nucliadb 6.2.1.post3312__py3-none-any.whl → 6.2.1.post3326__py3-none-any.whl

nucliadb/common/models_utils/from_proto.py

@@ -23,12 +23,13 @@ from typing import Any
 
 from google.protobuf.json_format import MessageToDict
 
-from nucliadb_models.common import Classification, FieldID, FieldTypeName
+from nucliadb_models.common import Classification, FieldID, FieldTypeName, QuestionAnswers
 from nucliadb_models.conversation import Conversation, FieldConversation
 from nucliadb_models.entities import EntitiesGroup, EntitiesGroupSummary, Entity
 from nucliadb_models.extracted import (
     ExtractedText,
     FieldComputedMetadata,
+    FieldMetadata,
     FieldQuestionAnswers,
     FileExtractedData,
     LargeComputedMetadata,
@@ -236,6 +237,15 @@ def field_question_answers(
     return FieldQuestionAnswers(**value)
 
 
+def question_answers(message: resources_pb2.QuestionAnswers) -> QuestionAnswers:
+    value = MessageToDict(
+        message,
+        preserving_proto_field_name=True,
+        including_default_value_fields=True,
+    )
+    return QuestionAnswers(**value)
+
+
 def extracted_text(message: resources_pb2.ExtractedText) -> ExtractedText:
     return ExtractedText(
         **MessageToDict(
@@ -304,10 +314,9 @@ def field_computed_metadata(
 ) -> FieldComputedMetadata:
     if shortened:
         shorten_fieldmetadata(message)
-    metadata = convert_fieldmetadata_pb_to_dict(message.metadata)
+    metadata = field_metadata(message.metadata)
     split_metadata = {
-        split: convert_fieldmetadata_pb_to_dict(metadata_split)
-        for split, metadata_split in message.split_metadata.items()
+        split: field_metadata(metadata_split) for split, metadata_split in message.split_metadata.items()
     }
     value = MessageToDict(
         message,
@@ -319,9 +328,9 @@ def field_computed_metadata(
     return FieldComputedMetadata(**value)
 
 
-def convert_fieldmetadata_pb_to_dict(
+def field_metadata(
     message: resources_pb2.FieldMetadata,
-) -> dict[str, Any]:
+) -> FieldMetadata:
     # Backwards compatibility with old entities format
     # TODO: Remove once deprecated fields are removed
     # If we recieved processor entities in the new field and the old field is empty, we copy them to the old field
@@ -347,7 +356,7 @@ def convert_fieldmetadata_pb_to_dict(
     value["relations"] = [
         convert_pb_relation_to_api(rel) for relations in message.relations for rel in relations.relations
     ]
-    return value
+    return FieldMetadata(**value)
 
 
 def conversation(message: resources_pb2.Conversation) -> Conversation:

nucliadb/search/api/v1/__init__.py

@@ -28,4 +28,5 @@ from . import suggest  # noqa
 from . import summarize  # noqa
 from .resource import ask as ask_resource  # noqa
 from .resource import search as search_resource  # noqa
+from .resource import ingestion_agents as ingestion_agents_resource  # noqa
 from .router import api  # noqa

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 Optional, Union
+from typing import Union
 
 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
@@ -104,8 +104,3 @@ async def resource_ask_endpoint_by_slug(
         x_synchronous,
         resource=resource_id,
     )
-
-
-async def get_resource_uuid_by_slug(kbid: str, slug: str) -> Optional[str]:
-    async with datamanagers.with_ro_transaction() as txn:
-        return await datamanagers.resources.get_resource_uuid_from_slug(txn, kbid=kbid, slug=slug)

