feat(FastEmbed): Support for SPLADE Sparse Embedder (#579)

* fix(opensearch): bulk error without create key

* feat(FastEmbed): Scaffold for SPLADE Sparse Embedding Support

* Revert "fix(opensearch): bulk error without create key"

This reverts commit afc8e79.

* feat(FastEmbed): __all__ fix

* feat(FastEmbed): fix one test

* feat(FastEmbed): fix one test

* feat(FastEmbed): fix a second test

* feat(FastEmbed): removed old TODO (fixed)

* feat(FastEmbed): fixing all test + doc

* fix output typing

* Fix output component

* feat(FastEmbed): renaming SPLADE to Sparse because it makes more sense

* feat(FastEmbed): hatch run all lint

* feat(FastEmbed): modify PR for haystack 2.1.0 with proper sparse vectors

* try testing with Haystack main branch

* update model name

* Update integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/fastembed_sparse_document_embedder.py

Co-authored-by: Stefano Fiorucci <[email protected]>

* Update integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/fastembed_sparse_document_embedder.py

Co-authored-by: Stefano Fiorucci <[email protected]>

* Update integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/fastembed_sparse_document_embedder.py

Co-authored-by: Stefano Fiorucci <[email protected]>

* Update integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/fastembed_sparse_document_embedder.py

Co-authored-by: Stefano Fiorucci <[email protected]>

* Update integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/fastembed_sparse_text_embedder.py

Co-authored-by: Stefano Fiorucci <[email protected]>

* feat(FastEmbed): remove prefix/suffix

* feat(FastEmbed): fix linting

* feat(FastEmbed): suggestion for progress bar

* feat(FastEmbed): return Haystack's SparseEmbedding instead of Dict

* feat(FastEmbed): fix lint

* feat(Fastembed): run output type from dict to haystack sparseembedding class

* feat(FastEmbed): reduce default sparse batch size

* Update integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/fastembed_sparse_text_embedder.py

Co-authored-by: Stefano Fiorucci <[email protected]>

* feat(FastEmbed): fix test

* updates after 2.0.1 release

* small fixes; naive example

---------

Co-authored-by: anakin87 <[email protected]>

.github/workflows/fastembed.yml

@@ -43,7 +43,7 @@ jobs:
 
       - name: Run tests
         id: tests
-        run: hatch run cov
+        run: hatch run cov        
 
       - name: Nightly - run unit tests with Haystack main branch
         if: github.event_name == 'schedule'
@@ -60,4 +60,4 @@ jobs:
             core-integrations failure: 
             ${{ (steps.tests.conclusion == 'nightly-haystack-main') && 'nightly-haystack-main' || 'tests' }}
              - ${{ github.workflow }}
-          api-key: ${{ secrets.CORE_DATADOG_API_KEY }}
+          api-key: ${{ secrets.CORE_DATADOG_API_KEY }}

integrations/fastembed/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.

integrations/fastembed/example/example.py → integrations/fastembed/examples/example.py

@@ -26,7 +26,7 @@
 
 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.',

integrations/fastembed/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()}")

integrations/fastembed/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:

integrations/fastembed/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]
@@ -155,7 +155,7 @@ 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 = ["haystack_integrations"]

integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/__init__.py

@@ -2,6 +2,13 @@
 #
 # 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"]
+__all__ = [
+    "FastembedDocumentEmbedder",
+    "FastembedTextEmbedder",
+    "FastembedSparseDocumentEmbedder",
+    "FastembedSparseTextEmbedder",
+]

integrations/fastembed/src/haystack_integrations/components/embedders/fastembed/embedding_backend/fastembed_backend.py

@@ -1,8 +1,10 @@
 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:
@@ -50,3 +52,58 @@ def embed(self, data: List[str], progress_bar=True, **kwargs) -> List[List[float
         ):
             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

integrations/fastembed/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}