From 7a11bd098f4c117df1a25e28ab5a8be3024e4080 Mon Sep 17 00:00:00 2001
From: fzowl <zoltan@voyageai.com>
Date: Sat, 21 Mar 2026 15:06:24 +0100
Subject: [PATCH] Migrate to Pydantic v2 APIs for Python 3.14 support

The previous fix (#46) swapped the pydantic import order but still fell
back to the broken pydantic.v1 path since Extra was removed in v2.

- Remove pydantic.v1 fallback and Extra import
- Replace Extra.forbid with string "forbid"
- Replace min_items with min_length (v2 rename)
- Replace .parse_obj() with .model_validate()
- Replace .dict() with .model_dump()
- Bump pydantic lower bound to >=2.7.4 (already required transitively
  by langchain-text-splitters via langchain-core)

Fixes #36
---
 pyproject.toml                           |  4 ++--
 voyageai/client.py                       |  2 +-
 voyageai/client_async.py                 |  2 +-
 voyageai/object/multimodal_embeddings.py | 20 ++++++++------------
 4 files changed, 12 insertions(+), 16 deletions(-)

diff --git a/pyproject.toml b/pyproject.toml
index a7c3664..dc2143f 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "voyageai"
-version = "0.3.8"
+version = "0.3.9"
 description = ""
 authors = ["Yujie Qian <yujieq@voyageai.com>"]
 readme = "README.md"
@@ -16,7 +16,7 @@ numpy = [
 ]
 aiolimiter = "*"
 pillow = "*"
-pydantic = ">=1.10.8"
+pydantic = ">=2.7.4"
 tokenizers = ">=0.14.0"
 langchain-text-splitters = ">=0.3.8"
 ffmpeg-python = "*"
diff --git a/voyageai/client.py b/voyageai/client.py
index 851cd5d..de058cc 100644
--- a/voyageai/client.py
+++ b/voyageai/client.py
@@ -170,7 +170,7 @@ def multimodal_embed(
                         truncation=truncation,
                         output_dtype=output_dtype,
                         output_dimension=output_dimension,
-                    ).dict(),
+                    ).model_dump(),
                     **self._params,
                 )
 
diff --git a/voyageai/client_async.py b/voyageai/client_async.py
index cebdaa3..5ae2455 100644
--- a/voyageai/client_async.py
+++ b/voyageai/client_async.py
@@ -170,7 +170,7 @@ async def multimodal_embed(
                         truncation=truncation,
                         output_dtype=output_dtype,
                         output_dimension=output_dimension,
-                    ).dict(),
+                    ).model_dump(),
                     **self._params,
                 )
         if response is None:
diff --git a/voyageai/object/multimodal_embeddings.py b/voyageai/object/multimodal_embeddings.py
index 1f4013b..a7f5e65 100644
--- a/voyageai/object/multimodal_embeddings.py
+++ b/voyageai/object/multimodal_embeddings.py
@@ -5,16 +5,12 @@
 
 import PIL.Image
 import PIL.ImageFile
+from pydantic import BaseModel, Field, ValidationError
 
 from voyageai import error
 from voyageai.api_resources import VoyageResponse
 from voyageai.video_utils import Video
 
-try:
-    from pydantic import BaseModel, Extra, Field, ValidationError
-except ImportError:
-    from pydantic.v1 import BaseModel, Extra, Field, ValidationError
-
 
 class MultimodalEmbeddingsObject:
     def __init__(self, response: Optional[VoyageResponse] = None):
@@ -51,7 +47,7 @@ class MultimodalInputSegmentText(BaseModel):
     text: str
 
     class Config:
-        extra = Extra.forbid
+        extra = "forbid"
 
 
 class MultimodalInputSegmentImageURL(BaseModel):
@@ -59,7 +55,7 @@ class MultimodalInputSegmentImageURL(BaseModel):
     image_url: str
 
     class Config:
-        extra = Extra.forbid
+        extra = "forbid"
 
 
 class MultimodalInputSegmentImageBase64(BaseModel):
@@ -67,7 +63,7 @@ class MultimodalInputSegmentImageBase64(BaseModel):
     image_base64: str
 
     class Config:
-        extra = Extra.forbid
+        extra = "forbid"
 
 
 class MultimodalInputSegmentVideoURL(BaseModel):
@@ -75,7 +71,7 @@ class MultimodalInputSegmentVideoURL(BaseModel):
     video_url: str
 
     class Config:
-        extra = Extra.forbid
+        extra = "forbid"
 
 
 class MultimodalInputSegmentVideoBase64(BaseModel):
@@ -83,7 +79,7 @@ class MultimodalInputSegmentVideoBase64(BaseModel):
     video_base64: str
 
     class Config:
-        extra = Extra.forbid
+        extra = "forbid"
 
 
 class MultimodalInput(BaseModel):
@@ -98,7 +94,7 @@ class MultimodalInput(BaseModel):
             ],
             Field(discriminator="type"),
         ]
-    ] = Field(..., min_items=1)
+    ] = Field(..., min_length=1)
 
 
 class MultimodalInputRequest(BaseModel):
@@ -195,7 +191,7 @@ 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.parse_obj(input_data)
+            return MultimodalInput.model_validate(input_data)
         except ValidationError as ve:
             raise ValueError(f"Validation error for input at index {idx}: {ve}") from ve