opensearch-haystack 8.1.2__tar.gz → 8.2.1__tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (36) hide show
  1. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/CHANGELOG.md +26 -0
  2. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/PKG-INFO +1 -1
  3. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/pyproject.toml +1 -0
  4. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/opensearch/bm25_retriever.py +12 -0
  5. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/opensearch/embedding_retriever.py +12 -0
  6. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/opensearch/metadata_retriever.py +12 -0
  7. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/opensearch/open_search_hybrid_retriever.py +15 -2
  8. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/opensearch/sql_retriever.py +12 -0
  9. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/document_stores/opensearch/document_store.py +68 -25
  10. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/document_stores/opensearch/filters.py +5 -2
  11. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_bm25_retriever.py +23 -1
  12. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_document_store.py +111 -65
  13. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_document_store_async.py +150 -0
  14. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_embedding_retriever.py +23 -1
  15. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_filters.py +53 -0
  16. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_metadata_retriever.py +23 -1
  17. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_open_search_hybrid_retriever.py +102 -92
  18. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_sql_retriever.py +22 -0
  19. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/.gitignore +0 -0
  20. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/LICENSE +0 -0
  21. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/README.md +0 -0
  22. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/docker-compose.yml +0 -0
  23. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/pydoc/config_docusaurus.yml +0 -0
  24. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/opensearch/__init__.py +0 -0
  25. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/opensearch/utils.py +0 -0
  26. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/components/retrievers/py.typed +0 -0
  27. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/document_stores/opensearch/__init__.py +0 -0
  28. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/document_stores/opensearch/auth.py +0 -0
  29. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/document_stores/opensearch/opensearch_scripts.py +0 -0
  30. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/src/haystack_integrations/document_stores/py.typed +0 -0
  31. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/__init__.py +0 -0
  32. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/conftest.py +0 -0
  33. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_auth.py +0 -0
  34. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_document_store_common.py +0 -0
  35. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_metadata_retriever_async.py +0 -0
  36. {opensearch_haystack-8.1.2 → opensearch_haystack-8.2.1}/tests/test_utils.py +0 -0
@@ -1,5 +1,31 @@
1
1
  # Changelog
2
2
 
3
+ ## [integrations/opensearch-v8.2.0] - 2026-07-22
4
+
5
+ ### 🚀 Features
6
+
7
+ - Opensearch - add methods to close resources (#3640)
8
+
9
+ ### 🐛 Bug Fixes
10
+
11
+ - Adapt optimum, ragas, elasticsearch, opensearch to APIs removed in Haystack 3.0 (#3538)
12
+
13
+ ### 🚜 Refactor
14
+
15
+ - Elasticsearch - split sync/async paths (#3633)
16
+
17
+ ### 📚 Documentation
18
+
19
+ - Replace old haystack core imports with haystack_integrations paths (#3545)
20
+
21
+
22
+ ## [integrations/opensearch-v8.1.2] - 2026-06-24
23
+
24
+ ### 🐛 Bug Fixes
25
+
26
+ - Serialize OpenSearchBM25Retriever all_terms_must_match (#3454)
27
+
28
+
3
29
  ## [integrations/opensearch-v8.1.1] - 2026-05-20
4
30
 
5
31
  ### 🐛 Bug Fixes
@@ -1,6 +1,6 @@
1
1
  Metadata-Version: 2.4
2
2
  Name: opensearch-haystack
3
- Version: 8.1.2
3
+ Version: 8.2.1
4
4
  Summary: Haystack 2.x Document Store for OpenSearch
5
5
  Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/opensearch#readme
6
6
  Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues
@@ -139,6 +139,7 @@ ignore = [
139
139
  "PLR0912",
140
140
  "PLR0913",
141
141
  "PLR0915",
142
+ "PLR0917",
142
143
  # Ignore asserts
143
144
  "S101",
144
145
  # Allow `Any` - used legitimately for dynamic types and SDK boundaries
@@ -157,6 +157,18 @@ class OpenSearchBM25Retriever:
157
157
  data["init_parameters"]["filter_policy"] = FilterPolicy.from_str(data["init_parameters"]["filter_policy"])
158
158
  return default_from_dict(cls, data)
159
159
 
160
+ def close(self) -> None:
161
+ """
162
+ Release the synchronous resources of the underlying Document Store.
