fix(cohere): use dataclass replace to avoid modifying input documents (#2755)

This PR fixes the Cohere document embedder and ranker to not modify input
Documents in place when setting embeddings or scores.

Instead of mutating the original documents:
  doc.embedding = embeddings
  doc.score = score

We now create new document instances using dataclass replace:
  replace(doc, embedding=embeddings)
  replace(doc, score=score)

This follows the established pattern from haystack-ai/haystack#9693 and
aligns with other integrations (FastEmbed, Optimum, Nvidia, Bedrock).

Related to: #2174

Author: GunaPalanivel

integrations/cohere/src/haystack_integrations/components/embedders/cohere/document_embedder.py

@@ -1,6 +1,7 @@
 # SPDX-FileCopyrightText: 2023-present deepset GmbH <info@deepset.ai>
 #
 # SPDX-License-Identifier: Apache-2.0
+from dataclasses import replace
 from typing import Any
 
 from haystack import Document, component, default_from_dict, default_to_dict
@@ -195,10 +196,11 @@ def run(self, documents: list[Document]) -> dict[str, list[Document] | dict[str,
             self.embedding_type,
         )
 
+        new_documents = []
         for doc, embeddings in zip(documents, all_embeddings, strict=True):
-            doc.embedding = embeddings
+            new_documents.append(replace(doc, embedding=embeddings))
 
-        return {"documents": documents, "meta": metadata}
+        return {"documents": new_documents, "meta": metadata}
 
     @component.output_types(documents=list[Document], meta=dict[str, Any])
     async def run_async(self, documents: list[Document]) -> dict[str, list[Document] | dict[str, Any]]:
@@ -228,7 +230,8 @@ async def run_async(self, documents: list[Document]) -> dict[str, list[Document]
             embedding_type=self.embedding_type,
         )
 
+        new_documents = []
         for doc, embeddings in zip(documents, all_embeddings, strict=True):
-            doc.embedding = embeddings
+            new_documents.append(replace(doc, embedding=embeddings))
 
-        return {"documents": documents, "meta": metadata}
+        return {"documents": new_documents, "meta": metadata}

integrations/cohere/src/haystack_integrations/components/rankers/cohere/ranker.py

@@ -1,3 +1,4 @@
+from dataclasses import replace
 from typing import Any
 
 from haystack import Document, component, default_from_dict, default_to_dict, logging
@@ -162,6 +163,5 @@ def run(self, query: str, documents: list[Document], top_k: int | None = None) -
         sorted_docs = []
         for idx, score in zip(indices, scores, strict=True):
             doc = documents[idx]
-            doc.score = score
-            sorted_docs.append(documents[idx])
+            sorted_docs.append(replace(doc, score=score))
         return {"documents": sorted_docs}

integrations/cohere/tests/test_document_embedder.py

@@ -195,6 +195,51 @@ async def test_run_async(self, mock_get_response):
             assert doc_with_embedding.meta == doc.meta
             assert doc_with_embedding.embedding == embedding
 
+    @patch("haystack_integrations.components.embedders.cohere.document_embedder.get_response")
+    def test_run_does_not_modify_original_documents(self, mock_get_response):
+        embedder = CohereDocumentEmbedder(api_key=Secret.from_token("test-api-key"))
+
+        embeddings = [[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]]
+        mock_get_response.return_value = (embeddings, {"api_version": "1.0"})
+
+        docs = [
+            Document(content="I love cheese", meta={"topic": "Cuisine"}),
+            Document(content="A transformer is a deep learning architecture", meta={"topic": "ML"}),
+        ]
+
+        result = embedder.run(docs)
+
+        # Check that the original documents are not modified
+        for doc in docs:
+            assert doc.embedding is None
+
+        # Check that the returned documents have embeddings
+        for doc_with_embedding, embedding in zip(result["documents"], embeddings, strict=True):
+            assert doc_with_embedding.embedding == embedding
+
+    @pytest.mark.asyncio
+    @patch("haystack_integrations.components.embedders.cohere.document_embedder.get_async_response")
+    async def test_run_async_does_not_modify_original_documents(self, mock_get_response):
+        embedder = CohereDocumentEmbedder(api_key=Secret.from_token("test-api-key"))
+
+        embeddings = [[0.1, 0.2, 0.3], [0.4, 0.5, 0.6]]
+        mock_get_response.return_value = (embeddings, {"api_version": "1.0"})
+
+        docs = [
+            Document(content="I love cheese", meta={"topic": "Cuisine"}),
+            Document(content="A transformer is a deep learning architecture", meta={"topic": "ML"}),
+        ]
+
+        result = await embedder.run_async(docs)
+
+        # Check that the original documents are not modified
+        for doc in docs:
+            assert doc.embedding is None
+
+        # Check that the returned documents have embeddings
+        for doc_with_embedding, embedding in zip(result["documents"], embeddings, strict=True):
+            assert doc_with_embedding.embedding == embedding
+
     @pytest.mark.skipif(
         not os.environ.get("COHERE_API_KEY", None) and not os.environ.get("CO_API_KEY", None),
         reason="Export an env var called COHERE_API_KEY/CO_API_KEY containing the Cohere API key to run this test.",

integrations/cohere/tests/test_ranker.py

@@ -295,6 +295,27 @@ def test_run_topk_set_in_init(self, monkeypatch, mock_ranker_response):  # noqa:
             Document(id="efgh", content="doc2", score=0.95),
         ]
 
+    def test_run_does_not_modify_original_documents(self, monkeypatch, mock_ranker_response):  # noqa: ARG002
+        monkeypatch.setenv("CO_API_KEY", "test-api-key")
+        ranker = CohereRanker(top_k=2)
+        query = "test"
+        documents = [
+            Document(id="abcd", content="doc1"),
+            Document(id="efgh", content="doc2"),
+            Document(id="ijkl", content="doc3"),
+        ]
+
+        ranker_results = ranker.run(query, documents)
+
+        # Check that the original documents are not modified
+        for doc in documents:
+            assert doc.score is None
+
+        # Check that the returned documents have scores
+        reranked_docs = ranker_results["documents"]
+        for doc in reranked_docs:
+            assert doc.score is not None
+
     @pytest.mark.skipif(
         not os.environ.get("COHERE_API_KEY", None) and not os.environ.get("CO_API_KEY", None),
         reason="Export an env var called COHERE_API_KEY/CO_API_KEY containing the Cohere API key to run this test.",