents"
 # Tests can use magic values, assertions, and relative imports
 "tests/**/*" = ["PLR2004", "S101", "TID252"]
 # examples can contain "print" commands
-"examples/**/*" = ["T201"]
+"examples/**/*" = ["T201", "E501"]
 
 [tool.coverage.run]
-source_pkgs = ["src", "tests"]
+source = ["haystack_integrations"]
 branch = true
-parallel = true
-
-
-[tool.coverage.paths]
-fastembed_haystack = ["src/haystack_integrations", "*/fastembed-haystack/src"]
-tests = ["tests", "*/fastembed-haystack/tests"]
+parallel = false
 
 [tool.coverage.report]
+omit = ["*/tests/*", "*/__init__.py"]
+show_missing=true
 exclude_lines = [
   "no cov",
   "if __name__ == .__main__.:",

fastembed_haystack-0.1.0/src/haystack_integrations/components/embedders/fastembed/__init__.py

@@ -0,0 +1,14 @@
+# SPDX-FileCopyrightText: 2024-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+from .fastembed_document_embedder import FastembedDocumentEmbedder
+from .fastembed_sparse_document_embedder import FastembedSparseDocumentEmbedder
+from .fastembed_sparse_text_embedder import FastembedSparseTextEmbedder
+from .fastembed_text_embedder import FastembedTextEmbedder
+
+__all__ = [
+    "FastembedDocumentEmbedder",
+    "FastembedTextEmbedder",
+    "FastembedSparseDocumentEmbedder",
+    "FastembedSparseTextEmbedder",
+]

fastembed_haystack-0.1.0/src/haystack_integrations/components/embedders/fastembed/embedding_backend/fastembed_backend.py

@@ -0,0 +1,109 @@
+from typing import ClassVar, Dict, List, Optional
+
+from haystack.dataclasses.sparse_embedding import SparseEmbedding
+from tqdm import tqdm
+
+from fastembed import TextEmbedding
+from fastembed.sparse.sparse_text_embedding import SparseTextEmbedding
+
+
+class _FastembedEmbeddingBackendFactory:
+    """
+    Factory class to create instances of fastembed embedding backends.
+    """
+
+    _instances: ClassVar[Dict[str, "_FastembedEmbeddingBackend"]] = {}
+
+    @staticmethod
+    def get_embedding_backend(
+        model_name: str,
+        cache_dir: Optional[str] = None,
+        threads: Optional[int] = None,
+    ):
+        embedding_backend_id = f"{model_name}{cache_dir}{threads}"
+
+        if embedding_backend_id in _FastembedEmbeddingBackendFactory._instances:
+            return _FastembedEmbeddingBackendFactory._instances[embedding_backend_id]
+
+        embedding_backend = _FastembedEmbeddingBackend(model_name=model_name, cache_dir=cache_dir, threads=threads)
+        _FastembedEmbeddingBackendFactory._instances[embedding_backend_id] = embedding_backend
+        return embedding_backend
+
+
+class _FastembedEmbeddingBackend:
+    """
+    Class to manage fastembed embeddings.
+    """
+
+    def __init__(
+        self,
+        model_name: str,
+        cache_dir: Optional[str] = None,
+        threads: Optional[int] = None,
+    ):
+        self.model = TextEmbedding(model_name=model_name, cache_dir=cache_dir, threads=threads)
+
+    def embed(self, data: List[str], progress_bar=True, **kwargs) -> List[List[float]]:
+        # the embed method returns a Iterable[np.ndarray], so we convert it to a list of lists
+        embeddings = []
+        embeddings_iterable = self.model.embed(data, **kwargs)
+        for np_array in tqdm(
+            embeddings_iterable, disable=not progress_bar, desc="Calculating embeddings", total=len(data)
+        ):
+            embeddings.append(np_array.tolist())
+        return embeddings
+
+
+class _FastembedSparseEmbeddingBackendFactory:
+    """
+    Factory class to create instances of fastembed sparse embedding backends.
+    """
+
+    _instances: ClassVar[Dict[str, "_FastembedSparseEmbeddingBackend"]] = {}
+
+    @staticmethod
+    def get_embedding_backend(
+        model_name: str,
+        cache_dir: Optional[str] = None,
+        threads: Optional[int] = None,
+    ):
+        embedding_backend_id = f"{model_name}{cache_dir}{threads}"
+
+        if embedding_backend_id in _FastembedSparseEmbeddingBackendFactory._instances:
+            return _FastembedSparseEmbeddingBackendFactory._instances[embedding_backend_id]
+
+        embedding_backend = _FastembedSparseEmbeddingBackend(
+            model_name=model_name, cache_dir=cache_dir, threads=threads
+        )
+        _FastembedSparseEmbeddingBackendFactory._instances[embedding_backend_id] = embedding_backend
+        return embedding_backend
+
+
+class _FastembedSparseEmbeddingBackend:
+    """
+    Class to manage fastembed sparse embeddings.
+    """
+
+    def __init__(
+        self,
+        model_name: str,
+        cache_dir: Optional[str] = None,
+        threads: Optional[int] = None,
+    ):
+        self.model = SparseTextEmbedding(model_name=model_name, cache_dir=cache_dir, threads=threads)
+
+    def embed(self, data: List[List[str]], progress_bar=True, **kwargs) -> List[SparseEmbedding]:
+        # The embed method returns a Iterable[SparseEmbedding], so we convert to Haystack SparseEmbedding type.
+        # Each SparseEmbedding contains an `indices` key containing a list of int and
+        # an `values` key containing a list of floats.
+
+        sparse_embeddings = []
+        sparse_embeddings_iterable = self.model.embed(data, **kwargs)
+        for sparse_embedding in tqdm(
+            sparse_embeddings_iterable, disable=not progress_bar, desc="Calculating sparse embeddings", total=len(data)
+        ):
+            sparse_embeddings.append(
+                SparseEmbedding(indices=sparse_embedding.indices.tolist(), values=sparse_embedding.values.tolist())
+            )
+
+        return sparse_embeddings