163
+ """
164
+ self._document_store.close()
165
+
166
+ async def close_async(self) -> None:
167
+ """
168
+ Release the asynchronous resources of the underlying Document Store.
169
+ """
170
+ await self._document_store.close_async()
171
+
160
172
  def _prepare_bm25_args(
161
173
  self,
162
174
  *,
@@ -163,6 +163,18 @@ class OpenSearchEmbeddingRetriever:
163
163
  data["init_parameters"]["filter_policy"] = FilterPolicy.from_str(data["init_parameters"]["filter_policy"])
164
164
  return default_from_dict(cls, data)
165
165
 
166
+ def close(self) -> None:
167
+ """
168
+ Release the synchronous resources of the underlying Document Store.
169
+ """
170
+ self._document_store.close()
171
+
172
+ async def close_async(self) -> None:
173
+ """
174
+ Release the asynchronous resources of the underlying Document Store.
175
+ """
176
+ await self._document_store.close_async()
177
+
166
178
  @component.output_types(documents=list[Document])
167
179
  def run(
168
180
  self,
@@ -181,6 +181,18 @@ class OpenSearchMetadataRetriever:
181
181
  )
182
182
  return default_from_dict(cls, data)
183
183
 
184
+ def close(self) -> None:
185
+ """
186
+ Release the synchronous resources of the underlying Document Store.
187
+ """
188
+ self._document_store.close()
189
+
190
+ async def close_async(self) -> None:
191
+ """
192
+ Release the asynchronous resources of the underlying Document Store.
193
+ """
194
+ await self._document_store.close_async()
195
+
184
196
  @component.output_types(metadata=list[dict[str, Any]])
185
197
  def run(
186
198
  self,
@@ -37,7 +37,9 @@ class OpenSearchHybridRetriever:
37
37
 
38
38
  ```python
39
39
  from haystack import Document
40
- from haystack.components.embedders import SentenceTransformersTextEmbedder, SentenceTransformersDocumentEmbedder
40
+ # Requires: pip install sentence-transformers-haystack
41
+ from haystack_integrations.components.embedders.sentence_transformers import SentenceTransformersTextEmbedder
42
+ from haystack_integrations.components.embedders.sentence_transformers import SentenceTransformersDocumentEmbedder
41
43
  from haystack_integrations.components.retrievers.opensearch import OpenSearchHybridRetriever
42
44
  from haystack_integrations.document_stores.opensearch import OpenSearchDocumentStore
43
45
 
@@ -59,7 +61,6 @@ class OpenSearchHybridRetriever:
59
61
 
60
62
  # Embed the documents and add them to the document store
61
63
  doc_embedder = SentenceTransformersDocumentEmbedder(model="sentence-transformers/all-MiniLM-L6-v2")
62
- doc_embedder.warm_up()
63
64
  docs = doc_embedder.run(docs)
64
65
  doc_store.write_documents(docs['documents'])
65
66
 
@@ -356,3 +357,15 @@ class OpenSearchHybridRetriever:
356
357
  data["init_parameters"]["join_mode"] = join_mode
357
358
 
358
359
  return default_from_dict(cls, data)
360
+
361
+ def close(self) -> None:
362
+ """
363
+ Release the synchronous resources of the underlying Document Store.
364
+ """
365
+ self.document_store.close()
366
+
367
+ async def close_async(self) -> None:
368
+ """
369
+ Release the asynchronous resources of the underlying Document Store.
370
+ """
371
+ await self.document_store.close_async()
@@ -80,6 +80,18 @@ class OpenSearchSQLRetriever:
80
80
  )
81
81
  return default_from_dict(cls, data)
82
82
 
83
+ def close(self) -> None:
84
+ """
85
+ Release the synchronous resources of the underlying Document Store.
86
+ """
87
+ self._document_store.close()
88
+
89
+ async def close_async(self) -> None:
90
+ """
91
+ Release the asynchronous resources of the underlying Document Store.