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

@@ -0,0 +1,123 @@
+# 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 typing import Union
+
+from fastapi import Header, Request, Response
+from fastapi_versioning import version
+
+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
+from nucliadb.search.search.ingestion_agents import run_agents
+from nucliadb_models.agents.ingestion import (
+    AppliedDataAugmentation,
+    NewTextField,
+    ResourceAgentsRequest,
+    ResourceAgentsResponse,
+)
+from nucliadb_models.agents.ingestion import AugmentedField as PublicAugmentedField
+from nucliadb_models.resource import NucliaDBRoles
+from nucliadb_utils.authentication import requires_one
+
+
+@api.post(
+    f"/{KB_PREFIX}/{{kbid}}/{RESOURCE_PREFIX}/{{rid}}/run-agents",
+    status_code=200,
+    summary="Run Agents on Resource",
+    description="Run Agents on Resource",
+    tags=["Ingestion Agents"],
+    response_model_exclude_unset=True,
+    response_model=ResourceAgentsResponse,
+)
+@requires_one([NucliaDBRoles.READER])
+@version(1)
+async def run_agents_by_uuid(
+    request: Request,
+    response: Response,
+    kbid: str,
+    rid: str,
+    item: ResourceAgentsRequest,
+    x_nucliadb_user: str = Header(""),
+) -> Union[ResourceAgentsResponse, HTTPClientError]:
+    return await _run_agents_endpoint(kbid, rid, x_nucliadb_user, item)
+
+
+@api.post(
+    f"/{KB_PREFIX}/{{kbid}}/{RESOURCE_SLUG_PREFIX}/{{slug}}/run-agents",
+    status_code=200,
+    summary="Run Agents on Resource (by slug)",
+    description="Run Agents on Resource (by slug)",
+    tags=["Ingestion Agents"],
+    response_model_exclude_unset=True,
+    response_model=ResourceAgentsResponse,
+)
+@requires_one([NucliaDBRoles.READER])
+@version(1)
+async def run_agents_by_slug(
+    request: Request,
+    response: Response,
+    kbid: str,
+    slug: str,
+    item: ResourceAgentsRequest,
+    x_nucliadb_user: str = Header(""),
+) -> Union[ResourceAgentsResponse, HTTPClientError]:
+    resource_id = await get_resource_uuid_by_slug(kbid, 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)
+
+
+async def _run_agents_endpoint(
+    kbid: str, resource_id: str, user_id: str, item: ResourceAgentsRequest
+) -> Union[ResourceAgentsResponse, HTTPClientError]:
+    try:
+        run_agents_response: RunAgentsResponse = await run_agents(
+            kbid, resource_id, user_id, filters=item.filters
+        )
+    except ResourceNotFoundError:
+        return HTTPClientError(status_code=404, detail="Resource not found")
+    response = ResourceAgentsResponse(results={})
+    for field_id, augmented_field in run_agents_response.results.items():
+        response.results[field_id] = _parse_augmented_field(augmented_field)
+    return response
+
+
+def _parse_augmented_field(augmented_field: AugmentedField) -> PublicAugmentedField:
+    ada = augmented_field.applied_data_augmentation
+    return PublicAugmentedField(
+        metadata=from_proto.field_metadata(augmented_field.metadata),
+        applied_data_augmentation=AppliedDataAugmentation(
+            changed=ada.changed,
+            qas=from_proto.question_answers(ada.qas) if ada.qas else None,
+            new_text_fields=[
+                NewTextField(
+                    text_field=from_proto.field_text(ntf["text_field"]),
+                    destination=ntf["destination"],
+                )
+                for ntf in ada.new_text_fields
+            ],
+        ),
+        input_nuclia_tokens=augmented_field.input_nuclia_tokens,
+        output_nuclia_tokens=augmented_field.output_nuclia_tokens,
+        time=augmented_field.time,
+    )

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

@@ -0,0 +1,28 @@
+# 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 typing import Optional
+
+from nucliadb.common import datamanagers
+
+
+async def get_resource_uuid_by_slug(kbid: str, slug: str) -> Optional[str]:
+    async with datamanagers.with_ro_transaction() as txn:
+        return await datamanagers.resources.get_resource_uuid_from_slug(txn, kbid=kbid, slug=slug)

nucliadb/search/predict.py

@@ -17,6 +17,7 @@
 # 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/>.
 #
+import base64
 import json
 import os
 import random
@@ -31,6 +32,12 @@ from pydantic import ValidationError
 
 from nucliadb.common import datamanagers
 from nucliadb.search import logger
+from nucliadb.search.predict_models import (
+    AppliedDataAugmentation,
+    AugmentedField,
+    RunAgentsRequest,
+    RunAgentsResponse,
+)
 from nucliadb.tests.vectors import Q, Qm2023
 from nucliadb_models.internal.predict import (
     Ner,
@@ -47,6 +54,7 @@ from nucliadb_models.search import (
     SummarizedResponse,
     SummarizeModel,
 )
+from nucliadb_protos.resources_pb2 import FieldMetadata
 from nucliadb_protos.utils_pb2 import RelationNode
 from nucliadb_telemetry import errors, metrics
 from nucliadb_utils.exceptions import LimitsExceededError
@@ -401,6 +409,24 @@ class PredictEngine:
         data = await resp.json()
         return RerankResponse.model_validate(data)
 
