[Multimodal] Revise dependency versions (#15)

pyproject.toml

@@ -1,25 +1,25 @@
 [tool.poetry]
 name = "voyageai"
-version = "0.3.0rc0"
+version = "0.3.0"
 description = ""
 authors = ["Yujie Qian <[email protected]>"]
 readme = "README.md"
 
 [tool.poetry.dependencies]
-python = "^3.7.1"
-requests = "^2.20"
-aiohttp = "^3.5"
-tenacity = ">=8.0.1"
-numpy = ">=1.11"
-aiolimiter = "^1.1.0"
-pillow = "^9.5.0"
-pydantic = "^2.5.3"
-tokenizers = "^0.20.0"
+python = "^3.9,<3.13"
+requests = "*"
+aiohttp = "*"
+tenacity = "*"
+oldest-supported-numpy = "*"
+aiolimiter = "*"
+pillow = "*"
+pydantic = ">=1.10.8"
+tokenizers = ">=0.14.0"
 
 [tool.poetry.group.test.dependencies]
 pytest = "^7.4.2"
 pytest-asyncio = "^0.21.1"
 
 [build-system]
 requires = ["poetry-core"]
-build-backend = "poetry.core.masonry.api"
+build-backend = "poetry.core.masonry.api"

tests/test_client_multimodal.py

@@ -105,9 +105,9 @@ def embedding_dimension():
 
 sample_input_list_text_01 = ["this is an image of a blue sailboat on a lake."]
 
-sample_input_list_img_01 = [Image.open("example_image_01.jpg")]
+sample_input_list_img_01 = [Image.open("tests/example_image_01.jpg")]
 
-sample_input_list_img_02 = [Image.open("example_image_01.jpg").resize((256, 256))]
+sample_input_list_img_02 = [Image.open("tests/example_image_01.jpg").resize((256, 256))]
 
 sample_input_list_img_03 = [Image.new("L", (400, 400), color=128)]
 

voyageai/_base.py

@@ -13,7 +13,7 @@
 import voyageai
 import voyageai.error as error
 from voyageai.object.multimodal_embeddings import MultimodalInputRequest, MultimodalInputSegmentText, \
-    MultimodalInputSegmentImageURL, MultimodalInputSegmentImageBase64
+    MultimodalInputSegmentImageURL, MultimodalInputSegmentImageBase64, MultimodalEmbeddingsObject
 from voyageai.util import default_api_key
 from voyageai.object import EmbeddingsObject, RerankingObject
 
@@ -35,7 +35,7 @@ def _get_client_config(
     return data_dict
 
 
-class _BaseClient:
+class _BaseClient(ABC):
     """Voyage AI Client
 
     Args:
@@ -58,15 +58,14 @@ def __init__(
             "request_timeout": timeout,
         }
 
-
     @abstractmethod    
     def embed(
         self,
         texts: List[str],
         model: Optional[str] = None,
         input_type: Optional[str] = None,
         truncation: bool = True,
-    ) -> Union[EmbeddingsObject, Awaitable[EmbeddingsObject]]:
+    ) -> EmbeddingsObject:
         pass
 
     @abstractmethod
@@ -77,7 +76,17 @@ def rerank(
         model: str,
         top_k: Optional[int] = None,
         truncation: bool = True,
-    ) -> Union[RerankingObject, Awaitable[RerankingObject]]:
+    ) -> RerankingObject:
+        pass
+
+    @abstractmethod
+    def multimodal_embed(
+        self,
+        inputs: Union[List[Dict], List[List[Union[str, PIL.Image.Image]]]],
+        model: str,
+        input_type: Optional[str] = None,
+        truncation: bool = True,
+    ) -> MultimodalEmbeddingsObject:
         pass
 
     @functools.lru_cache()
@@ -130,7 +139,7 @@ def count_usage(
         self,
         inputs: Union[List[Dict], List[List[Union[str, PIL.Image.Image]]]],
         model: str,
-    ) -> dict[str, int]:
+    ) -> Dict[str, int]:
         """
         This method returns estimated usage metrics for the provided input.
         Currently, only multimodal models are supported. Image URL segments are not supported.

voyageai/client.py

@@ -120,7 +120,7 @@ def multimodal_embed(
                         model=model,
                         input_type=input_type,
                         truncation=truncation,
-                    ).model_dump(),
+                    ).dict(),
                     **self._params,
                 )
 

voyageai/client_async.py

@@ -122,7 +122,7 @@ async def multimodal_embed(
                         model=model,
                         input_type=input_type,
                         truncation=truncation,
-                    ).model_dump(),
+                    ).dict(),
                     **self._params,
                 )
 

voyageai/object/multimodal_embeddings.py

@@ -3,12 +3,15 @@
 import PIL.ImageFile
 from io import BytesIO
 from enum import Enum
-from pydantic import BaseModel, Field, Extra, ValidationError
 from typing import List, Optional, Union, Dict, Literal, Annotated
 
 from voyageai import error
 from voyageai.api_resources import VoyageResponse
 
+try:
+    from pydantic.v1 import BaseModel, Field, Extra, ValidationError
+except ImportError:
+    from pydantic import BaseModel, Field, Extra, ValidationError
 
 class MultimodalEmbeddingsObject:
     def __init__(self, response: Optional[VoyageResponse] = None):
@@ -74,7 +77,7 @@ class MultimodalInput(BaseModel):
             ],
             Field(discriminator="type"),
         ]
-    ] = Field(..., min_length=1)
+    ] = Field(..., min_items=1)
 
 
 class MultimodalInputRequest(BaseModel):
@@ -167,7 +170,7 @@ def _process_dict_input(cls, input_data: Dict, idx: int) -> MultimodalInput:
             raise ValueError(f"Input at index {idx} is missing the 'content' field.")
 
         try:
-            return MultimodalInput.model_validate(input_data)
+            return MultimodalInput.parse_obj(input_data)
         except ValidationError as ve:
             raise ValueError(f"Validation error for input at index {idx}: {ve}") from ve