nucliadb 6.9.1.post5192__py3-none-any.whl → 6.10.0.post5705__py3-none-any.whl

nucliadb/search/api/v1/catalog.py

@@ -19,7 +19,6 @@
 #
 import json
 from time import time
-from typing import Optional, Union
 
 from fastapi import Request, Response
 from fastapi_versioning import version
@@ -75,31 +74,28 @@ async def catalog_get(
     response: Response,
     kbid: str,
     query: str = fastapi_query(SearchParamDefaults.query),
-    filter_expression: Optional[str] = fastapi_query(SearchParamDefaults.catalog_filter_expression),
+    filter_expression: str | None = fastapi_query(SearchParamDefaults.catalog_filter_expression),
     filters: list[str] = fastapi_query(SearchParamDefaults.filters),
     faceted: list[str] = fastapi_query(SearchParamDefaults.faceted),
     sort_field: SortField = fastapi_query(SearchParamDefaults.sort_field),
-    sort_limit: Optional[int] = fastapi_query(SearchParamDefaults.sort_limit),
     sort_order: SortOrder = fastapi_query(SearchParamDefaults.sort_order),
     page_number: int = fastapi_query(SearchParamDefaults.catalog_page_number),
     page_size: int = fastapi_query(SearchParamDefaults.catalog_page_size),
-    with_status: Optional[ResourceProcessingStatus] = fastapi_query(
+    with_status: ResourceProcessingStatus | None = fastapi_query(
         SearchParamDefaults.with_status, deprecated="Use filters instead"
     ),
     debug: bool = fastapi_query(SearchParamDefaults.debug, include_in_schema=False),
-    range_creation_start: Optional[DateTime] = fastapi_query(SearchParamDefaults.range_creation_start),
-    range_creation_end: Optional[DateTime] = fastapi_query(SearchParamDefaults.range_creation_end),
-    range_modification_start: Optional[DateTime] = fastapi_query(
+    range_creation_start: DateTime | None = fastapi_query(SearchParamDefaults.range_creation_start),
+    range_creation_end: DateTime | None = fastapi_query(SearchParamDefaults.range_creation_end),
+    range_modification_start: DateTime | None = fastapi_query(
         SearchParamDefaults.range_modification_start
     ),
-    range_modification_end: Optional[DateTime] = fastapi_query(
-        SearchParamDefaults.range_modification_end
-    ),
-    hidden: Optional[bool] = fastapi_query(SearchParamDefaults.hidden),
+    range_modification_end: DateTime | None = fastapi_query(SearchParamDefaults.range_modification_end),
+    hidden: bool | None = fastapi_query(SearchParamDefaults.hidden),
     show: list[ResourceProperties] = fastapi_query(
         SearchParamDefaults.show, default=[ResourceProperties.BASIC, ResourceProperties.ERRORS]
     ),
-) -> Union[CatalogResponse, HTTPClientError]:
+) -> CatalogResponse | HTTPClientError:
     try:
         expr = (
             CatalogFilterExpression.model_validate_json(filter_expression) if filter_expression else None
@@ -125,7 +121,7 @@ async def catalog_get(
         show=show,
     )
     if sort_field:
-        item.sort = SortOptions(field=sort_field, limit=sort_limit, order=sort_order)
+        item.sort = SortOptions(field=sort_field, order=sort_order)
     return await catalog(kbid, item)
 
 
@@ -144,14 +140,14 @@ async def catalog_post(
     request: Request,
     kbid: str,
     item: CatalogRequest,
-) -> Union[CatalogResponse, HTTPClientError]:
+) -> CatalogResponse | HTTPClientError:
     return await catalog(kbid, item)
 
 
 async def catalog(
     kbid: str,
     item: CatalogRequest,
-) -> Union[HTTPClientError, CatalogResponse]:
+) -> HTTPClientError | CatalogResponse:
     """
     Catalog endpoint is a simplified version of the search endpoint, it only
     returns bm25 results on titles and it does not support vector search.