92
+ """
93
+ await self._document_store.close_async()
94
+
83
95
  @component.output_types(result=dict[str, Any])
84
96
  def run(
85
97
  self,
@@ -5,6 +5,7 @@
5
5
  # ruff: noqa: FBT001, FBT002 boolean-type-hint-positional-argument and boolean-default-value-positional-argument
6
6
 
7
7
  from collections.abc import Generator, Mapping
8
+ from contextlib import suppress
8
9
  from dataclasses import replace
9
10
  from math import exp
10
11
  from typing import Any, Literal
@@ -162,7 +163,6 @@ class OpenSearchDocumentStore:
162
163
  # the document store is instantiated.
163
164
  self._client: OpenSearch | None = None
164
165
  self._async_client: AsyncOpenSearch | None = None
165
- self._initialized = False
166
166
 
167
167
  def _get_default_mappings(self) -> dict[str, Any]:
168
168
  default_mappings: dict[str, Any] = {
@@ -303,8 +303,6 @@ class OpenSearchDocumentStore:
303
303
  timeout=self._timeout,
304
304
  **self._kwargs,
305
305
  )
306
- self._initialized = True
307
-
308
306
  self._ensure_index_exists()
309
307
 
310
308
  async def _ensure_initialized_async(self) -> None:
@@ -322,9 +320,26 @@ class OpenSearchDocumentStore:
322
320
  connection_class=AsyncHttpConnection,
323
321
  **self._kwargs,
324
322
  )
325
- self._initialized = True
326
323
  await self._ensure_index_exists_async()
327
324
 
325
+ def close(self) -> None:
326
+ """
327
+ Release the associated synchronous resources.
328
+ """
329
+ if self._client is not None:
330
+ with suppress(Exception):
331
+ self._client.close()
332
+ self._client = None
333
+
334
+ async def close_async(self) -> None:
335
+ """
336
+ Release the associated asynchronous resources.
337
+ """
338
+ if self._async_client is not None:
339
+ with suppress(Exception):
340
+ await self._async_client.close()
341
+ self._async_client = None
342
+
328
343
  @staticmethod
329
344
  def _extract_nested_fields_from_mapping(mapping_properties: dict[str, Any]) -> set[str]:
330
345
  return {name for name, defn in mapping_properties.items() if defn.get("type") == "nested"}
@@ -2038,16 +2053,19 @@ class OpenSearchDocumentStore:
2038
2053
  self,
2039
2054
  metadata_field: str,
2040
2055
  search_term: str | None = None,
2041
- size: int | None = 10000,
2056
+ size: int | None = 10,
2042
2057
  after: dict[str, Any] | None = None,
2043
2058
  ) -> tuple[list[str], dict[str, Any] | None]:
2044
2059
  """
2045
- Returns unique values for a metadata field, optionally filtered by a search term in the content.
2060
+ Returns unique values for a metadata field, optionally filtered by a search term.
2046
2061
 
2047
2062
  Uses composite aggregations for proper pagination beyond 10k results.
2048
2063
 
2049
2064
  :param metadata_field: The metadata field to get unique values for.
2050
- :param search_term: Optional search term to filter documents by matching in the content field.
2065
+ :param search_term: Optional term to filter the returned values by, matching as a case-insensitive substring
2066
+ of the metadata field's own value (not the document content). NOTE: The matching is done with a server-side
2067
+ script to accomplish the substring matching on the value of the field and this operation is quite expensive
2068
+ for a large corpus.
2051
2069
  :param size: The number of unique values to return per page. Defaults to 10000.
2052
2070
  :param after: Optional pagination key from the previous response. Use None for the first page.
2053
2071
  For subsequent pages, pass the `after_key` from the previous response.
