feat: add batch request support for text-embedding-v3 model (#26375)

PR title: “langchain: add batch request support for text-embedding-v3
model”

PR message:

• Description: This PR introduces batch request support for the
text-embedding-v3 model within LangChain. The new functionality allows
users to process multiple text inputs in a single request, improving
efficiency and performance for high-volume applications.
	•	Issue: This PR addresses #<issue_number> (if applicable).
• Dependencies: No new external dependencies are required for this
change.
• Twitter handle: If announced on Twitter, please mention me at
@yourhandle.

Add tests and docs:

1. Added unit tests to cover the batch request functionality, ensuring
it operates without requiring network access.
2. Included an example notebook demonstrating the batch request feature,
located in docs/docs/integrations.

Lint and test: All required formatting and linting checks have been
performed using make format and make lint. The changes have been
verified with make test to ensure compatibility.

Additional notes:

	•	The changes are fully backwards compatible.
• No modifications were made to pyproject.toml, ensuring no new
dependencies were added.
• The update only affects the langchain package and does not involve
other packages.

---------

Co-authored-by: Chester Curme <chester.curme@gmail.com>

libs/community/langchain_community/embeddings/dashscope.py

@@ -23,6 +23,8 @@ from tenacity import (
 
 logger = logging.getLogger(__name__)
 
+BATCH_SIZE = {"text-embedding-v1": 25, "text-embedding-v2": 25, "text-embedding-v3": 6}
+
 
 def _create_retry_decorator(embeddings: DashScopeEmbeddings) -> Callable[[Any], Any]:
     multiplier = 1
@@ -49,9 +51,12 @@ def embed_with_retry(embeddings: DashScopeEmbeddings, **kwargs: Any) -> Any:
         i = 0
         input_data = kwargs["input"]
         input_len = len(input_data) if isinstance(input_data, list) else 1
+        batch_size = BATCH_SIZE.get(kwargs["model"], 25)
         while i < input_len:
             kwargs["input"] = (
-                input_data[i : i + 25] if isinstance(input_data, list) else input_data
+                input_data[i : i + batch_size]
+                if isinstance(input_data, list)
+                else input_data
             )
             resp = embeddings.client.call(**kwargs)
             if resp.status_code == 200:
@@ -67,7 +72,7 @@ def embed_with_retry(embeddings: DashScopeEmbeddings, **kwargs: Any) -> Any:
                     f"code: {resp.code} \n message: {resp.message}",
                     response=resp,
                 )
-            i += 25
+            i += batch_size
         return result
 
     return _embed_with_retry(**kwargs)