nucliadb/search/api/v1/find.py

@@ -18,7 +18,6 @@
 # along with this program. If not, see <http://www.gnu.org/licenses/>.
 #
 import json
-from typing import Optional, Union
 
 from fastapi import Body, Header, Query, Request, Response
 from fastapi.openapi.models import Example
@@ -46,7 +45,6 @@ from nucliadb_models.search import (
     KnowledgeboxFindResults,
     NucliaDBClientType,
     RankFusionName,
-    Reranker,
     RerankerName,
     ResourceProperties,
     SearchParamDefaults,
@@ -84,33 +82,31 @@ async def find_knowledgebox(
     response: Response,
     kbid: str,
     query: str = fastapi_query(SearchParamDefaults.query),
-    filter_expression: Optional[str] = fastapi_query(SearchParamDefaults.filter_expression),
+    filter_expression: str | None = fastapi_query(SearchParamDefaults.filter_expression),
     fields: list[str] = fastapi_query(SearchParamDefaults.fields),
     filters: list[str] = fastapi_query(SearchParamDefaults.filters),
-    top_k: Optional[int] = fastapi_query(SearchParamDefaults.top_k),
-    min_score: Optional[float] = Query(
+    top_k: int | None = fastapi_query(SearchParamDefaults.top_k),
+    min_score: float | None = Query(
         default=None,
-        description="Minimum similarity score to filter vector index results. If not specified, the default minimum score of the semantic model associated to the Knowledge Box will be used. Check out the documentation for more information on how to use this parameter: https://docs.nuclia.dev/docs/rag/advanced/search#minimum-score",  # noqa: E501
+        description="Minimum similarity score to filter vector index results. If not specified, the default minimum score of the semantic model associated to the Knowledge Box will be used. Check out the documentation for more information on how to use this parameter: https://docs.nuclia.dev/docs/rag/advanced/search#minimum-score",
         deprecated=True,
     ),
-    min_score_semantic: Optional[float] = Query(
+    min_score_semantic: float | None = Query(
         default=None,
-        description="Minimum semantic similarity score to filter vector index results. If not specified, the default minimum score of the semantic model associated to the Knowledge Box will be used. Check out the documentation for more information on how to use this parameter: https://docs.nuclia.dev/docs/rag/advanced/search#minimum-score",  # noqa: E501
+        description="Minimum semantic similarity score to filter vector index results. If not specified, the default minimum score of the semantic model associated to the Knowledge Box will be used. Check out the documentation for more information on how to use this parameter: https://docs.nuclia.dev/docs/rag/advanced/search#minimum-score",
     ),
     min_score_bm25: float = Query(
         default=0,
         description="Minimum bm25 score to filter paragraph and document index results",
         ge=0,
     ),
-    vectorset: Optional[str] = fastapi_query(SearchParamDefaults.vectorset),
-    range_creation_start: Optional[DateTime] = fastapi_query(SearchParamDefaults.range_creation_start),
-    range_creation_end: Optional[DateTime] = fastapi_query(SearchParamDefaults.range_creation_end),
-    range_modification_start: Optional[DateTime] = fastapi_query(
+    vectorset: str | None = fastapi_query(SearchParamDefaults.vectorset),
+    range_creation_start: DateTime | None = fastapi_query(SearchParamDefaults.range_creation_start),
+    range_creation_end: DateTime | None = fastapi_query(SearchParamDefaults.range_creation_end),
+    range_modification_start: DateTime | None = fastapi_query(
         SearchParamDefaults.range_modification_start
     ),
-    range_modification_end: Optional[DateTime] = fastapi_query(
-        SearchParamDefaults.range_modification_end
-    ),
+    range_modification_end: DateTime | None = fastapi_query(SearchParamDefaults.range_modification_end),
     features: list[FindOptions] = fastapi_query(
         SearchParamDefaults.search_features,
         default=[
@@ -127,19 +123,18 @@ async def find_knowledgebox(
     extracted: list[ExtractedDataTypeName] = fastapi_query(SearchParamDefaults.extracted),
     with_duplicates: bool = fastapi_query(SearchParamDefaults.with_duplicates),
     with_synonyms: bool = fastapi_query(SearchParamDefaults.with_synonyms),
-    autofilter: bool = fastapi_query(SearchParamDefaults.autofilter),
     security_groups: list[str] = fastapi_query(SearchParamDefaults.security_groups),
     show_hidden: bool = fastapi_query(SearchParamDefaults.show_hidden),
     rank_fusion: RankFusionName = fastapi_query(SearchParamDefaults.rank_fusion),
-    reranker: Union[RerankerName, Reranker] = fastapi_query(SearchParamDefaults.reranker),
-    search_configuration: Optional[str] = Query(
+    reranker: RerankerName = fastapi_query(SearchParamDefaults.reranker),
+    search_configuration: str | None = Query(
         default=None,
         description="Load find parameters from this configuration. Parameters in the request override parameters from the configuration.",
     ),
     x_ndb_client: NucliaDBClientType = Header(NucliaDBClientType.API),
     x_nucliadb_user: str = Header(""),
     x_forwarded_for: str = Header(""),
-) -> Union[KnowledgeboxFindResults, HTTPClientError]:
+) -> KnowledgeboxFindResults | HTTPClientError:
     try:
         expr = FilterExpression.model_validate_json(filter_expression) if filter_expression else None
 
