feat: llm export format (#1716)

Co-authored-by: Nicolas Hervé <nicolas.herve@kili-technology.com>

Author: HNicolas

src/kili/presentation/client/label.py

@@ -1126,7 +1126,7 @@ def export_labels_as_df(
     def export_labels(
         self,
         project_id: str,
-        filename: str,
+        filename: Optional[str],
         fmt: LabelFormat,
         asset_ids: Optional[List[str]] = None,
         layout: SplitOption = "split",
@@ -1137,7 +1137,7 @@ def export_labels(
         annotation_modifier: Optional[CocoAnnotationModifier] = None,
         asset_filter_kwargs: Optional[Dict[str, object]] = None,
         normalized_coordinates: Optional[bool] = None,
-    ) -> None:
+    ) -> Optional[List[Dict[str, Union[List[str], str]]]]:
         # pylint: disable=line-too-long
         """Export the project labels with the requested format into the requested output path.
 
@@ -1225,7 +1225,7 @@ def is_rectangle(coco_annotation, coco_image, kili_annotation):
             resolved_asset_ids = cast(List[AssetId], asset_ids)
 
         try:
-            export_labels(
+            return export_labels(
                 self,  # pyright: ignore[reportGeneralTypeIssues]
                 asset_ids=resolved_asset_ids,
                 project_id=ProjectId(project_id),
@@ -1243,3 +1243,4 @@ def is_rectangle(coco_annotation, coco_image, kili_annotation):
             )
         except NoCompatibleJobError as excp:
             warnings.warn(str(excp), stacklevel=2)
+            return None

src/kili/services/export/__init__.py

@@ -1,7 +1,7 @@
 """Service for exporting kili objects."""
 
 from pathlib import Path
-from typing import TYPE_CHECKING, Dict, List, Optional, Type
+from typing import TYPE_CHECKING, Dict, List, Optional, Type, Union
 
 from typing_extensions import get_args
 
@@ -31,14 +31,14 @@ def export_labels(  # pylint: disable=too-many-arguments, too-many-locals
     label_format: LabelFormat,
     split_option: SplitOption,
     single_file: bool,
-    output_file: str,
+    output_file: Optional[str],
     disable_tqdm: Optional[bool],
     log_level: LogLevel,
     with_assets: bool,
     annotation_modifier: Optional[CocoAnnotationModifier],
     asset_filter_kwargs: Optional[Dict[str, object]],
     normalized_coordinates: Optional[bool],
-) -> None:
+) -> Optional[List[Dict[str, Union[List[str], str]]]]:
     """Export the selected assets into the required format, and save it into a file archive."""
     kili.kili_api_gateway.get_project(project_id, ["id"])
 
@@ -49,7 +49,7 @@ def export_labels(  # pylint: disable=too-many-arguments, too-many-locals
         label_format=label_format,
         split_option=split_option,
         single_file=single_file,
-        output_file=Path(output_file),
+        output_file=Path(output_file) if output_file is not None else None,
         with_assets=with_assets,
         annotation_modifier=annotation_modifier,
         asset_filter_kwargs=asset_filter_kwargs,
@@ -77,8 +77,8 @@ def export_labels(  # pylint: disable=too-many-arguments, too-many-locals
             get_args(LabelFormat)
         )  # ensures full mapping
         exporter_class = format_exporter_selector_mapping[label_format]
-        exporter_class(
+        return exporter_class(
             export_params, kili, logger, disable_tqdm, content_repository
         ).export_project()
-    else:
-        raise ValueError(f'Label format "{label_format}" is not implemented or does not exist.')
+
+    raise ValueError(f'Label format "{label_format}" is not implemented or does not exist.')

src/kili/services/export/format/base.py

@@ -7,7 +7,7 @@
 from abc import ABC, abstractmethod
 from datetime import datetime
 from pathlib import Path
-from typing import TYPE_CHECKING, Dict, List, NamedTuple, Optional, Tuple
+from typing import TYPE_CHECKING, Dict, List, NamedTuple, Optional, Tuple, Union
 
 from kili.domain.asset import AssetId
 from kili.domain.project import ProjectId
@@ -41,7 +41,7 @@ class ExportParams(NamedTuple):
     label_format: LabelFormat
     split_option: SplitOption
     single_file: bool
-    output_file: Path
+    output_file: Optional[Path]
     with_assets: bool
     annotation_modifier: Optional[CocoAnnotationModifier]
     asset_filter_kwargs: Optional[Dict[str, object]]
@@ -107,6 +107,10 @@ def compatible_jobs(self) -> Tuple[str, ...]:
     def process_and_save(self, assets: List[Dict], output_filename: Path) -> None:
         """Converts the asset and save them into an archive file."""
 