+    @predict_observer.wrap({"type": "run_agents"})
+    async def run_agents(self, kbid: str, item: RunAgentsRequest) -> RunAgentsResponse:
+        try:
+            self.check_nua_key_is_configured_for_onprem()
+        except NUAKeyMissingError:
+            error = "Nuclia Service account is not defined. Summarize operation could not be performed"
+            logger.warning(error)
+            raise SendToPredictError(error)
+        resp = await self.make_request(
+            "POST",
+            url=self.get_predict_url("/run-agents", kbid),
+            json=item.model_dump(),
+            headers=self.get_predict_headers(kbid),
+        )
+        await self.check_response(resp, expected_status=200)
+        data = await resp.json()
+        return RunAgentsResponse.model_validate(data)
+
 
 class DummyPredictEngine(PredictEngine):
     default_semantic_threshold = 0.7
@@ -529,6 +555,21 @@ class DummyPredictEngine(PredictEngine):
         )
         return response
 
+    async def run_agents(self, kbid: str, item: RunAgentsRequest) -> RunAgentsResponse:
+        self.calls.append(("run_agents", (kbid, item)))
+        fm = FieldMetadata()
+        ada = AppliedDataAugmentation()
+        serialized_fm = base64.b64encode(fm.SerializeToString()).decode("utf-8")
+        augmented_field = AugmentedField(
+            metadata=serialized_fm,  # type: ignore
+            applied_data_augmentation=ada,
+            input_nuclia_tokens=1.0,
+            output_nuclia_tokens=1.0,
+            time=1.0,
+        )
+        response = RunAgentsResponse(results={"field_id": augmented_field})
+        return response
+
 
 def get_answer_generator(response: aiohttp.ClientResponse):
     """