@@ -166,7 +161,6 @@ async def find_knowledgebox(
             extracted=extracted,
             with_duplicates=with_duplicates,
             with_synonyms=with_synonyms,
-            autofilter=autofilter,
             security=security,
             show_hidden=show_hidden,
             rank_fusion=rank_fusion,
@@ -198,7 +192,7 @@ async def find_post_knowledgebox(
     x_ndb_client: NucliaDBClientType = Header(NucliaDBClientType.API),
     x_nucliadb_user: str = Header(""),
     x_forwarded_for: str = Header(""),
-) -> Union[KnowledgeboxFindResults, HTTPClientError]:
+) -> KnowledgeboxFindResults | HTTPClientError:
     return await _find_endpoint(response, kbid, item, x_ndb_client, x_nucliadb_user, x_forwarded_for)
 
 
@@ -209,7 +203,7 @@ async def _find_endpoint(
     x_ndb_client: NucliaDBClientType,
     x_nucliadb_user: str,
     x_forwarded_for: str,
-) -> Union[KnowledgeboxFindResults, HTTPClientError]:
+) -> KnowledgeboxFindResults | HTTPClientError:
     if item.search_configuration is not None:
         search_config = await datamanagers.atomic.search_configurations.get(
             kbid=kbid, name=item.search_configuration

nucliadb/search/api/v1/hydrate.py

@@ -18,7 +18,7 @@
 # along with this program. If not, see <http://www.gnu.org/licenses/>.
 #
 import asyncio
-from typing import Awaitable, Optional, Union
+from collections.abc import Awaitable
 
 from async_lru import alru_cache
 from fastapi import Request, Response
@@ -81,13 +81,13 @@ class HydratedBuilder:
         self._resources: dict[str, HydratedResource] = {}
         self._fields: dict[
             str,
-            Union[
-                HydratedTextField,
-                HydratedFileField,
-                HydratedLinkField,
-                HydratedConversationField,
-                HydratedGenericField,
-            ],
+            (
+                HydratedTextField
+                | HydratedFileField
+                | HydratedLinkField
+                | HydratedConversationField
+                | HydratedGenericField
+            ),
         ] = {}
         self._paragraphs: dict[str, HydratedParagraph] = {}
 
@@ -100,13 +100,13 @@ class HydratedBuilder:
         self,
     ) -> dict[
         str,
-        Union[
-            HydratedTextField,
-            HydratedFileField,
-            HydratedLinkField,
-            HydratedConversationField,
-            HydratedGenericField,
-        ],
+        (
+            HydratedTextField
+            | HydratedFileField
+            | HydratedLinkField
+            | HydratedConversationField
+            | HydratedGenericField
+        ),
     ]:
         return self._fields
 
@@ -127,13 +127,13 @@ class HydratedBuilder:
     def add_field(
         self,
         field_id: FieldId,
-        field: Union[
-            HydratedTextField,
-            HydratedFileField,
-            HydratedLinkField,
-            HydratedConversationField,
-            HydratedGenericField,
-        ],
+        field: (
+            HydratedTextField
+            | HydratedFileField
+            | HydratedLinkField
+            | HydratedConversationField
+            | HydratedGenericField
+        ),
     ):
         self._fields[field_id.full()] = field
 
@@ -233,7 +233,7 @@ class Hydrator:
 
             if field_id not in field_tasks:
                 field_tasks[field_id] = asyncio.create_task(
-                    self._limited_concurrency(hydrate_field(resource, field_id, self.config.field))
+                    self._limited_concurrency(hydrate_field(field, field_id, self.config.field))
                 )
 
             if rid not in resource_tasks:
@@ -323,6 +323,6 @@ class Hydrator:
         async with self.max_ops:
             return await aw
 
