refactor(client): use auto generated arg_types

client/starwhale/api/_impl/service/types/llm.py

@@ -1,17 +1,16 @@
 from __future__ import annotations
 
 import inspect
-from typing import Any, Dict, List, Callable, Optional
+from typing import Any, List, Callable, Optional
 
 from pydantic import BaseModel
 
 from starwhale.base.client.models.models import (
     ComponentValueSpecInt,
-    ComponentSpecValueType,
     ComponentValueSpecFloat,
 )
 
-from .types import ServiceType
+from .types import ServiceType, generate_type_definition
 
 
 class MessageItem(BaseModel):
@@ -31,17 +30,7 @@ class Query(BaseModel):
 class LLMChat(ServiceType):
     name = "llm_chat"
     args = {}
-
-    # TODO use pydantic model annotations generated arg_types
-    arg_types: Dict[str, ComponentSpecValueType] = {
-        "user_input": ComponentSpecValueType.string,
-        "history": ComponentSpecValueType.list,  # list of Message
-        "top_k": ComponentSpecValueType.int,
-        "top_p": ComponentSpecValueType.float,
-        "temperature": ComponentSpecValueType.float,
-        "max_new_tokens": ComponentSpecValueType.int,
-    }
-
+    arg_types = generate_type_definition(Query)
     Message = MessageItem
 
     def __init__(

client/starwhale/api/_impl/service/types/text_to_img.py

@@ -1,16 +1,18 @@
 from __future__ import annotations
 
 import inspect
-from typing import Any, Dict, Callable, Optional
+from typing import Any, Callable, Optional
 
 from pydantic import BaseModel
 
 from starwhale.base.client.models.models import (
     ComponentValueSpecInt,
-    ComponentSpecValueType,
     ComponentValueSpecFloat,
 )
-from starwhale.api._impl.service.types.types import ServiceType
+from starwhale.api._impl.service.types.types import (
+    ServiceType,
+    generate_type_definition,
+)
 
 
 class Query(BaseModel):
@@ -28,18 +30,7 @@ class Query(BaseModel):
 class TextToImage(ServiceType):
     name = "text_to_image"
     args = {}
-
-    arg_types: Dict[str, ComponentSpecValueType] = {
-        "prompt": ComponentSpecValueType.string,
-        "negative_prompt": ComponentSpecValueType.string,
-        "sampling_steps": ComponentSpecValueType.int,
-        "width": ComponentSpecValueType.int,
-        "height": ComponentSpecValueType.int,
-        "seed": ComponentSpecValueType.int,
-        "batch_size": ComponentSpecValueType.int,
-        "batch_count": ComponentSpecValueType.int,
-        "guidance_scale": ComponentSpecValueType.float,
-    }
+    arg_types = generate_type_definition(Query)
 
     def __init__(
         self,

client/starwhale/api/_impl/service/types/types.py

@@ -2,8 +2,9 @@
 
 import abc
 import inspect
-from typing import Any, Dict, List, Union, Callable
+from typing import Any, Dict, List, Type, Union, Callable
 
+from pydantic import BaseModel
 from typing_extensions import Protocol
 
 from starwhale.utils import console
@@ -164,3 +165,62 @@ def all_components_are_gradio(
             all([isinstance(out, gradio.components.Component) for out in outputs]),
         ]
     )
+
+
+def generate_type_definition(
+    model: Type[BaseModel],
+) -> Dict[str, ComponentSpecValueType]:
+    """
+    Generate the type definition for a given model.
+
+    Args:
+        model (Type[BaseModel]): The model for which to generate the type definition.
+
+    Returns:
+        Dict[str, ComponentSpecValueType]: The generated type definition.
+
+    Raises:
+        ValueError: If the field type is not supported.
+    """
+    type_definition: Dict[str, ComponentSpecValueType] = {}
+    # we will get multiple type field_type, e.g. 'str', typing.Optional[str], typing.List[str], <class 'str'>
+    # if we use from __future__ import annotations, we will get 'str'
+    # else we will get typing.Optional[str] or <class 'str'>
+    for field_name, field_type in model.__annotations__.items():
+        field_type = str(field_type)
+
+        while True:
+            # remove the typing. prefix
+            if field_type.startswith("typing."):
+                field_type = field_type[7:]
+                continue
+            # remove the typing.Optional[] wrapper
+            if field_type.startswith("Optional["):
+                field_type = field_type[9:-1]
+                continue
+            # remove the typing[] wrapper
+            if field_type.startswith("typing["):
+                field_type = field_type[7:-1]
+                continue
+            # remove the <class ''> wrapper
+            if field_type.startswith("<class '"):
+                field_type = field_type[8:-2]
+                continue
+            break
+
+        if field_type == "str":
+            type_definition[field_name] = ComponentSpecValueType.string
+        elif field_type == "int":
+            type_definition[field_name] = ComponentSpecValueType.int
+        elif field_type == "float":
+            type_definition[field_name] = ComponentSpecValueType.float
+        elif field_type == "bool":
+            type_definition[field_name] = ComponentSpecValueType.bool
+        elif field_type.startswith("typing.List[") or field_type.startswith("List["):
+            type_definition[field_name] = ComponentSpecValueType.list
+        else:
+            # we do not support union type for now:
+            raise ValueError(
+                f"Unsupported field type: {field_type} with name {field_name}"
+            )
+    return type_definition

client/tests/sdk/test_types.py

@@ -1,10 +1,14 @@
+from typing import List, Optional
+
 import pytest
+from pydantic import BaseModel
 
 from starwhale.api._impl.service.types import all_components_are_gradio
 from starwhale.base.client.models.models import (
     ComponentValueSpecInt,
     ComponentSpecValueType,
 )
+from starwhale.api._impl.service.types.types import generate_type_definition
 from starwhale.api._impl.service.types.text_to_img import TextToImage
 
 
@@ -55,3 +59,33 @@ def test_text_to_image():
     assert t.args == {
         "sampling_steps": ComponentValueSpecInt(default_val=1),
     }
+
+
+def test_generate_type_definition():
+    class MyType(BaseModel):
+        a: int
+        b: str
+        c: float
+        d: bool
+        e: Optional[int] = None
+        f: Optional[str] = None
+        g: Optional[float] = None
+        h: Optional[bool] = None
+        i: List[int]
+        j: List[str]
+        k: Optional[List[int]] = None
+
+    types = generate_type_definition(MyType)
+    assert types == {
+        "a": ComponentSpecValueType.int,
+        "b": ComponentSpecValueType.string,
+        "c": ComponentSpecValueType.float,
+        "d": ComponentSpecValueType.bool,
+        "e": ComponentSpecValueType.int,
+        "f": ComponentSpecValueType.string,
+        "g": ComponentSpecValueType.float,
+        "h": ComponentSpecValueType.bool,
+        "i": ComponentSpecValueType.list,
+        "j": ComponentSpecValueType.list,
+        "k": ComponentSpecValueType.list,
+    }