fix(ollama): fix the bug in streaming output of ollama chat model (#1082)

Author: qbc2016

src/agentscope/model/_ollama_model.py

@@ -261,7 +261,7 @@ async def _parse_ollama_stream_completion_response(
                     print(f"Error parsing tool call input: {e}")
 
             # Generate response when there's new content or at final chunk
-            if chunk.done and contents:
+            if chunk.done or contents:
                 res = ChatResponse(
                     content=contents,
                     usage=usage,

tests/model_anthropic_test.py

@@ -318,7 +318,7 @@ async def mock_stream() -> AsyncGenerator:
             async for response in result:
                 responses.append(response)
 
-            self.assertGreater(len(responses), 0)
+            self.assertEqual(len(responses), 2)
             final_response = responses[-1]
             self.assertIsInstance(final_response, ChatResponse)
             expected_content = [

tests/model_ollama_test.py

@@ -281,7 +281,7 @@ async def test_streaming_response_processing(self) -> None:
             async for response in result:
                 responses.append(response)
 
-            self.assertGreaterEqual(len(responses), 1)
+            self.assertEqual(len(responses), 2)
             final_response = responses[-1]
             expected_content = [TextBlock(type="text", text="Hello there!")]
             self.assertEqual(final_response.content, expected_content)

tests/model_openai_test.py

@@ -244,7 +244,7 @@ async def test_streaming_response_processing(self) -> None:
             async for response in result:
                 responses.append(response)
 
-            self.assertGreaterEqual(len(responses), 1)
+            self.assertEqual(len(responses), 2)
             final_response = responses[-1]
             expected_content = [TextBlock(type="text", text="Hello there!")]
             self.assertEqual(final_response.content, expected_content)