-    @alru_cache(maxsize=None)
-    async def cached_download_page_preview(self, field: Field, page: int) -> Optional[Image]:
+    @alru_cache(maxsize=50)
+    async def cached_download_page_preview(self, field: Field, page: int) -> Image | None:
         return await download_page_preview(field, page)

nucliadb/search/api/v1/knowledgebox.py

@@ -18,7 +18,6 @@
 # along with this program. If not, see <http://www.gnu.org/licenses/>.
 #
 import asyncio
-from typing import Optional
 
 from fastapi import HTTPException, Request
 from fastapi_versioning import version
@@ -178,7 +177,7 @@ async def get_node_index_counts(kbid: str) -> tuple[IndexCounts, list[str]]:
         )
 
     try:
-        results: Optional[list[Shard]] = await asyncio.wait_for(
+        results: list[Shard] | None = await asyncio.wait_for(
             asyncio.gather(*ops, return_exceptions=True),  # type: ignore
             timeout=settings.search_timeout,
         )

nucliadb/search/api/v1/predict_proxy.py

@@ -18,7 +18,6 @@
 # along with this program. If not, see <http://www.gnu.org/licenses/>.
 #
 import json
-from typing import Union
 
 from fastapi import Header, Request
 from fastapi.responses import Response, StreamingResponse
@@ -68,7 +67,7 @@ async def predict_proxy_endpoint(
     x_nucliadb_user: str = Header(""),
     x_ndb_client: NucliaDBClientType = Header(NucliaDBClientType.API),
     x_forwarded_for: str = Header(""),
-) -> Union[Response, StreamingResponse, HTTPClientError]:
+) -> Response | StreamingResponse | HTTPClientError:
     try:
         payload = await request.json()
     except json.JSONDecodeError:

nucliadb/search/api/v1/resource/ask.py

@@ -17,14 +17,14 @@
 # You should have received a copy of the GNU Affero General Public License
 # along with this program. If not, see <http://www.gnu.org/licenses/>.
 #
-from typing import Union
+from uuid import UUID
 
 from fastapi import Header, Request, Response
 from fastapi_versioning import version
 from starlette.responses import StreamingResponse
 
+from nucliadb.common import datamanagers
 from nucliadb.models.responses import HTTPClientError
-from nucliadb.search.api.v1.resource.utils import get_resource_uuid_by_slug
 from nucliadb.search.api.v1.router import KB_PREFIX, RESOURCE_SLUG_PREFIX, api
 from nucliadb_models.resource import NucliaDBRoles
 from nucliadb_models.search import AskRequest, NucliaDBClientType, SyncAskResponse
@@ -47,7 +47,7 @@ from ..ask import create_ask_response
 async def resource_ask_endpoint_by_uuid(
     request: Request,
     kbid: str,
-    rid: str,
+    rid: UUID,
     item: AskRequest,
     x_show_consumption: bool = Header(default=False),
     x_ndb_client: NucliaDBClientType = Header(NucliaDBClientType.API),
@@ -58,7 +58,7 @@ async def resource_ask_endpoint_by_uuid(
         description="When set to true, outputs response as JSON in a non-streaming way. "
         "This is slower and requires waiting for entire answer to be ready.",
     ),
-) -> Union[StreamingResponse, HTTPClientError, Response]:
+) -> StreamingResponse | HTTPClientError | Response:
     current_user: NucliaUser = request.user
     # If present, security groups from AuthorizationBackend overrides any
     # security group of the payload
@@ -75,7 +75,7 @@ async def resource_ask_endpoint_by_uuid(
         client_type=x_ndb_client,
         origin=x_forwarded_for,
         x_synchronous=x_synchronous,
-        resource=rid,
+        resource=str(rid),
         extra_predict_headers={"X-Show-Consumption": str(x_show_consumption).lower()},
     )
 
@@ -104,8 +104,8 @@ async def resource_ask_endpoint_by_slug(
         description="When set to true, outputs response as JSON in a non-streaming way. "
         "This is slower and requires waiting for entire answer to be ready.",
     ),
-) -> Union[StreamingResponse, HTTPClientError, Response]:
-    resource_id = await get_resource_uuid_by_slug(kbid, slug)
+) -> StreamingResponse | HTTPClientError | Response:
+    resource_id = await datamanagers.atomic.resources.get_resource_uuid_from_slug(kbid=kbid, slug=slug)
     if resource_id is None:
         return HTTPClientError(status_code=404, detail="Resource not found")
 

nucliadb/search/api/v1/resource/ingestion_agents.py

@@ -17,14 +17,13 @@
 # You should have received a copy of the GNU Affero General Public License
 # along with this program. If not, see <http://www.gnu.org/licenses/>.
 #
-from typing import Union
 
 from fastapi import Header, Request, Response
 from fastapi_versioning import version
 