nucliadb/search/predict_models.py

@@ -0,0 +1,160 @@
+# 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 base64 import b64decode, b64encode
+from enum import Enum
+from typing import Optional, TypedDict
+
+from google.protobuf.message import DecodeError, Message
+from pydantic import BaseModel, ConfigDict, Field, field_validator
+
+from nucliadb_protos.resources_pb2 import FieldMetadata, FieldText, QuestionAnswers
+
+
+class FieldInfo(BaseModel):
+    """
+    Model to represent the field information required
+    """
+
+    text: str = Field(..., title="The text of the field")
+    metadata: str = Field(
+        title="The metadata of the field as a base64 string serialized nucliadb_protos.resources.FieldMetadata protobuf",
+    )
+    field_id: str = Field(
+        ...,
+        title="The field ID of the field (rid/field_type/field[/split]) or any unique identifier",
+    )
+
+
+class OperationType(str, Enum):
+    graph = "graph"
+    label = "label"
+    ask = "ask"
+    qa = "qa"
+    extract = "extract"
+    prompt_guard = "prompt_guard"
+    llama_guard = "llama_guard"
+
+
+class NameOperationFilter(BaseModel):
+    operation_type: OperationType = Field(..., description="Type of the operation")
+    task_names: list[str] = Field(
+        default_factory=list,
+        description="list of task names. If None or empty, all tasks for that operation are applied.",
+    )
+
+
+class RunAgentsRequest(BaseModel):
+    """
+    Model to represent a request for the Augment model
+    The text will be augmented with the Knowledge Box's configured Data Augmentation Agents
+    """
+
+    fields: list[FieldInfo] = Field(
+        ...,
+        title="The fields to be augmented with the Knowledge Box's Data Augmentation Agents",
+    )
+    user_id: str = Field(..., title="The user ID of the user making the request")
+    filters: Optional[list[NameOperationFilter]] = Field(
+        default=None,
+        title="Filters to select which Data Augmentation Agents are applied to the text. If empty, all configured agents for the Knowledge Box are applied.",
+    )
+
+
+class NewTextField(TypedDict):
+    text_field: FieldText
+    destination: str
+
+
+class AppliedDataAugmentation(BaseModel):
+    model_config = ConfigDict(
+        # Since we have protos as fields, we need to enable arbitrary_types_allowed
+        arbitrary_types_allowed=True,
+    )
+    qas: Optional[QuestionAnswers] = Field(
+        default=None,
+        description="Question and answers generated by the Question Answers agent",
+    )
+    new_text_fields: list[NewTextField] = Field(
+        default_factory=list,
+        description="New text fields. Only generated by the Labeler agent as of now.",
+    )
+    changed: bool = Field(
+        default=True,
+        description="Indicates if the FieldMetadata was changed by the agents",
+    )
+
+    @field_validator("qas", mode="before")
+    def validate_qas(cls, qas: Optional[str]) -> Optional[QuestionAnswers]:
+        if qas is None:
+            return None
+        try:
+            return QuestionAnswers.FromString(b64decode(qas))
+        except DecodeError:
+            raise ValueError("Invalid QuestionAnswers protobuf")
+
+    @field_validator("new_text_fields", mode="before")
+    def validate_new_text_fields(cls, new_text_fields: list[NewTextField]) -> list[NewTextField]:
+        try:
+            return [
+                NewTextField(
+                    text_field=FieldText.FromString(b64decode(text_field["text_field"])),  # type: ignore
+                    destination=text_field["destination"],
+                )
+                for text_field in new_text_fields
+            ]
+        except DecodeError:
+            raise ValueError("Invalid NewTextField value")
+
+
+class AugmentedField(BaseModel):
+    # Since we have protos as fields, we need to enable arbitrary_types_allowed
+    model_config = ConfigDict(
+        arbitrary_types_allowed=True,
+        # Custom encoding to be able to serialize protobuf messages
+        json_encoders={Message: lambda m: b64encode(m.SerializeToString()).decode()},
+    )
+    metadata: FieldMetadata = Field(
+        ...,
+        title="The updated metadata of the field as a base64 string serialized nucliadb_protos.resources.FieldMetadata protobuf",
+    )
+    applied_data_augmentation: AppliedDataAugmentation = Field(
+        ..., title="The results of the Applied Data Augmentation"
+    )
+    input_nuclia_tokens: float = Field(
+        ..., title="The number of input Nuclia tokens consumed for the field"
+    )
+    output_nuclia_tokens: float = Field(
+        ..., title="The number of output Nuclia tokens consumed for the field"
+    )
+    time: float = Field(..., title="The time taken to execute the Data Augmentation agents to the field")
+
+    @field_validator("metadata", mode="before")
+    def validate_metadata(cls, metadata: str) -> FieldMetadata:
+        try:
+            return FieldMetadata.FromString(b64decode(metadata))
+        except DecodeError:
+            raise ValueError("Invalid FieldMetadata protobuf")
+
+
+class RunAgentsResponse(BaseModel):
+    results: dict[str, AugmentedField] = Field(
+        ...,
+        title="Pairs of augmented FieldMetadata and Data Augmentation results by field id",
+    )

