Merge 3dd7ca7f52 into 2a9fad611f

litellm/llms/bedrock/chat/invoke_transformations/amazon_openai_transformation.py

@@ -82,7 +82,12 @@ class AmazonBedrockOpenAIConfig(OpenAIGPTConfig, BaseAWSLLM):
 
         Uses the standard Bedrock invoke endpoint format.
         """
-        model_id = self._get_openai_model_id(model)
+        model_id = optional_params.pop("model_id", None) or litellm_params.get("model_id")
+        if model_id is not None:
+            model_id = CommonUtils.encode_bedrock_runtime_modelid_arn(str(model_id))
+        else:
+            model_id = self._get_openai_model_id(model)
+            model_id = CommonUtils.encode_bedrock_runtime_modelid_arn(model_id)
 
         # Get AWS region
         aws_region_name: Final = self._get_aws_region_name(optional_params=optional_params, model=model)
@@ -95,8 +100,7 @@ class AmazonBedrockOpenAIConfig(OpenAIGPTConfig, BaseAWSLLM):
             aws_region_name=aws_region_name,
         )
 
-        # Encode model ID for ARNs (e.g., :imported-model/ -> :imported-model%2F)
-        model_id = CommonUtils.encode_bedrock_runtime_modelid_arn(model_id)
+        # Model ID is already encoded above
 
         invoke_route: Final = "invoke-with-response-stream" if stream else "invoke"
         return join_api_base(endpoint_url, f"model/{model_id}/{invoke_route}")
@@ -141,8 +145,9 @@ class AmazonBedrockOpenAIConfig(OpenAIGPTConfig, BaseAWSLLM):
         Removes AWS-specific params and stream param (handled separately in URL),
         then delegates to parent class for standard OpenAI request transformation.
         """
-        # Remove stream from optional_params as it's handled separately in URL
+        # Remove stream and model_id from optional_params as they're handled separately in URL
         optional_params.pop("stream", None)
+        optional_params.pop("model_id", None)
 
         # Remove AWS-specific params that shouldn't be in the request body
         inference_params: Final = {k: v for k, v in optional_params.items() if k not in self.aws_authentication_params}

tests/unit/llms/bedrock/chat/invoke_transformations/test_amazon_openai_transformation.py

@@ -51,3 +51,29 @@ def test_openai_imported_model_preserves_message_types_and_request_parameters()
         "https://bedrock-runtime.us-east-1.amazonaws.com/model/"
         "arn:aws:bedrock:us-east-1:117159858402:imported-model%2Fm4gc1mrfuddy/invoke"
     )
+
+
+def test_openai_imported_model_with_model_id_application_inference_profile() -> None:
+    model: Final = "bedrock/openai/global.openai.gpt-6-luna"
+    profile_arn: Final = "arn:aws:bedrock:ap-northeast-2:123456789012:application-inference-profile/abc123def456"
+    config: Final = AmazonBedrockOpenAIConfig()
+    request: Final = config.transform_request(
+        model=model,
+        messages=[{"role": "user", "content": "hi"}],
+        optional_params={"model_id": profile_arn, "max_tokens": 100},
+        litellm_params={"model_id": profile_arn},
+        headers={},
+    )
+    url: Final = config.get_complete_url(
+        api_base=None,
+        api_key=None,
+        model=model,
+        optional_params={"aws_region_name": "ap-northeast-2", "model_id": profile_arn},
+        litellm_params={"model_id": profile_arn},
+        stream=False,
+    )
+    assert "model_id" not in request
+    assert url == (
+        "https://bedrock-runtime.ap-northeast-2.amazonaws.com/model/"
+        "arn:aws:bedrock:ap-northeast-2:123456789012:application-inference-profile%2Fabc123def456/invoke"
+    )