+    def process(self, assets: List[Dict]) -> List[Dict[str, Union[List[str], str]]]:
+        """Converts the asset."""
+        raise ValueError("Output file is required for this export format.")
+
     def make_archive(self, root_folder: Path, output_filename: Path) -> Path:
         """Make the export archive."""
         path_folder = root_folder / self.project_id
@@ -148,7 +152,7 @@ def write_remote_content_file(remote_content: List[str], images_folder: Path) ->
 
     def export_project(
         self,
-    ) -> None:
+    ) -> Optional[List[Dict[str, Union[List[str], str]]]]:
         """Export a project to a json.
 
         Return the name of the exported archive file.
@@ -175,7 +179,10 @@ def export_project(
 
             self._check_geotiff_export_compatibility(assets)
 
-            self.process_and_save(assets, self.output_file)
+            if self.output_file is None:
+                return self.process(assets)
+
+            return self.process_and_save(assets, self.output_file)
 
     def _check_and_ensure_asset_access(self) -> None:
         """Check asset access.

src/kili/services/export/format/llm/__init__.py

@@ -3,7 +3,7 @@
 import json
 import logging
 from pathlib import Path
-from typing import Dict, List, Union
+from typing import Dict, List, Optional, Union
 
 from kili.services.export.exceptions import NotCompatibleInputType
 from kili.services.export.format.base import AbstractExporter
@@ -25,8 +25,6 @@ def _check_project_compatibility(self) -> None:
                 f"Project with input type \"{self.project['inputType']}\" not compatible with LLM"
                 " export format."
             )
-        if not self.single_file:
-            raise ValueError("LLM export are always single file.")
 
     def _is_job_compatible(self, job: Job) -> bool:
         """Check job compatibility with the LLM format."""
@@ -41,23 +39,41 @@ def _save_assets_export(self, assets: List[Dict], output_filename: Path) -> None
         with output_filename.open("wb") as output_file:
             output_file.write(export_json.encode("utf-8"))
 
-    def process_and_save(self, assets: List[Dict], output_filename: Path) -> None:
+    def process_and_save(
+        self, assets: List[Dict], output_filename: Path
+    ) -> Optional[List[Dict[str, Union[List[str], str]]]]:
         """LLM specific process and save."""
+        result = self._process(assets)
+        self._save_assets_export(result, output_filename)
+
+    def process(self, assets: List[Dict]) -> List[Dict[str, Union[List[str], str]]]:
+        """LLM specific process."""
+        return self._process(assets)
+
+    def _process(self, assets: List[Dict]) -> List[Dict[str, Union[List[str], str]]]:
         result = []
         for asset in assets:
             jobs_config = self.project["jsonInterface"]["jobs"]
-            json_response = asset["latestLabel"]["jsonResponse"]
+            latest_label = asset["latestLabel"]
             result.append(
                 {
                     "raw_data": _format_raw_data(asset),
                     "status": asset["status"],
                     "external_id": asset["externalId"],
                     "metadata": asset["jsonMetadata"],
-                    "labels": [_format_json_response(jobs_config, json_response)],
+                    "labels": [
+                        {
+                            "author": latest_label["author"]["email"],
+                            "created_at": latest_label["createdAt"],
+                            "label_type": latest_label["labelType"],
+                            "label": _format_json_response(
+                                jobs_config, latest_label["jsonResponse"]
+                            ),
+                        }
+                    ],
                 }
             )
-
-        self._save_assets_export(result, output_filename)
+        return result
 
 
 def _format_json_response(