chroma-haystack 4.3.0__tar.gz → 4.3.2__tar.gz

{chroma_haystack-4.3.0 → chroma_haystack-4.3.2}/CHANGELOG.md

@@ -1,5 +1,36 @@
 # Changelog
 
+## [integrations/chroma-v4.3.1] - 2026-06-29
+
+### 🐛 Bug Fixes
+
+- Serialize ChromaDocumentStore metadata in to_dict (#3445)
+
+
+## [integrations/chroma-v4.3.0] - 2026-04-29
+
+### 🚀 Features
+
+- *(chroma)* Support DuplicatePolicy in write_documents and use async mixin tests (#3245)
+
+### 🚜 Refactor
+
+- Weaviate, chroma, elasticsearch, opensearch, azure_ai_search use `_normalize_metadata_field_name` from haystack.utils (#2953)
+
+### 🧪 Testing
+
+- `ChromaDocumentStore` use Mixin tests (#3026)
+- Track test coverage for all integrations (#3065)
+- Better categorize some Document Stores tests (#3085)
+- Chroma - add unit tests (#3175)
+
+### 🧹 Chores
+
+- Add missing -> None return type annotations to chroma __init__ methods (#2976)
+- Enforce ruff docstring rules (D102/D103/D205/D209/D213/D417/D419) in first 10 integrations (#3008)
+- Increase lower pins for 3.14 support in some integrations + test with 3.14 (#3033)
+
+
 ## [integrations/chroma-v4.2.0] - 2026-03-11
 
 ### 🚀 Features

{chroma_haystack-4.3.0 → chroma_haystack-4.3.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: chroma-haystack
-Version: 4.3.0
+Version: 4.3.2
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/chroma#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues
 Project-URL: Source, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/chroma

{chroma_haystack-4.3.0 → chroma_haystack-4.3.2}/src/haystack_integrations/document_stores/chroma/document_store.py

@@ -140,9 +140,10 @@ class ChromaDocumentStore:
 
             self._client = client  # store client for potential future use
 
-            self._metadata = self._metadata or {}
-            if "hnsw:space" not in self._metadata:
-                self._metadata["hnsw:space"] = self._distance_function
+            # Build the collection metadata locally so `self._metadata` stays exactly as the user passed it.
+            # This keeps `to_dict()` deterministic and avoids mutating a user-supplied dict in place.
+            collection_metadata = dict(self._metadata) if self._metadata else {}
+            collection_metadata.setdefault("hnsw:space", self._distance_function)
 
             existing_collection_names = [c.name for c in client.list_collections()]
             if self._collection_name in existing_collection_names:
@@ -151,14 +152,14 @@ class ChromaDocumentStore:
                     embedding_function=self._embedding_func,
                 )
 
-                if self._metadata != self._collection.metadata:
+                if collection_metadata != self._collection.metadata:
                     logger.warning(
                         "Collection already exists. The `distance_function` and `metadata` parameters will be ignored."
                     )
             else:
                 self._collection = client.create_collection(
                     name=self._collection_name,
-                    metadata=self._metadata,
+                    metadata=collection_metadata,
                     embedding_function=self._embedding_func,
                 )
 
@@ -188,9 +189,10 @@ class ChromaDocumentStore:
 
             self._async_client = client  # store client for potential future use
 
-            self._metadata = self._metadata or {}
-            if "hnsw:space" not in self._metadata:
-                self._metadata["hnsw:space"] = self._distance_function
+            # Build the collection metadata locally so `self._metadata` stays exactly as the user passed it.
+            # This keeps `to_dict()` deterministic and avoids mutating a user-supplied dict in place.
+            collection_metadata = dict(self._metadata) if self._metadata else {}
+            collection_metadata.setdefault("hnsw:space", self._distance_function)
 
             collection = await client.list_collections()
             existing_collection_names = [c.name for c in collection]
@@ -200,14 +202,14 @@ class ChromaDocumentStore:
                     embedding_function=self._embedding_func,
                 )
 
-                if self._metadata != self._async_collection.metadata:
+                if collection_metadata != self._async_collection.metadata:
                     logger.warning(
                         "Collection already exists. The `distance_function` and `metadata` parameters will be ignored."
                     )
             else:
                 self._async_collection = await client.create_collection(
                     name=self._collection_name,
-                    metadata=self._metadata,
+                    metadata=collection_metadata,
                     embedding_function=self._embedding_func,
                 )
 
@@ -1373,6 +1375,7 @@ class ChromaDocumentStore:
             host=self._host,
             port=self._port,
             distance_function=self._distance_function,
+            metadata=self._metadata,
             client_settings=self._client_settings,
             **self._embedding_function_params,
         )