nucliadb/search/search/ingestion_agents.py

@@ -0,0 +1,88 @@
+# 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/>.
+#
+import asyncio
+from base64 import b64encode
+from typing import Optional
+
+from nucliadb.common import datamanagers
+from nucliadb.ingest.fields.base import Field
+from nucliadb.search.predict_models import (
+    FieldInfo,
+    NameOperationFilter,
+    OperationType,
+    RunAgentsRequest,
+    RunAgentsResponse,
+)
+from nucliadb.search.search.exceptions import ResourceNotFoundError
+from nucliadb.search.utilities import get_predict
+from nucliadb_models.agents.ingestion import AgentsFilter
+from nucliadb_protos.resources_pb2 import FieldMetadata
+
+
+async def run_agents(
+    kbid: str, rid: str, user_id: str, filters: Optional[list[AgentsFilter]] = None
+) -> RunAgentsResponse:
+    fields = await fetch_resource_fields(kbid, rid)
+
+    item = RunAgentsRequest(user_id=user_id, filters=_parse_filters(filters), fields=fields)
+
+    predict = get_predict()
+    return await predict.run_agents(kbid, item)
+
+
+def _parse_filters(filters: Optional[list[AgentsFilter]]) -> Optional[list[NameOperationFilter]]:
+    if filters is None:
+        return None
+    return [
+        NameOperationFilter(
+            operation_type=OperationType(filter.type.value), task_names=filter.task_names
+        )
+        for filter in filters
+    ]
+
+
+async def fetch_resource_fields(kbid: str, rid: str) -> list[FieldInfo]:
+    async with datamanagers.with_ro_transaction() as txn:
+        resource = await datamanagers.resources.get_resource(txn, kbid=kbid, rid=rid)
+        if resource is None:
+            raise ResourceNotFoundError()
+        fields = await resource.get_fields(force=True)
+
+    tasks: list[asyncio.Task] = []
+    for field in fields.values():
+        tasks.append(asyncio.create_task(_hydrate_field_info(field)))
+    await asyncio.gather(*tasks)
+
+    return [task.result() for task in tasks]
+
+
+async def _hydrate_field_info(field_obj: Field) -> FieldInfo:
+    extracted_text = await field_obj.get_extracted_text(force=True)
+    fcmw = await field_obj.get_field_metadata(force=True)
+    if fcmw is None:
+        metadata = FieldMetadata()
+    else:
+        metadata = fcmw.metadata
+    serialized_metadata = b64encode(metadata.SerializeToString()).decode()
+    return FieldInfo(
+        text=extracted_text.text if extracted_text is not None else "",
+        metadata=serialized_metadata,
+        field_id=f"{field_obj.type}/{field_obj.id}",
+    )

{nucliadb-6.2.1.post3312.dist-info → nucliadb-6.2.1.post3326.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: nucliadb
-Version: 6.2.1.post3312
+Version: 6.2.1.post3326
 Summary: NucliaDB
 Author-email: Nuclia <nucliadb@nuclia.com>
 License: AGPL
@@ -20,11 +20,11 @@ Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3 :: Only
 Requires-Python: <4,>=3.9
 Description-Content-Type: text/markdown
-Requires-Dist: nucliadb-telemetry[all]>=6.2.1.post3312
-Requires-Dist: nucliadb-utils[cache,fastapi,storages]>=6.2.1.post3312
-Requires-Dist: nucliadb-protos>=6.2.1.post3312
-Requires-Dist: nucliadb-models>=6.2.1.post3312
-Requires-Dist: nidx-protos>=6.2.1.post3312
+Requires-Dist: nucliadb-telemetry[all]>=6.2.1.post3326
+Requires-Dist: nucliadb-utils[cache,fastapi,storages]>=6.2.1.post3326
+Requires-Dist: nucliadb-protos>=6.2.1.post3326
+Requires-Dist: nucliadb-models>=6.2.1.post3326
+Requires-Dist: nidx-protos>=6.2.1.post3326
 Requires-Dist: nucliadb-admin-assets>=1.0.0.post1224
 Requires-Dist: nuclia-models>=0.24.2
 Requires-Dist: uvicorn

