[FEATURE] Add support to update record fields

.github/workflows/argilla.yml

@@ -24,7 +24,7 @@ jobs:
   build:
     services:
       argilla-server:
-        image: argilladev/argilla-hf-spaces:develop
+        image: argilladev/argilla-hf-spaces:pr-5685
         ports:
           - 6900:6900
         env:

argilla-server/CHANGELOG.md

@@ -16,6 +16,11 @@ These are the section headers that we use:
 
 ## [Unreleased]()
 
+### Added
+
+- Added support to update record fields in `PATCH /api/v1/records/:record_id` endpoint. ([#5685](https://github.com/argilla-io/argilla/pull/5685))
+- Added support to update record fields in `PUT /api/v1/datasets/:dataset_id/records/bulk` endpoint. ([#5685](https://github.com/argilla-io/argilla/pull/5685))
+
 ### Changed
 
 - Changed default python version to 3.13. ([#5649](https://github.com/argilla-io/argilla/pull/5649))

argilla-server/src/argilla_server/api/handlers/v1/datasets/records.py

@@ -264,7 +264,7 @@ async def delete_dataset_records(
     if num_records > DELETE_DATASET_RECORDS_LIMIT:
         raise UnprocessableEntityError(f"Cannot delete more than {DELETE_DATASET_RECORDS_LIMIT} records at once")
 
-    await datasets.delete_records(db, search_engine, dataset, record_ids)
+    await records.delete_records(db, search_engine, dataset, record_ids)
 
 
 @router.post(

argilla-server/src/argilla_server/api/handlers/v1/records.py

@@ -25,7 +25,7 @@
 from argilla_server.api.schemas.v1.responses import Response, ResponseCreate
 from argilla_server.api.schemas.v1.suggestions import Suggestion as SuggestionSchema
 from argilla_server.api.schemas.v1.suggestions import SuggestionCreate, Suggestions
-from argilla_server.contexts import datasets
+from argilla_server.contexts import datasets, records
 from argilla_server.database import get_async_db
 from argilla_server.errors.future.base_errors import NotFoundError, UnprocessableEntityError
 from argilla_server.models import Dataset, Question, Record, Suggestion, User
@@ -74,16 +74,21 @@ async def update_record(
         db,
         record_id,
         options=[
-            selectinload(Record.dataset).selectinload(Dataset.questions),
-            selectinload(Record.dataset).selectinload(Dataset.metadata_properties),
+            selectinload(Record.dataset).options(
+                selectinload(Dataset.questions),
+                selectinload(Dataset.metadata_properties),
+                selectinload(Dataset.vectors_settings),
+                selectinload(Dataset.fields),
+            ),
             selectinload(Record.suggestions),
+            selectinload(Record.responses),
             selectinload(Record.vectors),
         ],
     )
 
     await authorize(current_user, RecordPolicy.update(record))
 
-    return await datasets.update_record(db, search_engine, record, record_update)
+    return await records.update_record(db, search_engine, record, record_update)
 
 
 @router.post("/records/{record_id}/responses", status_code=status.HTTP_201_CREATED, response_model=Response)
@@ -233,4 +238,4 @@ async def delete_record(
 
     await authorize(current_user, RecordPolicy.delete(record))
 
-    return await datasets.delete_record(db, search_engine, record)
+    return await records.delete_record(db, search_engine, record)

argilla-server/src/argilla_server/api/schemas/v1/records.py

@@ -16,7 +16,6 @@
 from typing import Annotated, Any, Dict, List, Literal, Optional, Union
 from uuid import UUID
 
-
 from argilla_server.api.schemas.v1.chat import ChatFieldValue
 from argilla_server.api.schemas.v1.commons import UpdateSchema
 from argilla_server.api.schemas.v1.metadata_properties import MetadataPropertyName
@@ -151,17 +150,12 @@ def prevent_nan_values(cls, metadata: Optional[Dict[str, Any]]) -> Optional[Dict
 
 
 class RecordUpdate(UpdateSchema):
-    metadata_: Optional[Dict[str, Any]] = Field(None, alias="metadata")
+    fields: Optional[Dict[str, FieldValueCreate]] = None
+    metadata: Optional[Dict[str, Any]] = None
     suggestions: Optional[List[SuggestionCreate]] = None
     vectors: Optional[Dict[str, List[float]]]
 
-    @property
-    def metadata(self) -> Optional[Dict[str, Any]]:
-        # Align with the RecordCreate model. Both should have the same name for the metadata field.
-        # TODO(@frascuchon): This will be properly adapted once the bulk records refactor is completed.
-        return self.metadata_
-
-    @validator("metadata_")
+    @validator("metadata")
     @classmethod
     def prevent_nan_values(cls, metadata: Optional[Dict[str, Any]]) -> Optional[Dict[str, Any]]:
         if metadata is None:
@@ -173,15 +167,20 @@ def prevent_nan_values(cls, metadata: Optional[Dict[str, Any]]) -> Optional[Dict
 
         return {k: v for k, v in metadata.items() if v == v}  # By definition, NaN != NaN
 