{fastembed_haystack-0.0.5 → fastembed_haystack-0.1.0}/src/haystack_integrations/components/embedders/fastembed/fastembed_document_embedder.py

@@ -8,7 +8,7 @@ from .embedding_backend.fastembed_backend import _FastembedEmbeddingBackendFacto
 @component
 class FastembedDocumentEmbedder:
     """
-    A component for computing Document embeddings using Fastembed embedding models.
+    FastembedDocumentEmbedder computes Document embeddings using Fastembed embedding models.
     The embedding of each Document is stored in the `embedding` field of the Document.
 
     Usage example:
@@ -48,6 +48,7 @@ class FastembedDocumentEmbedder:
     print(f"Document Text: {result['documents'][0].content}")
     print(f"Document Embedding: {result['documents'][0].embedding}")
     print(f"Embedding Dimension: {len(result['documents'][0].embedding)}")
+    ```
     """  # noqa: E501
 
     def __init__(
@@ -67,11 +68,11 @@ class FastembedDocumentEmbedder:
         Create an FastembedDocumentEmbedder component.
 
         :param model: Local path or name of the model in Hugging Face's model hub,
-            such as ``'BAAI/bge-small-en-v1.5'``.
-        :param cache_dir (str, optional): The path to the cache directory.
+            such as `BAAI/bge-small-en-v1.5`.
+        :param cache_dir: The path to the cache directory.
                 Can be set using the `FASTEMBED_CACHE_PATH` env variable.
                 Defaults to `fastembed_cache` in the system's temp directory.
-        :param threads (int, optional): The number of threads single onnxruntime session can use. Defaults to None.
+        :param threads: The number of threads single onnxruntime session can use. Defaults to None.
         :param prefix: A string to add to the beginning of each text.
         :param suffix: A string to add to the end of each text.
         :param batch_size: Number of strings to encode at once.
@@ -97,7 +98,9 @@ class FastembedDocumentEmbedder:
 
     def to_dict(self) -> Dict[str, Any]:
         """
-        Serialize this component to a dictionary.
+        Serializes the component to a dictionary.
+        :returns:
+            Dictionary with serialized data.
         """
         return default_to_dict(
             self,
@@ -115,7 +118,7 @@ class FastembedDocumentEmbedder:
 
     def warm_up(self):
         """
-        Load the embedding backend.
+        Initializes the component.
         """
         if not hasattr(self, "embedding_backend"):
             self.embedding_backend = _FastembedEmbeddingBackendFactory.get_embedding_backend(
@@ -128,18 +131,21 @@ class FastembedDocumentEmbedder:
             meta_values_to_embed = [
                 str(doc.meta[key]) for key in self.meta_fields_to_embed if key in doc.meta and doc.meta[key] is not None
             ]
-            text_to_embed = [
-                self.prefix + self.embedding_separator.join([*meta_values_to_embed, doc.content or ""]) + self.suffix,
-            ]
+            text_to_embed = (
+                self.prefix + self.embedding_separator.join([*meta_values_to_embed, doc.content or ""]) + self.suffix
+            )
 
-            texts_to_embed.append(text_to_embed[0])
+            texts_to_embed.append(text_to_embed)
         return texts_to_embed
 
     @component.output_types(documents=List[Document])
     def run(self, documents: List[Document]):
         """
-        Embed a list of Documents.
-        The embedding of each Document is stored in the `embedding` field of the Document.
+        Embeds a list of Documents.
+
+        :param documents: List of Documents to embed.
+        :returns: A dictionary with the following keys:
+            - `documents`: List of Documents with each Document's `embedding` field set to the computed embeddings.
         """
         if not isinstance(documents, list) or documents and not isinstance(documents[0], Document):
