fix(bedrock): add Bedrock-native format conversion to OpenAI format

instructor/process_response.py

@@ -754,6 +754,18 @@ def _prepare_bedrock_converse_kwargs_internal(
 
     See: https://boto3.amazonaws.com/v1/documentation/api/latest/reference/services/bedrock-runtime/client/converse.html
     """
+    # Handle Bedrock-native system parameter format: system=[{'text': '...'}]
+    # Convert to OpenAI format by adding to messages as system role
+    if "system" in call_kwargs and isinstance(call_kwargs["system"], list):
+        system_content = call_kwargs.pop("system")
+        if system_content and isinstance(system_content[0], dict) and "text" in system_content[0]:
+            # Convert system=[{'text': '...'}] to OpenAI format
+            system_text = system_content[0]["text"]
+            if "messages" not in call_kwargs:
+                call_kwargs["messages"] = []
+            # Insert system message at beginning
+            call_kwargs["messages"].insert(0, {"role": "system", "content": system_text})
+
     # Bedrock expects 'modelId' over 'model'
     if "model" in call_kwargs and "modelId" not in call_kwargs:
         call_kwargs["modelId"] = call_kwargs.pop("model")
@@ -842,7 +854,10 @@ def _prepare_bedrock_converse_kwargs_internal(
                 if "content" in current_message_for_api:
                     if isinstance(content, str):
                         current_message_for_api["content"] = [{"text": content}]
-                    else:  # Content is not a string (e.g., None, list, int).
+                    elif isinstance(content, list) and content and isinstance(content[0], dict) and "text" in content[0]:
+                        # Handle Bedrock-native content format: [{'text': "..."}]
+                        current_message_for_api["content"] = content
+                    else:  # Content is not a string or supported list format (e.g., None, int, unsupported list).
                         # This matches the original behavior which raised for any non-string content.
                         raise NotImplementedError(
                             "Non-text prompts are not currently supported in the Bedrock provider."

tests/test_process_response.py

@@ -1,6 +1,6 @@
 from typing_extensions import TypedDict
 from pydantic import BaseModel
-from instructor.process_response import handle_response_model
+from instructor.process_response import handle_response_model, _prepare_bedrock_converse_kwargs_internal
 
 
 def test_typed_dict_conversion() -> None:
@@ -16,3 +16,100 @@
 
     _, pydantic_user_tool_definition = handle_response_model(User)
     assert user_tool_definition == pydantic_user_tool_definition
+
+
+def test_bedrock_system_parameter_conversion() -> None:
+    """Test conversion of Bedrock-native system=[{'text': '...'}] to OpenAI format"""
+    call_kwargs = {
+        "system": [{'text': 'You are a helpful assistant.'}],
+        "messages": [{'role': 'user', 'content': 'Hello'}]
+    }
+
+    result = _prepare_bedrock_converse_kwargs_internal(call_kwargs)
+
+    # System parameter should be removed and converted to system message in messages
+    assert "system" not in result
+    assert len(result["messages"]) == 2
+    assert result["messages"][0] == {"role": "system", "content": "You are a helpful assistant."}
+    assert result["messages"][1] == {'role': 'user', 'content': 'Hello'}
+
+
+def test_bedrock_content_list_format() -> None:
+    """Test handling of content=[{'text': '...'}] format in messages"""
+    call_kwargs = {
+        "messages": [
+            {'role': 'user', 'content': [{'text': "Extract: Jason is 22 years old"}]}
+        ]
+    }
+
+    result = _prepare_bedrock_converse_kwargs_internal(call_kwargs)
+
+    # Content should remain in list format for Bedrock
+    assert len(result["messages"]) == 1
+    assert result["messages"][0]["content"] == [{'text': "Extract: Jason is 22 years old"}]
+
+
+def test_bedrock_combined_system_and_content_format() -> None:
+    """Test the exact issue scenario: system=[{'text': '...'}] + content=[{'text': '...'}]"""
+    call_kwargs = {
+        "modelId": "anthropic.claude-3-haiku-20240307-v1:0",
+        "system": [{'text': 'You are a helpful assistant.'}],
+        "messages": [{'role': 'user', 'content': [{'text': "Extract: Jason is 22 years old"}]}]
+    }
+
+    result = _prepare_bedrock_converse_kwargs_internal(call_kwargs)
+
+    # Should convert system parameter and preserve content format
+    assert "system" not in result
+    assert result["modelId"] == "anthropic.claude-3-haiku-20240307-v1:0"
+    assert len(result["messages"]) == 2
+    assert result["messages"][0] == {"role": "system", "content": "You are a helpful assistant."}
+    assert result["messages"][1]["role"] == "user"
+    assert result["messages"][1]["content"] == [{'text': "Extract: Jason is 22 years old"}]
+
+
+def test_bedrock_backward_compatibility_openai_format() -> None:
+    """Test that existing OpenAI format still works"""
+    call_kwargs = {
+        "messages": [
+            {"role": "system", "content": "You are a helpful assistant."},
+            {"role": "user", "content": "Extract: Jason is 22 years old"}
+        ]
+    }
+
+    result = _prepare_bedrock_converse_kwargs_internal(call_kwargs)
+
+    # Should process OpenAI format correctly (string content converted to text objects)
+    assert len(result["messages"]) == 2
+    assert result["messages"][0]["role"] == "system"
+    assert result["messages"][0]["content"] == "You are a helpful assistant."  # System content stays as string
+    assert result["messages"][1]["role"] == "user"
+    assert result["messages"][1]["content"] == [{"text": "Extract: Jason is 22 years old"}]
+
+
+def test_bedrock_model_id_conversion() -> None:
+    """Test that model parameter is converted to modelId"""
+    call_kwargs = {
+        "model": "anthropic.claude-3-haiku-20240307-v1:0",
+        "messages": []
+    }
+
+    result = _prepare_bedrock_converse_kwargs_internal(call_kwargs)
+
+    assert "model" not in result
+    assert result["modelId"] == "anthropic.claude-3-haiku-20240307-v1:0"
+
+
+def test_bedrock_invalid_content_format() -> None:
+    """Test that invalid content formats still raise NotImplementedError"""
+    call_kwargs = {
+        "messages": [
+            {'role': 'user', 'content': 12345}  # Invalid content type
+        ]
+    }
+
+    try:
+        _prepare_bedrock_converse_kwargs_internal(call_kwargs)
+        assert False, "Should have raised NotImplementedError"
+    except NotImplementedError as e:
+        assert "Non-text prompts are not currently supported" in str(e)