@@ -2060,12 +2078,6 @@ class OpenSearchDocumentStore:
2060
2078
 
2061
2079
  field_name = _normalize_metadata_field_name(metadata_field)
2062
2080
 
2063
- # filter by search_term if provided
2064
- query: dict[str, Any] = {"match_all": {}}
2065
- if search_term:
2066
- # Use match_phrase for exact phrase matching to avoid tokenization issues
2067
- query = {"match_phrase": {"content": search_term}}
2068
-
2069
2081
  # Build composite aggregation for proper pagination
2070
2082
  composite_agg: dict[str, Any] = {
2071
2083
  "size": size,
@@ -2074,8 +2086,7 @@ class OpenSearchDocumentStore:
2074
2086
  if after is not None:
2075
2087
  composite_agg["after"] = after
2076
2088
 
2077
- body = {
2078
- "query": query,
2089
+ body: dict[str, Any] = {
2079
2090
  "aggs": {
2080
2091
  "unique_values": {
2081
2092
  "composite": composite_agg,
@@ -2083,6 +2094,24 @@ class OpenSearchDocumentStore:
2083
2094
  },
2084
2095
  "size": 0, # we only need aggregations, not documents
2085
2096
  }
2097
+ if search_term:
2098
+ # Composite aggregation terms sources don't support `include`/`exclude` (that's only valid on
2099
+ # standalone `terms` aggregations), and a `regexp` query only works on keyword/text fields, not
2100
+ # numeric ones. A doc-value script query works uniformly across field types by stringifying the
2101
+ # field's value, matching the case-insensitive substring semantics documented above. The term is
2102
+ # lower-cased once here rather than per-document in the script.
2103
+ body["query"] = {
2104
+ "script": {
2105
+ "script": {
2106
+ "source": (
2107
+ "def v = doc[params.field]; "
2108
+ "if (v.size() == 0) { return false; } "
2109
+ "return v.value.toString().toLowerCase().contains(params.term)"
2110
+ ),
2111
+ "params": {"field": field_name, "term": search_term.lower()},
2112
+ }
2113
+ }
2114
+ }
2086
2115
 
2087
2116
  result = self._client.search(index=self._index, body=body)
2088
2117
  aggregations = result.get("aggregations", {})
@@ -2108,12 +2137,15 @@ class OpenSearchDocumentStore:
2108
2137
  after: dict[str, Any] | None = None,
2109
2138
  ) -> tuple[list[str], dict[str, Any] | None]:
2110
2139
  """
2111
- Asynchronously returns unique values for a metadata field, optionally filtered by a search term in the content.
2140
+ Asynchronously returns unique values for a metadata field, optionally filtered by a search term.
2112
2141
 
2113
2142
  Uses composite aggregations for proper pagination beyond 10k results.
2114
2143
 
2115
2144
  :param metadata_field: The metadata field to get unique values for.
2116
- :param search_term: Optional search term to filter documents by matching in the content field.
2145
+ :param search_term: Optional term to filter the returned values by, matching as a case-insensitive substring
2146
+ of the metadata field's own value (not the document content). NOTE: The matching is done with a server-side
2147
+ script to accomplish the substring matching on the value of the field and this operation is quite expensive
2148
+ for a large corpus.
2117
2149
  :param size: The number of unique values to return per page. Defaults to 10000.
2118
2150
  :param after: Optional pagination key from the previous response. Use None for the first page.
2119
2151
  For subsequent pages, pass the `after_key` from the previous response.
@@ -2126,12 +2158,6 @@ class OpenSearchDocumentStore:
2126
2158
 
2127
2159
  field_name = _normalize_metadata_field_name(metadata_field)
2128
2160
 
2129
- # filter by search_term if provided
2130
- query: dict[str, Any] = {"match_all": {}}
2131
- if search_term:
2132
- # Use match_phrase for exact phrase matching to avoid tokenization issues
2133
- query = {"match_phrase": {"content": search_term}}
2134
-
2135
2161
  # Build composite aggregation for proper pagination
2136
2162
  composite_agg: dict[str, Any] = {
2137
2163
  "size": size,
@@ -2140,8 +2166,7 @@ class OpenSearchDocumentStore:
2140
2166
  if after is not None:
2141
2167
  composite_agg["after"] = after
2142
2168
 
2143
- body = {
2144
- "query": query,
2169
+ body: dict[str, Any] = {
2145
2170
  "aggs": {
2146
2171
  "unique_values": {
2147
2172
  "composite": composite_agg,
@@ -2149,6 +2174,24 @@ class OpenSearchDocumentStore:
2149
2174
  },
2150
2175
  "size": 0, # we only need aggregations, not documents
2151
2176
  }
2177
+ if search_term:
2178
+ # Composite aggregation terms sources don't support `include`/`exclude` (that's only valid on
2179
+ # standalone `terms` aggregations), and a `regexp` query only works on keyword/text fields, not
2180
+ # numeric ones. A doc-value script query works uniformly across field types by stringifying the
2181
+ # field's value, matching the case-insensitive substring semantics documented above. The term is
2182
+ # lower-cased once here rather than per-document in the script.
