From 143d7b88a91f89cd0c21e84bbb394747ecf9eedf Mon Sep 17 00:00:00 2001
From: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Date: Fri, 24 Oct 2025 18:21:38 +0000
Subject: [PATCH] Fix Bedrock cross-region inference profile support
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

- Add support for model_id parameter in BedrockCompletion
- When model_id is provided, use it instead of model for API calls
- This fixes issue #3791 where cross-region inference profiles
  (which require ARN as model_id) were not working
- Add comprehensive tests for cross-region inference profiles
- All existing bedrock tests pass

Co-Authored-By: João <joao@crewai.com>
---
 .../llms/providers/bedrock/completion.py      |  7 +-
 lib/crewai/tests/llms/bedrock/test_bedrock.py | 74 +++++++++++++++++++
 2 files changed, 79 insertions(+), 2 deletions(-)

diff --git a/lib/crewai/src/crewai/llms/providers/bedrock/completion.py b/lib/crewai/src/crewai/llms/providers/bedrock/completion.py
index 4607316c5..3e71bcddb 100644
--- a/lib/crewai/src/crewai/llms/providers/bedrock/completion.py
+++ b/lib/crewai/src/crewai/llms/providers/bedrock/completion.py
@@ -175,8 +175,11 @@ class BedrockCompletion(BaseLLM):
             guardrail_config: Guardrail configuration for content filtering
             additional_model_request_fields: Model-specific request parameters
             additional_model_response_field_paths: Custom response field paths
-            **kwargs: Additional parameters
+            **kwargs: Additional parameters (including model_id for cross-region inference)
         """
+        # Extract model_id from kwargs if provided (for cross-region inference profiles)
+        custom_model_id = kwargs.pop("model_id", None)
+        
         # Extract provider from kwargs to avoid duplicate argument
         kwargs.pop("provider", None)
 
@@ -230,7 +233,7 @@ class BedrockCompletion(BaseLLM):
         self.supports_streaming = True
 
         # Handle inference profiles for newer models
-        self.model_id = model
+        self.model_id = custom_model_id if custom_model_id else model
 
     def call(
         self,
diff --git a/lib/crewai/tests/llms/bedrock/test_bedrock.py b/lib/crewai/tests/llms/bedrock/test_bedrock.py
index 9fd172cc6..36e0434cb 100644
--- a/lib/crewai/tests/llms/bedrock/test_bedrock.py
+++ b/lib/crewai/tests/llms/bedrock/test_bedrock.py
@@ -18,6 +18,7 @@ def mock_aws_credentials():
         "AWS_SECRET_ACCESS_KEY": "test-secret-key",
         "AWS_DEFAULT_REGION": "us-east-1"
     }):
+        import crewai.llms.providers.bedrock.completion
         # Mock boto3 Session to prevent actual AWS connections
         with patch('crewai.llms.providers.bedrock.completion.Session') as mock_session_class:
             # Create mock session instance
@@ -736,3 +737,76 @@ def test_bedrock_client_error_handling():
         with pytest.raises(RuntimeError) as exc_info:
             llm.call("Hello")
         assert "throttled" in str(exc_info.value).lower()
+
+
+def test_bedrock_cross_region_inference_profile():
+    """
+    Test that Bedrock supports cross-region inference profiles with model_id parameter.
+    
+    This tests the fix for issue #3791 where cross-region inference profiles
+    (which require using ARN as model_id) were not working in version 1.20.0.
+    
+    When using cross-region inference profiles, users need to:
+    1. Set model to the base model name (e.g., "bedrock/anthropic.claude-sonnet-4-20250514-v1:0")
+    2. Set model_id to the inference profile ARN
+    
+    The BedrockCompletion should use the model_id parameter when provided,
+    not the model parameter, for the actual API call.
+    """
+    # Test with cross-region inference profile ARN
+    inference_profile_arn = "arn:aws:bedrock:us-east-1:123456789012:inference-profile/us.anthropic.claude-sonnet-4-20250514-v1:0"
+    
+    llm = LLM(
+        model="bedrock/anthropic.claude-sonnet-4-20250514-v1:0",
+        model_id=inference_profile_arn,
+        temperature=0.3,
+        max_tokens=4000,
+    )
+    
+    from crewai.llms.providers.bedrock.completion import BedrockCompletion
+    assert isinstance(llm, BedrockCompletion)
+    
+    assert llm.model_id == inference_profile_arn
+    assert llm.model == "anthropic.claude-sonnet-4-20250514-v1:0"
+    
+    # Verify that the client.converse call would use the correct model_id
+    with patch.object(llm.client, 'converse') as mock_converse:
+        mock_converse.return_value = {
+            'output': {
+                'message': {
+                    'role': 'assistant',
+                    'content': [{'text': 'Test response'}]
+                }
+            },
+            'usage': {
+                'inputTokens': 10,
+                'outputTokens': 5,
+                'totalTokens': 15
+            }
+        }
+        
+        llm.call("Test message")
+        
+        # Verify the converse call was made with the inference profile ARN
+        mock_converse.assert_called_once()
+        call_kwargs = mock_converse.call_args[1]
+        assert call_kwargs['modelId'] == inference_profile_arn
+
+
+def test_bedrock_model_id_parameter_takes_precedence():
+    """
+    Test that when both model and model_id are provided, model_id takes precedence
+    for the actual API call, while model is used for internal identification.
+    """
+    custom_model_id = "custom-model-identifier"
+    
+    llm = LLM(
+        model="bedrock/anthropic.claude-3-5-sonnet-20241022-v2:0",
+        model_id=custom_model_id,
+    )
+    
+    from crewai.llms.providers.bedrock.completion import BedrockCompletion
+    assert isinstance(llm, BedrockCompletion)
+    
+    assert llm.model_id == custom_model_id
+    assert llm.model == "anthropic.claude-3-5-sonnet-20241022-v2:0"