llama-index-llms-bedrock-converse 0.12.5__tar.gz → 0.12.6__tar.gz

{llama_index_llms_bedrock_converse-0.12.5 → llama_index_llms_bedrock_converse-0.12.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: llama-index-llms-bedrock-converse
-Version: 0.12.5
+Version: 0.12.6
 Summary: llama-index llms bedrock converse integration
 Author-email: Your Name <you@example.com>
 License-Expression: MIT

{llama_index_llms_bedrock_converse-0.12.5 → llama_index_llms_bedrock_converse-0.12.6}/llama_index/llms/bedrock_converse/utils.py

@@ -12,10 +12,12 @@ from typing import (
     Literal,
     Union,
 )
+from botocore.exceptions import ClientError
 from typing_extensions import TypedDict
 from tenacity import (
     before_sleep_log,
     retry,
+    retry_if_exception,
     retry_if_exception_type,
     stop_after_attempt,
     wait_exponential,
@@ -566,11 +568,38 @@ def _create_retry_decorator(client: Any, max_retries: int) -> Callable[[Any], An
         reraise=True,
         stop=stop_after_attempt(max_retries),
         wait=wait_exponential(multiplier=1, min=min_seconds, max=max_seconds),
-        retry=(retry_if_exception_type(client.exceptions.ThrottlingException)),
+        retry=(
+            retry_if_exception_type(
+                (
+                    client.exceptions.ThrottlingException,
+                    client.exceptions.InternalServerException,
+                    client.exceptions.ServiceUnavailableException,
+                    client.exceptions.ModelTimeoutException,
+                )
+            )
+        ),
         before_sleep=before_sleep_log(logger, logging.WARNING),
     )
 
 
+RETRYABLE_ERROR_CODES = frozenset(
+    {
+        "ThrottlingException",
+        "InternalServerException",
+        "ServiceUnavailableException",
+        "ModelTimeoutException",
+    }
+)
+
+
+def _is_retryable_client_error(exception: BaseException) -> bool:
+    """Check if an exception is a retryable ClientError from botocore."""
+    if isinstance(exception, ClientError):
+        error_code = exception.response.get("Error", {}).get("Code", "")
+        return error_code in RETRYABLE_ERROR_CODES
+    return False
+
+
 def _create_retry_decorator_async(max_retries: int) -> Callable[[Any], Any]:
     min_seconds = 4
     max_seconds = 10
@@ -588,9 +617,7 @@ def _create_retry_decorator_async(max_retries: int) -> Callable[[Any], Any]:
         reraise=True,
         stop=stop_after_attempt(max_retries),
         wait=wait_exponential(multiplier=1, min=min_seconds, max=max_seconds),
-        retry=(
-            retry_if_exception_type()
-        ),  # TODO: Add throttling exception in async version
+        retry=retry_if_exception(_is_retryable_client_error),
         before_sleep=before_sleep_log(logger, logging.WARNING),
     )
 

{llama_index_llms_bedrock_converse-0.12.5 → llama_index_llms_bedrock_converse-0.12.6}/pyproject.toml

@@ -29,7 +29,7 @@ dev = [
 
 [project]
 name = "llama-index-llms-bedrock-converse"
-version = "0.12.5"
+version = "0.12.6"
 description = "llama-index llms bedrock converse integration"
 authors = [{name = "Your Name", email = "you@example.com"}]
 requires-python = ">=3.9,<4.0"