             msg = (
@@ -151,13 +157,11 @@ class FastembedDocumentEmbedder:
             msg = "The embedding model has not been loaded. Please call warm_up() before running."
             raise RuntimeError(msg)
 
-        # TODO: once non textual Documents are properly supported, we should also prepare them for embedding here
-
         texts_to_embed = self._prepare_texts_to_embed(documents=documents)
         embeddings = self.embedding_backend.embed(
             texts_to_embed,
             batch_size=self.batch_size,
-            show_progress_bar=self.progress_bar,
+            progress_bar=self.progress_bar,
             parallel=self.parallel,
         )
 

fastembed_haystack-0.1.0/src/haystack_integrations/components/embedders/fastembed/fastembed_sparse_document_embedder.py

@@ -0,0 +1,160 @@
+from typing import Any, Dict, List, Optional
+
+from haystack import Document, component, default_to_dict
+
+from .embedding_backend.fastembed_backend import _FastembedSparseEmbeddingBackendFactory
+
+
+@component
+class FastembedSparseDocumentEmbedder:
+    """
+    FastembedSparseDocumentEmbedder computes Document embeddings using Fastembed sparse models.
+
+    Usage example:
+    ```python
+    # To use this component, install the "fastembed-haystack" package.
+    # pip install fastembed-haystack
+
+    from haystack_integrations.components.embedders.fastembed import FastembedSparseDocumentEmbedder
+    from haystack.dataclasses import Document
+
+    doc_embedder = FastembedSparseDocumentEmbedder(
+        model="prithvida/Splade_PP_en_v1",
+        batch_size=32,
+    )
+
+    doc_embedder.warm_up()
+
+    # Text taken from PubMed QA Dataset (https://huggingface.co/datasets/pubmed_qa)
+    document_list = [
+        Document(
+            content="Oxidative stress generated within inflammatory joints can produce autoimmune phenomena and joint destruction. Radical species with oxidative activity, including reactive nitrogen species, represent mediators of inflammation and cartilage damage.",
+            meta={
+                "pubid": "25,445,628",
+                "long_answer": "yes",
+            },
+        ),
+        Document(
+            content="Plasma levels of pancreatic polypeptide (PP) rise upon food intake. Although other pancreatic islet hormones, such as insulin and glucagon, have been extensively investigated, PP secretion and actions are still poorly understood.",
+            meta={
+                "pubid": "25,445,712",
+                "long_answer": "yes",
+            },
+        ),
+    ]
+
+    result = doc_embedder.run(document_list)
+    print(f"Document Text: {result['documents'][0].content}")
+    print(f"Document Embedding: {result['documents'][0].sparse_embedding}")
+    print(f"Embedding Dimension: {len(result['documents'][0].sparse_embedding)}")
+    ```
+    """  # noqa: E501
+
+    def __init__(
+        self,
+        model: str = "prithvida/Splade_PP_en_v1",
+        cache_dir: Optional[str] = None,
+        threads: Optional[int] = None,
+        batch_size: int = 32,
+        progress_bar: bool = True,
+        parallel: Optional[int] = None,
+        meta_fields_to_embed: Optional[List[str]] = None,
+        embedding_separator: str = "\n",
+    ):
+        """
+        Create an FastembedDocumentEmbedder component.
+
+        :param model: Local path or name of the model in Hugging Face's model hub,
+            such as `prithvida/Splade_PP_en_v1`.
+        :param cache_dir: The path to the cache directory.
+                Can be set using the `FASTEMBED_CACHE_PATH` env variable.
+                Defaults to `fastembed_cache` in the system's temp directory.
+        :param threads: The number of threads single onnxruntime session can use.
+        :param batch_size: Number of strings to encode at once.
