feat: configurable examples

gliner_api/__init__.py

@@ -0,0 +1,24 @@
+from pydantic import AliasChoices, BaseModel, Field
+
+
+class Entity(BaseModel):
+    start: int = Field(
+        ge=0,
+        description="Start index of the entity in the input text",
+    )
+    end: int = Field(
+        ge=0,
+        description="End index of the entity in the input text",
+    )
+    text: str = Field(
+        description="Text of the entity, extracted from the input text",
+    )
+    type: str = Field(
+        validation_alias=AliasChoices("type", "label"),
+        description="Entity type or label",
+    )
+    score: float = Field(
+        ge=0.0,
+        le=1.0,
+        description="Confidence score of the entity detection, between 0 and 1",
+    )

gliner_api/backend.py

@@ -10,11 +10,11 @@
 from gliner import GLiNER
 from starlette.status import HTTP_401_UNAUTHORIZED, HTTP_403_FORBIDDEN
 
+from gliner_api import Entity
 from gliner_api.config import Config, get_config
 from gliner_api.datamodel import (
     BatchRequest,
     BatchResponse,
-    Entity,
     ErrorMessage,
     HealthCheckResponse,
     InfoResponse,

gliner_api/datamodel.py

@@ -1,136 +1,88 @@
-from pydantic import AliasChoices, BaseModel, Field, TypeAdapter
+from pydantic import BaseModel, Field, TypeAdapter
 
+from gliner_api import Entity
 from gliner_api.config import Config, get_config
+from gliner_api.examples import Examples, get_examples
 
 config: Config = get_config()
+examples: Examples = get_examples()
 
 
 class ErrorMessage(BaseModel):
     error: str = Field(description="Short error code")
     detail: str = Field(description="Detailed error explanaiton")
 
 
-class Entity(BaseModel):
-    start: int = Field(
-        ge=0,
-        description="Start index of the entity in the input text",
-    )
-    end: int = Field(
-        ge=0,
-        description="End index of the entity in the input text",
-    )
-    text: str = Field(
-        description="Text of the entity, extracted from the input text",
-    )
-    type: str = Field(
-        validation_alias=AliasChoices("type", "label"),
-        description="Entity type or label",
-    )
-    score: float = Field(
-        ge=0.0,
-        le=1.0,
-        description="Confidence score of the entity detection, between 0 and 1",
-    )
-
-
 class InvokeRequest(BaseModel):
     text: str = Field(
         description="Input text to analyze for entities",
-        examples=["Steve Jobs founded Apple Inc. in Cupertino, CA on April 1, 1976."],
+        examples=[example.text for example in examples.invoke],
     )
     threshold: float = Field(
         default=config.default_threshold,
         description="Threshold for entity detection; if not set, uses default threshold (see gliner config from /api/info endpoint)",
-        examples=[0.5],
+        examples=[example.threshold for example in examples.invoke],
         ge=0.0,
         le=1.0,
     )
     entity_types: list[str] = Field(
         default=config.default_entities,
         description="List of entity types to detect; if not set, uses default entities (see gliner config from /api/info endpoint)",
-        examples=[["person", "organization", "location", "date"]],
+        examples=[example.entity_types for example in examples.invoke],
     )
     flat_ner: bool = Field(
         default=True,
         description="Whether to return flat entities (default: True). If False, returns nested entities.",
-        examples=[True],
+        examples=[example.flat_ner for example in examples.invoke],
     )
     multi_label: bool = Field(
         default=False,
         description="Whether to allow multiple labels per entity (default: False). If True, there can be multiple entities returned for the same span.",
-        examples=[False],
+        examples=[example.multi_label for example in examples.invoke],
     )
 
 
 class InvokeResponse(BaseModel):
     entities: list[Entity] = Field(
         description="List of detected entities in the input text",
-        examples=[
-            [
-                Entity(start=0, end=10, text="Steve Jobs", type="person", score=0.99),
-                Entity(start=19, end=24, text="Apple", type="organization", score=0.98),
-                Entity(start=28, end=37, text="Cupertino", type="location", score=0.98),
-                Entity(start=39, end=49, text="California", type="location", score=0.99),
-                Entity(start=53, end=66, text="April 1, 1976", type="date", score=0.68),
-            ]
-        ],
+        examples=[example.entities for example in examples.invoke],
     )
 