+from nucliadb.common import datamanagers
 from nucliadb.common.models_utils import from_proto
 from nucliadb.models.responses import HTTPClientError
-from nucliadb.search.api.v1.resource.utils import get_resource_uuid_by_slug
 from nucliadb.search.api.v1.router import KB_PREFIX, RESOURCE_PREFIX, RESOURCE_SLUG_PREFIX, api
 from nucliadb.search.predict_models import AugmentedField, RunAgentsResponse
 from nucliadb.search.search.exceptions import ResourceNotFoundError
@@ -58,7 +57,7 @@ async def run_agents_by_uuid(
     rid: str,
     item: ResourceAgentsRequest,
     x_nucliadb_user: str = Header(""),
-) -> Union[ResourceAgentsResponse, HTTPClientError]:
+) -> ResourceAgentsResponse | HTTPClientError:
     return await _run_agents_endpoint(kbid, rid, x_nucliadb_user, item)
 
 
@@ -80,8 +79,8 @@ async def run_agents_by_slug(
     slug: str,
     item: ResourceAgentsRequest,
     x_nucliadb_user: str = Header(""),
-) -> Union[ResourceAgentsResponse, HTTPClientError]:
-    resource_id = await get_resource_uuid_by_slug(kbid, slug)
+) -> ResourceAgentsResponse | HTTPClientError:
+    resource_id = await datamanagers.atomic.resources.get_resource_uuid_from_slug(kbid=kbid, slug=slug)
     if resource_id is None:
         return HTTPClientError(status_code=404, detail="Resource not found")
     return await _run_agents_endpoint(kbid, resource_id, x_nucliadb_user, item)
@@ -89,7 +88,7 @@ async def run_agents_by_slug(
 
 async def _run_agents_endpoint(
     kbid: str, resource_id: str, user_id: str, item: ResourceAgentsRequest
-) -> Union[ResourceAgentsResponse, HTTPClientError]:
+) -> ResourceAgentsResponse | HTTPClientError:
     try:
         run_agents_response: RunAgentsResponse = await run_agents(
             kbid,

nucliadb/search/api/v1/resource/search.py

@@ -18,7 +18,7 @@
 # along with this program. If not, see <http://www.gnu.org/licenses/>.
 #
 import json
-from typing import Optional, Union, cast
+from typing import cast
 
 from fastapi import Header, Request, Response
 from fastapi_versioning import version
@@ -62,25 +62,23 @@ async def resource_search(
     kbid: str,
     query: str,
     rid: str,
-    filter_expression: Optional[str] = fastapi_query(SearchParamDefaults.filter_expression),
+    filter_expression: str | None = fastapi_query(SearchParamDefaults.filter_expression),
     fields: list[str] = fastapi_query(SearchParamDefaults.fields),
     filters: list[str] = fastapi_query(SearchParamDefaults.filters),
     faceted: list[str] = fastapi_query(SearchParamDefaults.faceted),
-    sort: Optional[SortField] = fastapi_query(SearchParamDefaults.sort_field, alias="sort_field"),
+    sort: SortField | None = fastapi_query(SearchParamDefaults.sort_field, alias="sort_field"),
     sort_order: SortOrder = fastapi_query(SearchParamDefaults.sort_order),
-    top_k: Optional[int] = fastapi_query(SearchParamDefaults.top_k),
-    range_creation_start: Optional[DateTime] = fastapi_query(SearchParamDefaults.range_creation_start),
-    range_creation_end: Optional[DateTime] = fastapi_query(SearchParamDefaults.range_creation_end),
-    range_modification_start: Optional[DateTime] = fastapi_query(
+    top_k: int | None = fastapi_query(SearchParamDefaults.top_k),
+    range_creation_start: DateTime | None = fastapi_query(SearchParamDefaults.range_creation_start),
+    range_creation_end: DateTime | None = fastapi_query(SearchParamDefaults.range_creation_end),
+    range_modification_start: DateTime | None = fastapi_query(
         SearchParamDefaults.range_modification_start
     ),
-    range_modification_end: Optional[DateTime] = fastapi_query(
-        SearchParamDefaults.range_modification_end
-    ),
+    range_modification_end: DateTime | None = fastapi_query(SearchParamDefaults.range_modification_end),
     highlight: bool = fastapi_query(SearchParamDefaults.highlight),
     x_ndb_client: NucliaDBClientType = Header(NucliaDBClientType.API),
     debug: bool = fastapi_query(SearchParamDefaults.debug),
-) -> Union[ResourceSearchResults, HTTPClientError]:
+) -> ResourceSearchResults | HTTPClientError:
     top_k = top_k or SearchParamDefaults.top_k  # type: ignore
     top_k = cast(int, top_k)
 