+        :param progress_bar: If `True`, displays progress bar during embedding.
+        :param parallel:
+                If > 1, data-parallel encoding will be used, recommended for offline encoding of large datasets.
+                If 0, use all available cores.
+                If None, don't use data-parallel processing, use default onnxruntime threading instead.
+        :param meta_fields_to_embed: List of meta fields that should be embedded along with the Document content.
+        :param embedding_separator: Separator used to concatenate the meta fields to the Document content.
+        """
+
+        self.model_name = model
+        self.cache_dir = cache_dir
+        self.threads = threads
+        self.batch_size = batch_size
+        self.progress_bar = progress_bar
+        self.parallel = parallel
+        self.meta_fields_to_embed = meta_fields_to_embed or []
+        self.embedding_separator = embedding_separator
+
+    def to_dict(self) -> Dict[str, Any]:
+        """
+        Serializes the component to a dictionary.
+        :returns:
+            Dictionary with serialized data.
+        """
+        return default_to_dict(
+            self,
+            model=self.model_name,
+            cache_dir=self.cache_dir,
+            threads=self.threads,
+            batch_size=self.batch_size,
+            progress_bar=self.progress_bar,
+            parallel=self.parallel,
+            meta_fields_to_embed=self.meta_fields_to_embed,
+            embedding_separator=self.embedding_separator,
+        )
+
+    def warm_up(self):
+        """
+        Initializes the component.
+        """
+        if not hasattr(self, "embedding_backend"):
+            self.embedding_backend = _FastembedSparseEmbeddingBackendFactory.get_embedding_backend(
+                model_name=self.model_name, cache_dir=self.cache_dir, threads=self.threads
+            )
+
+    def _prepare_texts_to_embed(self, documents: List[Document]) -> List[str]:
+        texts_to_embed = []
+        for doc in documents:
+            meta_values_to_embed = [
+                str(doc.meta[key]) for key in self.meta_fields_to_embed if key in doc.meta and doc.meta[key] is not None
+            ]
+            text_to_embed = self.embedding_separator.join([*meta_values_to_embed, doc.content or ""])
+
+            texts_to_embed.append(text_to_embed)
+        return texts_to_embed
+
+    @component.output_types(documents=List[Document])
+    def run(self, documents: List[Document]):
+        """
+        Embeds a list of Documents.
+
+        :param documents: List of Documents to embed.
+        :returns: A dictionary with the following keys:
+            - `documents`: List of Documents with each Document's `sparse_embedding`
+                            field set to the computed embeddings.
+        """
+        if not isinstance(documents, list) or documents and not isinstance(documents[0], Document):
+            msg = (
+                "FastembedSparseDocumentEmbedder expects a list of Documents as input. "
+                "In case you want to embed a list of strings, please use the FastembedTextEmbedder."
+            )
+            raise TypeError(msg)
+        if not hasattr(self, "embedding_backend"):
+            msg = "The embedding model has not been loaded. Please call warm_up() before running."
+            raise RuntimeError(msg)
+
+        texts_to_embed = self._prepare_texts_to_embed(documents=documents)
+        embeddings = self.embedding_backend.embed(
+            texts_to_embed,
+            batch_size=self.batch_size,
+            show_progress_bar=self.progress_bar,
+            parallel=self.parallel,
+        )
+
+        for doc, emb in zip(documents, embeddings):
+            doc.sparse_embedding = emb
+        return {"documents": documents}