 
 class BatchRequest(BaseModel):
     texts: list[str] = Field(
         description="List of input texts to analyze for entities",
-        examples=[
-            [
-                "Steve Jobs founded Apple Inc. in Cupertino, CA on April 1, 1976.",
-                "Until her death in 2022, the head of the Windsor family, Queen Elizabeth, resided in London.",
-            ],
-        ],
+        examples=[example.texts for example in examples.batch],
+        min_length=1,
     )
     threshold: float = Field(
         default=config.default_threshold,
         description="Threshold for entity detection; if not set, uses default threshold (see gliner config from /api/info endpoint)",
-        examples=[0.3],
+        examples=[example.threshold for example in examples.batch],
         ge=0.0,
         le=1.0,
     )
     entity_types: list[str] = Field(
         default=config.default_entities,
         description="List of entity types to detect; if not set, uses default entities (see gliner config from /api/info endpoint)",
-        examples=[["person", "organization", "location", "date"]],
+        examples=[example.entity_types for example in examples.batch],
     )
     flat_ner: bool = Field(
         default=True,
         description="Whether to return flat entities (default: True). If False, returns nested entities.",
-        examples=[True],
+        examples=[example.flat_ner for example in examples.batch],
     )
     multi_label: bool = Field(
         default=False,
         description="Whether to allow multiple labels per entity (default: False). If True, there can be multiple entities returned for the same span.",
-        examples=[False],
+        examples=[example.multi_label for example in examples.batch],
     )
 
 
 class BatchResponse(BaseModel):
     entities: list[list[Entity]] = Field(
         description="List of lists of detected entities for each input text",
-        examples=[
-            [
-                [
-                    Entity(start=0, end=10, text="Steve Jobs", type="person", score=0.99),
-                    Entity(start=19, end=24, text="Apple", type="organization", score=0.98),
-                    Entity(start=28, end=37, text="Cupertino", type="location", score=0.98),
-                    Entity(start=39, end=49, text="California", type="location", score=0.99),
-                    Entity(start=53, end=66, text="April 1, 1976", type="date", score=0.68),
-                ],
-                [
-                    Entity(start=19, end=23, text="2022", type="date", score=0.38),
-                    Entity(start=41, end=55, text="Windsor family", type="organization", score=0.90),
-                    Entity(start=57, end=72, text="Queen Elizabeth", type="person", score=0.99),
-                    Entity(start=85, end=91, text="London", type="location", score=0.99),
-                ],
-            ]
-        ],
+        examples=[example.entities for example in examples.batch],
     )
 
 
@@ -145,37 +97,31 @@ class InfoResponse(BaseModel):
     model_id: str = Field(
         default=config.model_id,
         description="The Huggingface model ID for a GLiNER model.",
-        examples=["knowledgator/gliner-x-base"],
     )
     default_entities: list[str] = Field(
         default=config.default_entities,
         description="The default entities to be detected, used if request includes no specific entities.",
-        examples=[["person", "organization", "location", "date"]],
     )
     default_threshold: float = Field(
         default=config.default_threshold,
         description="The default threshold for entity detection, used if request includes no specific threshold.",
-        examples=[0.5],
         ge=0.0,
         le=1.0,
     )
     api_key_required: bool = Field(
         default=config.api_key is not None,
         description="Whether an API key is required for requests",
-        examples=[False],
     )
     configured_use_case: str = Field(
         default=config.use_case,
         description="The configured use case for this deployment",
-        examples=["general"],
     )
     onnx_enabled: bool = Field(
         default=config.onnx_enabled,
         description="Whether the GLiNER model is loaded as an ONNX model",
-        examples=[False],
     )
 
 
 # Define TypeAdapter for Entity list once and reuse it
-entity_list_adapter: TypeAdapter[list[Entity]] = TypeAdapter(list[Entity])
-deep_entity_list_adapter: TypeAdapter[list[list[Entity]]] = TypeAdapter(list[list[Entity]])
+entity_list_adapter: TypeAdapter[list[Entity]] = TypeAdapter(type=list[Entity])
+deep_entity_list_adapter: TypeAdapter[list[list[Entity]]] = TypeAdapter(type=list[list[Entity]])