+    def is_set(self, attribute: str) -> bool:
+        return attribute in self.__fields_set__
 
-class RecordUpdateWithId(RecordUpdate):
-    id: UUID
+    def has_changes(self) -> bool:
+        return self.dict(exclude_unset=True) != {}
 
 
 class RecordUpsert(RecordCreate):
     id: Optional[UUID]
     fields: Optional[Dict[str, FieldValueCreate]] = None
 
+    def is_set(self, attribute: str) -> bool:
+        return attribute in self.__fields_set__
+
 
 class RecordIncludeParam(BaseModel):
     relationships: Optional[List[RecordInclude]] = Field(None, alias="keys")
@@ -245,11 +244,6 @@ class RecordsCreate(BaseModel):
     items: List[RecordCreate] = Field(..., min_items=RECORDS_CREATE_MIN_ITEMS, max_items=RECORDS_CREATE_MAX_ITEMS)
 
 
-class RecordsUpdate(BaseModel):
-    # TODO: review this definition and align to create model
-    items: List[RecordUpdateWithId] = Field(..., min_items=RECORDS_UPDATE_MIN_ITEMS, max_items=RECORDS_UPDATE_MAX_ITEMS)
-
-
 class MetadataParsedQueryParam:
     def __init__(self, string: str):
         k, *v = string.split(":", maxsplit=1)

argilla-server/src/argilla_server/api/schemas/v1/records_bulk.py

@@ -29,7 +29,7 @@ class RecordsBulk(BaseModel):
     items: List[Record]
 
 
-class RecordsBulkWithUpdateInfo(RecordsBulk):
+class RecordsBulkWithUpdatedItemIds(RecordsBulk):
     updated_item_ids: List[UUID]
 
 

argilla-server/src/argilla_server/bulk/records_bulk.py

@@ -16,6 +16,7 @@
 from typing import Dict, List, Sequence, Tuple, Union
 from uuid import UUID
 
+from datetime import UTC
 from fastapi.encoders import jsonable_encoder
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
@@ -26,7 +27,7 @@
     RecordsBulk,
     RecordsBulkCreate,
     RecordsBulkUpsert,
-    RecordsBulkWithUpdateInfo,
+    RecordsBulkWithUpdatedItemIds,
 )
 from argilla_server.api.schemas.v1.responses import UserResponseCreate
 from argilla_server.api.schemas.v1.suggestions import SuggestionCreate
@@ -36,7 +37,7 @@
     fetch_records_by_ids_as_dict,
 )
 from argilla_server.errors.future import UnprocessableEntityError
-from argilla_server.models import Dataset, Record, Response, Suggestion, Vector, VectorSettings
+from argilla_server.models import Dataset, Record, Response, Suggestion, Vector
 from argilla_server.search_engine import SearchEngine
 from argilla_server.validators.records import RecordsBulkCreateValidator, RecordUpsertValidator
 
@@ -139,15 +140,11 @@ async def _upsert_records_vectors(
             autocommit=False,
         )
 
-    @classmethod
-    def _metadata_is_set(cls, record_create: RecordCreate) -> bool:
-        return "metadata" in record_create.__fields_set__
-
 
 class UpsertRecordsBulk(CreateRecordsBulk):
     async def upsert_records_bulk(
         self, dataset: Dataset, bulk_upsert: RecordsBulkUpsert, raise_on_error: bool = True
-    ) -> RecordsBulkWithUpdateInfo:
+    ) -> RecordsBulkWithUpdatedItemIds:
         found_records = await self._fetch_existing_dataset_records(dataset, bulk_upsert.items)
 
         records = []
@@ -170,9 +167,14 @@ async def upsert_records_bulk(
                     external_id=record_upsert.external_id,
                     dataset_id=dataset.id,
                 )
-            elif self._metadata_is_set(record_upsert):
-                record.metadata_ = record_upsert.metadata
-                record.updated_at = datetime.utcnow()
+            else:
+                if record_upsert.is_set("metadata"):
+                    record.metadata_ = record_upsert.metadata
+                if record_upsert.is_set("fields"):
+                    record.fields = jsonable_encoder(record_upsert.fields)
+
+                if self._db.is_modified(record):
+                    record.updated_at = datetime.now(UTC)
 
             records.append(record)
 
@@ -186,7 +188,7 @@ async def upsert_records_bulk(
         await _preload_records_relationships_before_index(self._db, records)
         await self._search_engine.index_records(dataset, records)
 
-        return RecordsBulkWithUpdateInfo(
+        return RecordsBulkWithUpdatedItemIds(
             items=records,
             updated_item_ids=[record.id for record in found_records.values()],
         )