{chroma_haystack-4.3.0 → chroma_haystack-4.3.2}/src/haystack_integrations/document_stores/chroma/filters.py

@@ -68,7 +68,7 @@ def _convert_filters(filters: dict[str, Any]) -> ChromaFilter:
             continue
         # if field is "id", it'll be passed to Chroma's ids filter
         elif field == "id":
-            if not value["$eq"]:
+            if "$eq" not in value or not value["$eq"]:
                 msg = f"id filter only supports '==' operator, got {value}"
                 raise ChromaDocumentStoreFilterError(msg)
             ids.append(value["$eq"])

{chroma_haystack-4.3.0 → chroma_haystack-4.3.2}/tests/test_document_store.py

@@ -101,6 +101,7 @@ class TestDocumentStoreUnit:
                 "port": None,
                 "api_key": "1234567890",
                 "distance_function": "l2",
+                "metadata": None,
                 "client_settings": {"anonymized_telemetry": False},
             },
         }
@@ -128,6 +129,44 @@ class TestDocumentStoreUnit:
         assert ds._embedding_function_params == {"api_key": "1234567890"}
         assert ds._client_settings == {"anonymized_telemetry": False}
 
+    def test_to_dict_and_from_dict_preserves_metadata(self):
+        """`metadata` configures the underlying Chroma collection (e.g.
+        ``hnsw:space`` and other HNSW index parameters). It must survive a
+        to_dict/from_dict round-trip, otherwise serializing a pipeline to YAML
+        silently drops the collection settings when the store is reloaded."""
+        metadata = {"hnsw:space": "cosine", "hnsw:search_ef": 200}
+        ds = ChromaDocumentStore(
+            collection_name="test_md",
+            embedding_function="HuggingFaceEmbeddingFunction",
+            api_key="1234567890",
+            metadata=metadata,
+        )
+
+        assert ds.to_dict()["init_parameters"]["metadata"] == metadata
+
+        restored = ChromaDocumentStore.from_dict(ds.to_dict())
+        assert restored._metadata == metadata
+
+    def test_ensure_initialized_does_not_mutate_metadata(self):
+        """Initialization derives the collection metadata (injecting ``hnsw:space``) into a
+        local copy, so ``self._metadata`` must stay exactly as the user passed it. Otherwise
+        ``to_dict()`` would become lifecycle-dependent and a user-supplied dict would be mutated
+        in place."""
+        # metadata=None must stay None after initialization
+        store = ChromaDocumentStore(collection_name="test_no_metadata", distance_function="cosine")
+        store._ensure_initialized()
+        assert store._metadata is None
+        assert store.to_dict()["init_parameters"]["metadata"] is None
+
+        # a user-supplied dict without "hnsw:space" must not be mutated
+        user_metadata = {"hnsw:search_ef": 200}
+        store = ChromaDocumentStore(
+            collection_name="test_user_metadata", distance_function="cosine", metadata=user_metadata
+        )
+        store._ensure_initialized()
+        assert store._metadata == {"hnsw:search_ef": 200}
+        assert "hnsw:space" not in user_metadata
+
     def test_same_collection_name_reinitialization(self):
         ChromaDocumentStore("test_1")
         ChromaDocumentStore("test_1")

{chroma_haystack-4.3.0 → chroma_haystack-4.3.2}/tests/test_filters.py

@@ -17,6 +17,13 @@ def test_id_filter_with_empty_value_raises():
         _convert_filters({"field": "id", "operator": "==", "value": ""})
 
 
+def test_id_filter_with_non_eq_operator_raises():
+    # A non-'==' operator produces e.g. {"$in": [...]} (no "$eq" key), which must
+    # surface the friendly ChromaDocumentStoreFilterError, not a raw KeyError('$eq').
+    with pytest.raises(ChromaDocumentStoreFilterError, match="id filter only supports"):
+        _convert_filters({"field": "id", "operator": "in", "value": ["1", "2"]})
+
+
 @pytest.mark.parametrize(
     ("condition", "match"),
     [

{chroma_haystack-4.3.0 → chroma_haystack-4.3.2}/tests/test_retriever.py

@@ -44,6 +44,7 @@ class TestChromaQueryTextRetriever:
                         "port": None,
                         "api_key": "1234567890",
                         "distance_function": "l2",
+                        "metadata": None,
                         "client_settings": None,
                     },
                 },
@@ -158,6 +159,7 @@ class TestChromaEmbeddingRetriever:
                         "port": None,
                         "api_key": "1234567890",
                         "distance_function": "l2",
+                        "metadata": None,
                         "client_settings": None,
                     },
                 },