Batching logic

Reformatting
Adding VoyageAI to enum

Author: fzowl

redisvl/utils/vectorize/__init__.py

@@ -36,3 +36,5 @@ def vectorizer_from_dict(vectorizer: dict) -> BaseVectorizer:
         return MistralAITextVectorizer(model)
     elif vectorizer_type == Vectorizers.vertexai:
         return VertexAITextVectorizer(model)
+    elif vectorizer_type == Vectorizers.voyageai:
+        return VoyageAITextVectorizer(model)

redisvl/utils/vectorize/base.py

@@ -14,6 +14,7 @@ class Vectorizers(Enum):
     mistral = "mistral"
     vertexai = "vertexai"
     hf = "hf"
+    voyageai = "voyageai"
 
 
 class BaseVectorizer(BaseModel, ABC):

redisvl/utils/vectorize/text/voyageai.py

@@ -7,6 +7,7 @@
 
 from redisvl.utils.vectorize.base import BaseVectorizer
 
+
 # ignore that voyageai isn't imported
 # mypy: disable-error-code="name-defined"
 
@@ -207,7 +208,15 @@ def embed_many(
             raise TypeError("Truncation (optional) parameter is a bool.")
 
         if batch_size is None:
-            batch_size = 72 if self.model in ["voyage-2", "voyage-02"] else 7
+            batch_size = (
+                72
+                if self.model in ["voyage-2", "voyage-02"]
+                else (
+                    30
+                    if self.model == "voyage-3-lite"
+                    else (10 if self.model == "voyage-3" else 7)
+                )
+            )
 
         embeddings: List = []
         for batch in self.batchify(texts, batch_size, preprocess):
@@ -277,7 +286,15 @@ async def aembed_many(
             raise TypeError("Truncation (optional) parameter is a bool.")
 
         if batch_size is None:
-            batch_size = 72 if self.model in ["voyage-2", "voyage-02"] else 7
+            batch_size = (
+                72
+                if self.model in ["voyage-2", "voyage-02"]
+                else (
+                    30
+                    if self.model == "voyage-3-lite"
+                    else (10 if self.model == "voyage-3" else 7)
+                )
+            )
 
         embeddings: List = []
         for batch in self.batchify(texts, batch_size, preprocess):