nucliadb/search/api/v1/retrieve.py

@@ -0,0 +1,130 @@
+# Copyright (C) 2021 Bosutech XXI S.L.
+#
+# nucliadb is offered under the AGPL v3.0 and as commercial software.
+# For commercial licensing, contact us at info@nuclia.com.
+#
+# AGPL:
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as
+# published by the Free Software Foundation, either version 3 of the
+# License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU Affero General Public License for more details.
+#
+# You should have received a copy of the GNU Affero General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>.
+#
+
+from time import time
+
+from fastapi import Header, HTTPException, Request
+from fastapi_versioning import version
+
+from nucliadb.common.exceptions import InvalidQueryError
+from nucliadb.common.external_index_providers.base import TextBlockMatch
+from nucliadb.common.models_utils import to_proto
+from nucliadb.search.api.v1.router import KB_PREFIX, api
+from nucliadb.search.search.query_parser.parsers.retrieve import parse_retrieve
+from nucliadb.search.search.retrieval import text_block_search
+from nucliadb_models.resource import NucliaDBRoles
+from nucliadb_models.retrieval import (
+    Metadata,
+    RetrievalMatch,
+    RetrievalRequest,
+    RetrievalResponse,
+    Scores,
+)
+from nucliadb_models.search import NucliaDBClientType
+from nucliadb_utils.authentication import requires
+from nucliadb_utils.utilities import get_audit
+
+
+@api.post(
+    f"/{KB_PREFIX}/{{kbid}}/retrieve",
+    status_code=200,
+    description="Search text blocks on a Knowledge Box",
+    include_in_schema=False,
+    tags=["Search"],
+)
+@requires(NucliaDBRoles.READER)
+@version(1)
+async def _retrieve_endpoint(
+    request: Request,
+    kbid: str,
+    item: RetrievalRequest,
+    x_ndb_client: NucliaDBClientType = Header(NucliaDBClientType.API),
+    x_nucliadb_user: str = Header(""),
+    x_forwarded_for: str = Header(""),
+) -> RetrievalResponse:
+    return await retrieve_endpoint(
+        kbid,
+        item,
+        x_ndb_client=x_ndb_client,
+        x_nucliadb_user=x_nucliadb_user,
+        x_forwarded_for=x_forwarded_for,
+    )
+
+
+async def retrieve_endpoint(
+    kbid: str,
+    item: RetrievalRequest,
+    *,
+    x_ndb_client: NucliaDBClientType,
+    x_nucliadb_user: str,
+    x_forwarded_for: str,
+) -> RetrievalResponse:
+    audit = get_audit()
+    start_time = time()
+
+    try:
+        retrieval = await parse_retrieve(kbid, item)
+    except InvalidQueryError as err:
+        raise HTTPException(
+            status_code=422,
+            detail=str(err),
+        )
+
+    text_blocks, pb_query, _, _ = await text_block_search(kbid, retrieval)
+
+    # cut the top K, we may have more due to extra results used for rank fusion
+    text_blocks = text_blocks[: retrieval.top_k]
+
+    # convert to response models
+    matches = [text_block_match_to_retrieval_match(text_block) for text_block in text_blocks]
+
+    if audit is not None:
+        retrieval_time = time() - start_time
+        audit.retrieve(
+            kbid,
+            x_nucliadb_user,
+            to_proto.client_type(x_ndb_client),
+            x_forwarded_for,
+            retrieval_time,
+            # TODO(decoupled-ask): add interesting things to audit
+        )
+
+    return RetrievalResponse(matches=matches)
+
+
+def text_block_match_to_retrieval_match(item: TextBlockMatch) -> RetrievalMatch:
+    return RetrievalMatch(
+        id=item.paragraph_id.full(),
+        score=Scores(
+            value=item.current_score.score,
+            source=item.current_score.source,
+            type=item.current_score.type,
+            history=item.scores,
+        ),
+        metadata=Metadata(
+            field_labels=item.field_labels,
+            paragraph_labels=item.paragraph_labels,
+            is_an_image=item.is_an_image,
+            is_a_table=item.is_a_table,
+            source_file=item.representation_file,
+            page=item.position.page_number,
+            in_page_with_visual=item.page_with_visual,
+        ),
+    )