Merge pull request #124 from pinecone-io/increase-upsert-defaults

change defaults

config/config.yaml

@@ -46,7 +46,7 @@ chat_engine:
         type: OpenAIRecordEncoder
         params:
           model_name: text-embedding-ada-002
-          batch_size: 100
+          batch_size: 400
 
       chunker:
         type: MarkdownChunker

src/canopy/knowledge_base/knowledge_base.py

@@ -490,7 +490,7 @@ def _query_index(self,
     def upsert(self,
                documents: List[Document],
                namespace: str = "",
-               batch_size: int = 100,
+               batch_size: int = 200,
                show_progress_bar: bool = False):
         """
         Upsert documents into the knowledge base.

src/canopy/knowledge_base/record_encoder/openai.py

@@ -17,7 +17,7 @@ class OpenAIRecordEncoder(DenseRecordEncoder):
     def __init__(self,
                  *,
                  model_name: str = "text-embedding-ada-002",
-                 batch_size: int = 100,
+                 batch_size: int = 400,
                  **kwargs):
         encoder = OpenAIEncoder(model_name)
         super().__init__(dense_encoder=encoder, batch_size=batch_size, **kwargs)

src/canopy_server/api_models.py

@@ -19,7 +19,7 @@ class ContextQueryRequest(BaseModel):
 
 class ContextUpsertRequest(BaseModel):
     documents: List[Document]
-    batch_size: int = 100
+    batch_size: int = 200
 
 
 class ContextDeleteRequest(BaseModel):