{nucliadb-6.2.1.post3312.dist-info → nucliadb-6.2.1.post3326.dist-info}/RECORD

@@ -90,7 +90,7 @@ nucliadb/common/maindb/local.py,sha256=uE9DIQX1yCNHNN8Tx4fPgSiuTtWpQhlfWkMJ8QZPa
 nucliadb/common/maindb/pg.py,sha256=FNq2clckJYj4Te-1svjQblqGoAF5OwJ5nwz2JtxD0d4,13645
 nucliadb/common/maindb/utils.py,sha256=zWLs82rWEVhpc1dYvdqTZiAcjZroB6Oo5MQaxMeFuKk,3301
 nucliadb/common/models_utils/__init__.py,sha256=cp15ZcFnHvpcu_5-aK2A4uUyvuZVV_MJn4bIXMa20ks,835
-nucliadb/common/models_utils/from_proto.py,sha256=CAT01eDn2CnN2ovOaEUOPjNG6G2zn1MyCNO8Xlkd7Ys,15733
+nucliadb/common/models_utils/from_proto.py,sha256=d-q6-JWvjVVxtI57itQVKRIY8Hzv-tBhDt-tr4aIKNM,15979
 nucliadb/common/models_utils/to_proto.py,sha256=97JvOR_3odu50YvzLa2CERfEN3w_QPmAVcCJwJB5m5A,2438
 nucliadb/export_import/__init__.py,sha256=y-Is0Bxa8TMV6UiOW0deC_D3U465P65CQ5RjBjIWnow,932
 nucliadb/export_import/datamanager.py,sha256=b9Vhf-WqJ8HosTdNpKXlGj-Vi7MHyMoPxL0VpfPlYlE,6720
@@ -181,13 +181,14 @@ nucliadb/search/__init__.py,sha256=tnypbqcH4nBHbGpkINudhKgdLKpwXQCvDtPchUlsyY4,1
 nucliadb/search/app.py,sha256=6UV7rO0f3w5bNFXLdQM8bwUwXayMGnM4hF6GGv7WPv4,4260
 nucliadb/search/lifecycle.py,sha256=DW8v4WUi4rZqc7xTOi3rE67W7877WG7fH9oTZbolHdE,2099
 nucliadb/search/openapi.py,sha256=t3Wo_4baTrfPftg2BHsyLWNZ1MYn7ZRdW7ht-wFOgRs,1016
-nucliadb/search/predict.py,sha256=j51l_Zb8K9j6YAXIx1aafspqvdeZuANpyv96_pVfUEg,20641
+nucliadb/search/predict.py,sha256=z2-RkhMkH-5T6PtFkfESxNof07XiS5FxicLHPRyCUXc,22284
+nucliadb/search/predict_models.py,sha256=m3abP2J-xZIhKa230OvnnBRXxL4oSlozl5PpA7VeVvE,5908
 nucliadb/search/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 nucliadb/search/run.py,sha256=aFb-CXRi_C8YMpP_ivNj8KW1BYhADj88y8K9Lr_nUPI,1402
 nucliadb/search/settings.py,sha256=vem3EcyYlTPSim0kEK-xe-erF4BZg0CT_LAb8ZRQAE8,1684
 nucliadb/search/utilities.py,sha256=9SsRDw0rJVXVoLBfF7rBb6q080h-thZc7u8uRcTiBeY,1037
 nucliadb/search/api/__init__.py,sha256=cp15ZcFnHvpcu_5-aK2A4uUyvuZVV_MJn4bIXMa20ks,835
-nucliadb/search/api/v1/__init__.py,sha256=NSbOVF6toiHX9WMpGgqpFrjJKT50EcHzOILp-2iHk5I,1249
+nucliadb/search/api/v1/__init__.py,sha256=8w6VhZ5rbzX1xLSXr336d2IE-O0dQiv-ba6UYdRKnHA,1325
 nucliadb/search/api/v1/ask.py,sha256=F2dR3-swb3Xz8MfZPYL3G65KY2R_mgef4YVBbu8kLi4,4352
 nucliadb/search/api/v1/catalog.py,sha256=TF19WN-qgZZLkqBwVH5xNsMxYTrmdEflPvy7qft_4lE,7010
 nucliadb/search/api/v1/feedback.py,sha256=kNLc4dHz2SXHzV0PwC1WiRAwY88fDptPcP-kO0q-FrQ,2620
@@ -200,8 +201,10 @@ nucliadb/search/api/v1/suggest.py,sha256=S0YUTAWukzZSYZJzN3T5MUgPM3599HQvG76GOCB
 nucliadb/search/api/v1/summarize.py,sha256=VAHJvE6V3xUgEBfqNKhgoxmDqCvh30RnrEIBVhMcNLU,2499
 nucliadb/search/api/v1/utils.py,sha256=5Ve-frn7LAE2jqAgB85F8RSeqxDlyA08--gS-AdOLS4,1434
 nucliadb/search/api/v1/resource/__init__.py,sha256=cp15ZcFnHvpcu_5-aK2A4uUyvuZVV_MJn4bIXMa20ks,835
