feat: DIA-1415: LabelStudio skill

adala/skills/__init__.py

@@ -3,4 +3,5 @@
 from .collection.entity_extraction import EntityExtraction
 from .collection.rag import RAGSkill
 from .collection.ontology_creation import OntologyCreator, OntologyMerger
+from .collection.label_studio import LabelStudioSkill
 from ._base import Skill, TransformSkill, AnalysisSkill, SynthesisSkill

adala/skills/collection/label_studio.py

@@ -0,0 +1,71 @@
+import logging
+from typing import Dict, Any, Type
+from functools import cached_property
+from adala.skills._base import TransformSkill
+from pydantic import BaseModel, Field, model_validator
+
+from adala.runtimes import Runtime, AsyncRuntime
+from adala.utils.internal_data import InternalDataFrame
+
+from label_studio_sdk.label_interface import LabelInterface
+from label_studio_sdk._extensions.label_studio_tools.core.utils.json_schema import json_schema_to_pydantic
+
+
+logger = logging.getLogger(__name__)
+
+
+class LabelStudioSkill(TransformSkill):
+
+    name: str = "label_studio"
+    input_template: str = "Annotate the input data according to the provided schema."
+    # TODO: remove output_template, fix calling @model_validator(mode='after') in the base class
+    output_template: str = "Output: {field_name}"
+    response_model: Type[BaseModel] = BaseModel  # why validate_response_model is called in the base class?
+    # ------------------------------
+    label_config: str = "<View></View>"
+
+    # TODO: implement postprocessing like in EntityExtractionSkill or to verify Taxonomy
+
+    @model_validator(mode='after')
+    def validate_response_model(self):
+
+        interface = LabelInterface(self.label_config)
+        logger.debug(f'Read labeling config {self.label_config}')
+
+        self.field_schema = interface.to_json_schema()
+        logger.debug(f'Converted labeling config to json schema: {self.field_schema}')
+
+        return self
+
+    def _create_response_model_from_field_schema(self):
+        pass
+
+    def apply(
+        self,
+        input: InternalDataFrame,
+        runtime: Runtime,
+    ) -> InternalDataFrame:
+
+        with json_schema_to_pydantic(self.field_schema) as ResponseModel:
+            return runtime.batch_to_batch(
+                input,
+                input_template=self.input_template,
+                output_template="",
+                instructions_template=self.instructions,
+                response_model=ResponseModel,
+            )
+
+    async def aapply(
+        self,
+        input: InternalDataFrame,
+        runtime: AsyncRuntime,
+    ) -> InternalDataFrame:
+
+        with json_schema_to_pydantic(self.field_schema) as ResponseModel:
+            return await runtime.batch_to_batch(
+                input,
+                input_template=self.input_template,
+                output_template="",
+                instructions_template=self.instructions,
+                response_model=ResponseModel,
+            )

pyproject.toml

@@ -36,7 +36,7 @@ celery = {version = "^5.3.6", extras = ["redis"]}
 kombu = ">=5.4.0rc2" # Pin version to fix https://github.com/celery/celery/issues/8030. TODO: remove when this fix will be included in celery
 uvicorn = "*"
 pydantic-settings = "^2.2.1"
-label-studio-sdk = {url = "https://github.com/HumanSignal/label-studio-sdk/archive/81ab3bbed0373fca1fd9166f186bfd0817512331.zip"}
+label-studio-sdk = {url = "https://github.com/HumanSignal/label-studio-sdk/archive/e94072130d90f7b89701b9234ce175425f36f23e.zip"}
 kafka-python = "^2.0.2"
 # https://github.com/geerlingguy/ansible-role-docker/issues/462#issuecomment-2144121102
 requests = "2.31.0"