2183
+ body["query"] = {
2184
+ "script": {
2185
+ "script": {
2186
+ "source": (
2187
+ "def v = doc[params.field]; "
2188
+ "if (v.size() == 0) { return false; } "
2189
+ "return v.value.toString().toLowerCase().contains(params.term)"
2190
+ ),
2191
+ "params": {"field": field_name, "term": search_term.lower()},
2192
+ }
2193
+ }
2194
+ }
2152
2195
 
2153
2196
  result = await self._async_client.search(index=self._index, body=body)
2154
2197
  aggregations = result.get("aggregations", {})
@@ -93,7 +93,7 @@ def _group_nested_conditions(
93
93
  inner.append(_parse_logical_condition(c, nested_fields=None))
94
94
  else:
95
95
  inner.append(_parse_comparison_condition(c, nested_fields))
96
- if len(inner) > 1:
96
+ if len(inner) > 1 and operator != "OR":
97
97
  inner = _normalize_ranges(inner)
98
98
  if len(inner) == 1:
99
99
  conditions.append({"nested": {"path": path, "query": inner[0]}})
@@ -120,7 +120,10 @@ def _parse_logical_condition(condition: dict[str, Any], nested_fields: set[str]
120
120
  else:
121
121
  conditions = [_parse_comparison_condition(c, nested_fields) for c in condition["conditions"]]
122
122
 
123
- if len(conditions) > 1:
123
+ # Range merging is only valid when the conditions are AND-combined (AND, and the
124
+ # inner `must` of NOT). Merging same-field ranges under OR would collapse e.g.
125
+ # `price < 10 OR price > 100` into a single impossible range, matching nothing.
126
+ if len(conditions) > 1 and operator != "OR":
124
127
  conditions = _normalize_ranges(conditions)
125
128
  if operator == "AND":
126
129
  return {"bool": {"must": conditions}}
@@ -2,7 +2,7 @@
2
2
  #
3
3
  # SPDX-License-Identifier: Apache-2.0
4
4
 
5
- from unittest.mock import Mock, patch
5
+ from unittest.mock import AsyncMock, Mock, patch
6
6
 
7
7
  import pytest
8
8
  from haystack.dataclasses import Document
@@ -159,6 +159,28 @@ def test_from_dict_not_defaults(_mock_opensearch_client):
159
159
  assert retriever._raise_on_failure is True
160
160
 
161
161
 
162
+ def test_close():
163
+ mock_store = Mock(spec=OpenSearchDocumentStore)
164
+ retriever = OpenSearchBM25Retriever(document_store=mock_store)
165
+
166
+ retriever.close()
167
+
168
+ mock_store.close.assert_called_once_with()
169
+ assert retriever._document_store is mock_store
170
+
171
+
172
+ @pytest.mark.asyncio
173
+ async def test_close_async():
174
+ mock_store = Mock(spec=OpenSearchDocumentStore)
175
+ mock_store.close_async = AsyncMock()
176
+ retriever = OpenSearchBM25Retriever(document_store=mock_store)
177
+
178
+ await retriever.close_async()
179
+
180
+ mock_store.close_async.assert_awaited_once_with()
181
+ assert retriever._document_store is mock_store
182
+
183
+
162
184
  def test_run():
163
185
  mock_store = Mock(spec=OpenSearchDocumentStore)
164
186
  mock_store._bm25_retrieval.return_value = [Document(content="Test doc")]