-nucliadb/search/api/v1/resource/ask.py,sha256=XMEP9_Uwy37yaXLcIYKMXGiZYNASD8RTByzQGjd9LPQ,3847
+nucliadb/search/api/v1/resource/ask.py,sha256=nsVzBSanSSlf0Ody6LSTjdEy75Vg283_YhbkAtWEjh8,3637
+nucliadb/search/api/v1/resource/ingestion_agents.py,sha256=fqqRCd8Wc9GciS5P98lcnihvTKStsZYYtOU-T1bc-6E,4771
 nucliadb/search/api/v1/resource/search.py,sha256=oSU5lwG7XRnD7oBFct31JaECGjTjX5R8mxNF1mskINc,4715
+nucliadb/search/api/v1/resource/utils.py,sha256=-NjZqAQtFEXKpIh8ui5S26ItnJ5rzmmG0BHxGSS9QPw,1141
 nucliadb/search/requesters/__init__.py,sha256=itSI7dtTwFP55YMX4iK7JzdMHS5CQVUiB1XzQu4UBh8,833
 nucliadb/search/requesters/utils.py,sha256=qL81UVPNgBftUMLpcxIYVr7ILsMqpKCo-9SY2EvAaXw,6681
 nucliadb/search/search/__init__.py,sha256=cp15ZcFnHvpcu_5-aK2A4uUyvuZVV_MJn4bIXMa20ks,835
@@ -214,6 +217,7 @@ nucliadb/search/search/find.py,sha256=AocqiH_mWvF_szUaW0ONqWrZAbX-k_VhM0Lpv7D669
 nucliadb/search/search/find_merge.py,sha256=3FnzKFEnVemg6FO_6zveulbAU7klvsiPEBvLrpBBMg8,17450
 nucliadb/search/search/graph_strategy.py,sha256=ahwcUTQZ0Ll-rnS285DO9PmRyiM-1p4BM3UvmOYVwhM,31750
 nucliadb/search/search/hydrator.py,sha256=-R37gCrGxkyaiHQalnTWHNG_FCx11Zucd7qA1vQCxuw,6985
+nucliadb/search/search/ingestion_agents.py,sha256=NeJr4EEX-bvFFMGvXOOwLv8uU7NuQ-ntJnnrhnKfMzY,3174
 nucliadb/search/search/merge.py,sha256=i_PTBFRqC5iTTziOMEltxLIlmokIou5hjjgR4BnoLBE,22635
 nucliadb/search/search/metrics.py,sha256=81X-tahGW4n2CLvUzCPdNxNClmZqUWZjcVOGCUHoiUM,2872
 nucliadb/search/search/paragraphs.py,sha256=pNAEiYqJGGUVcEf7xf-PFMVqz0PX4Qb-WNG-_zPGN2o,7799
@@ -331,8 +335,8 @@ nucliadb/writer/tus/local.py,sha256=7jYa_w9b-N90jWgN2sQKkNcomqn6JMVBOVeDOVYJHto,
 nucliadb/writer/tus/s3.py,sha256=vF0NkFTXiXhXq3bCVXXVV-ED38ECVoUeeYViP8uMqcU,8357
 nucliadb/writer/tus/storage.py,sha256=ToqwjoYnjI4oIcwzkhha_MPxi-k4Jk3Lt55zRwaC1SM,2903
 nucliadb/writer/tus/utils.py,sha256=MSdVbRsRSZVdkaum69_0wku7X3p5wlZf4nr6E0GMKbw,2556
-nucliadb-6.2.1.post3312.dist-info/METADATA,sha256=UfQFTvgOW3bRrzu96TjD3-T826GHe2rCnNFmT3O2plU,4291
-nucliadb-6.2.1.post3312.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-nucliadb-6.2.1.post3312.dist-info/entry_points.txt,sha256=XqGfgFDuY3zXQc8ewXM2TRVjTModIq851zOsgrmaXx4,1268
-nucliadb-6.2.1.post3312.dist-info/top_level.txt,sha256=hwYhTVnX7jkQ9gJCkVrbqEG1M4lT2F_iPQND1fCzF80,20
-nucliadb-6.2.1.post3312.dist-info/RECORD,,
+nucliadb-6.2.1.post3326.dist-info/METADATA,sha256=dlbBvdHdR_9AzOSB7w415CnvT_1xpxCNMDsxwPHoyek,4291
+nucliadb-6.2.1.post3326.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
+nucliadb-6.2.1.post3326.dist-info/entry_points.txt,sha256=XqGfgFDuY3zXQc8ewXM2TRVjTModIq851zOsgrmaXx4,1268
+nucliadb-6.2.1.post3326.dist-info/top_level.txt,sha256=hwYhTVnX7jkQ9gJCkVrbqEG1M4lT2F_iPQND1fCzF80,20
+nucliadb-6.2.1.post3326.dist-info/RECORD,,