Fix OpenAI compatibility issues for newer models and Azure config (#727)

Use max_completion_tokens for OpenAI and Azure OpenAI providers:
The OpenAI API deprecated max_tokens in favor of
max_completion_tokens for chat completions. Newer models
(gpt-4o, o1, o3) reject the old parameter with a 400 error.

AZURE_API_VERSION env var now overrides the default API version:
(falls back to 2024-12-01-preview).

Update tests to test for expected structures

tests/integration/test_text_completion_integration.py

@@ -93,7 +93,7 @@ class TestTextCompletionIntegration:
         
         assert call_args.kwargs['model'] == "gpt-3.5-turbo"
         assert call_args.kwargs['temperature'] == 0.7
-        assert call_args.kwargs['max_tokens'] == 1024
+        assert call_args.kwargs['max_completion_tokens'] == 1024
         assert len(call_args.kwargs['messages']) == 1
         assert call_args.kwargs['messages'][0]['role'] == "user"
         assert "You are a helpful assistant." in call_args.kwargs['messages'][0]['content'][0]['text']
@@ -134,7 +134,7 @@ class TestTextCompletionIntegration:
             call_args = mock_openai_client.chat.completions.create.call_args
             assert call_args.kwargs['model'] == config['model']
             assert call_args.kwargs['temperature'] == config['temperature']
-            assert call_args.kwargs['max_tokens'] == config['max_output']
+            assert call_args.kwargs['max_completion_tokens'] == config['max_output']
             
             # Reset mock for next iteration
             mock_openai_client.reset_mock()
@@ -286,7 +286,7 @@ class TestTextCompletionIntegration:
         # were removed in #561 as unnecessary parameters
         assert 'model' in call_args.kwargs
         assert 'temperature' in call_args.kwargs
-        assert 'max_tokens' in call_args.kwargs
+        assert 'max_completion_tokens' in call_args.kwargs
 
         # Verify result structure
         assert hasattr(result, 'text')
@@ -362,7 +362,7 @@ class TestTextCompletionIntegration:
         call_args = mock_openai_client.chat.completions.create.call_args
         assert call_args.kwargs['model'] == "gpt-4"
         assert call_args.kwargs['temperature'] == 0.8
-        assert call_args.kwargs['max_tokens'] == 2048
+        assert call_args.kwargs['max_completion_tokens'] == 2048
         # Note: top_p, frequency_penalty, and presence_penalty
         # were removed in #561 as unnecessary parameters
 

tests/integration/test_text_completion_streaming_integration.py

@@ -201,7 +201,7 @@ class TestTextCompletionStreaming:
         call_args = mock_streaming_openai_client.chat.completions.create.call_args
         assert call_args.kwargs['model'] == "gpt-4"
         assert call_args.kwargs['temperature'] == 0.5
-        assert call_args.kwargs['max_tokens'] == 2048
+        assert call_args.kwargs['max_completion_tokens'] == 2048
         assert call_args.kwargs['stream'] is True
 
         # Verify chunks have correct model

tests/unit/test_text_completion/test_azure_openai_processor.py

@@ -108,7 +108,7 @@ class TestAzureOpenAIProcessorSimple(IsolatedAsyncioTestCase):
                 }]
             }],
             temperature=0.0,
-            max_tokens=4192,
+            max_completion_tokens=4192,
             top_p=1
         )
 
@@ -399,7 +399,7 @@ class TestAzureOpenAIProcessorSimple(IsolatedAsyncioTestCase):
         # Verify other parameters
         assert call_args[1]['model'] == 'gpt-4'
         assert call_args[1]['temperature'] == 0.5
-        assert call_args[1]['max_tokens'] == 1024
+        assert call_args[1]['max_completion_tokens'] == 1024
         assert call_args[1]['top_p'] == 1
 
     @patch('trustgraph.model.text_completion.azure_openai.llm.AzureOpenAI')

tests/unit/test_text_completion/test_openai_processor.py

@@ -102,7 +102,7 @@ class TestOpenAIProcessorSimple(IsolatedAsyncioTestCase):
                 }]
             }],
             temperature=0.0,
-            max_tokens=4096
+            max_completion_tokens=4096
         )
 
     @patch('trustgraph.model.text_completion.openai.llm.OpenAI')
@@ -380,7 +380,7 @@ class TestOpenAIProcessorSimple(IsolatedAsyncioTestCase):
         # Verify other parameters
         assert call_args[1]['model'] == 'gpt-3.5-turbo'
         assert call_args[1]['temperature'] == 0.5
-        assert call_args[1]['max_tokens'] == 1024
+        assert call_args[1]['max_completion_tokens'] == 1024
 
 
     @patch('trustgraph.model.text_completion.openai.llm.OpenAI')

trustgraph-flow/trustgraph/model/text_completion/azure_openai/llm.py

@@ -20,7 +20,7 @@ default_ident = "text-completion"
 
 default_temperature = 0.0
 default_max_output = 4192
-default_api = "2024-12-01-preview"
+default_api = os.getenv("AZURE_API_VERSION", "2024-12-01-preview")
 default_endpoint = os.getenv("AZURE_ENDPOINT", None)
 default_token = os.getenv("AZURE_TOKEN", None)
 default_model = os.getenv("AZURE_MODEL", None)
@@ -90,7 +90,7 @@ class Processor(LlmService):
                     }
                 ],
                 temperature=effective_temperature,
-                max_tokens=self.max_output,
+                max_completion_tokens=self.max_output,
                 top_p=1,
             )
 
@@ -159,7 +159,7 @@ class Processor(LlmService):
                     }
                 ],
                 temperature=effective_temperature,
-                max_tokens=self.max_output,
+                max_completion_tokens=self.max_output,
                 top_p=1,
                 stream=True,
                 stream_options={"include_usage": True}

trustgraph-flow/trustgraph/model/text_completion/openai/llm.py

@@ -86,7 +86,7 @@ class Processor(LlmService):
                     }
                 ],
                 temperature=effective_temperature,
-                max_tokens=self.max_output,
+                max_completion_tokens=self.max_output,
             )
             
             inputtokens = resp.usage.prompt_tokens
@@ -152,7 +152,7 @@ class Processor(LlmService):
                     }
                 ],
                 temperature=effective_temperature,
-                max_tokens=self.max_output,
+                max_completion_tokens=self.max_output,
                 stream=True,
                 stream_options={"include_usage": True}
             )