gliner_api/examples.py

@@ -0,0 +1,127 @@
+from functools import lru_cache
+
+from pydantic import Field
+from pydantic_settings import (
+    BaseSettings,
+    JsonConfigSettingsSource,
+    PydanticBaseSettingsSource,
+    SettingsConfigDict,
+    YamlConfigSettingsSource,
+)
+
+from gliner_api import Entity
+from gliner_api.config import Config, get_config
+
+config: Config = get_config()
+
+
+class Examples(BaseSettings):
+    invoke: list["InvokeExample"] = Field(
+        default=[
+            {
+                "text": "Steve Jobs founded Apple Inc. in Cupertino, CA on April 1, 1976.",
+                "entities": [
+                    Entity(start=0, end=10, text="Steve Jobs", type="person", score=0.99),
+                    Entity(start=19, end=24, text="Apple", type="organization", score=0.98),
+                    Entity(start=28, end=37, text="Cupertino", type="location", score=0.98),
+                    Entity(start=39, end=49, text="California", type="location", score=0.99),
+                    Entity(start=53, end=66, text="April 1, 1976", type="date", score=0.68),
+                ],
+            },
+            {
+                "text": "Until her death in 2022, the head of the Windsor family, Queen Elizabeth, resided in London.",
+                "entity_types": ["person", "organization", "location", "date"],
+                "entities": [],
+            },
+            {
+                "text": "The Eiffel Tower was completed in 1889 and is located in Paris, France.",
+                "entity_types": ["building", "location", "date"],
+                "entities": [],
+            },
+            {
+                "text": "Barack Obama served as the 44th President of the United States from 2009 to 2017.",
+                "threshold": 0.4,
+                "entity_types": ["person", "organization", "location", "date", "job title"],
+                "flat_ner": False,
+                "entities": [],
+            },
+            {
+                "text": "Albert Einstein developed the theory of relativity, which revolutionized modern physics.",
+                "threshold": 0.2,
+                "entity_types": ["person", "research field", "topic", "physical law"],
+                "multi_label": True,
+                "entities": [],
+            },
+        ]
+    )
+    batch: list["BatchExample"] = Field(
+        default=[
+            {
+                "texts": [
+                    "Steve Jobs founded Apple Inc. in Cupertino, CA on April 1, 1976.",
+                    "Until her death in 2022, the head of the Windsor family, Queen Elizabeth, resided in London.",
+                ],
+                "entities": [
+                    [
+                        Entity(start=0, end=10, text="Steve Jobs", type="person", score=0.99),
+                        Entity(start=19, end=24, text="Apple", type="organization", score=0.98),
+                        Entity(start=28, end=37, text="Cupertino", type="location", score=0.98),
+                        Entity(start=39, end=49, text="California", type="location", score=0.99),
+                        Entity(start=53, end=66, text="April 1, 1976", type="date", score=0.68),
+                    ],
+                    [
+                        Entity(start=19, end=23, text="2022", type="date", score=0.38),
+                        Entity(start=41, end=55, text="Windsor family", type="organization", score=0.90),
+                        Entity(start=57, end=72, text="Queen Elizabeth", type="person", score=0.99),
+                        Entity(start=85, end=91, text="London", type="location", score=0.99),
+                    ],
+                ],
+            }
+        ]
+    )
+
+    model_config: SettingsConfigDict = SettingsConfigDict(
+        yaml_file="examples.yaml",
+        yaml_file_encoding="utf-8",
+        json_file="examples.json",
+        json_file_encoding="utf-8",
+    )
+
+    @classmethod
+    def settings_customise_sources(
+        cls,
+        settings_cls: type[BaseSettings],
+        init_settings: PydanticBaseSettingsSource,
+        env_settings: PydanticBaseSettingsSource,
+        dotenv_settings: PydanticBaseSettingsSource,
+        file_secret_settings: PydanticBaseSettingsSource,
+    ) -> tuple[PydanticBaseSettingsSource, ...]:
+        return (
+            init_settings,
+            YamlConfigSettingsSource(settings_cls=settings_cls),
+            JsonConfigSettingsSource(settings_cls=settings_cls),
+        )
+
+
+class InvokeExample(BaseSettings):
+    text: str
+    threshold: float = Field(ge=0.0, le=1.0, default=config.default_threshold)
+    entity_types: list[str] = Field(default=config.default_entities)
+    flat_ner: bool = True
+    multi_label: bool = False
+    entities: list[Entity]
+
+
+class BatchExample(BaseSettings):
+    texts: list[str]
+    threshold: float = Field(ge=0.0, le=1.0, default=config.default_threshold)
+    entity_types: list[str] = Field(default=config.default_entities)
+    flat_ner: bool = True
+    multi_label: bool = False
+    entities: list[list[Entity]]
+
+
+@lru_cache
+def get_examples() -> Examples:
+    """Get the examples for the API docs and the frontend."""
+    return Examples()