修复传入max_tokens=1时的报错 (#3068)

* 修复传入max_tokens=1时的报错

* 修复传入max_tokens=1时的报错

* 修复传入max_tokens=1时的报错

* 修复传入max_tokens=1时的报错

* 修复传入max_tokens=1时的报错

* 修复传入max_tokens=1时的报错

fastdeploy/output/token_processor.py

@@ -344,6 +344,7 @@ class TokenProcessor:
                     first_token_time=time.time() - task.inference_start_time,
                     time_in_queue=task.schedule_start_time - task.preprocess_end_time,
                     preprocess_cost_time=task.preprocess_end_time - task.preprocess_start_time,
+                    request_start_time=task.arrival_time,
                 )
 
                 self._record_first_token_metrics(task, current_time)
@@ -476,6 +477,7 @@ class TokenProcessor:
                     first_token_time=time.time() - task.inference_start_time,
                     time_in_queue=task.schedule_start_time - task.preprocess_end_time,
                     preprocess_cost_time=task.preprocess_end_time - task.preprocess_start_time,
+                    request_start_time=task.arrival_time,
                 )
 
                 self._record_first_token_metrics(task, current_time)

test/ci_use/EB_Lite/test_EB_Lite_serving.py

@@ -696,3 +696,25 @@ def test_non_streaming_chat_completion_disable_chat_template(openai_client, caps
     assert hasattr(disabled_response, "choices")
     assert len(disabled_response.choices) > 0
     assert enabled_response.choices[0].message.content == disabled_response.choices[0].message.content
+
+
+def test_non_streaming_min_max_token_equals_one(openai_client, capsys):
+    """
+    Test chat/completion when min_tokens equals max_tokens equals 1.
+    Verify it returns exactly one token.
+    """
+    # Test non-streaming chat
+    response = openai_client.chat.completions.create(
+        model="default",
+        messages=[{"role": "user", "content": "Hello"}],
+        max_tokens=1,
+        temperature=0.0,
+        stream=False,
+    )
+    assert hasattr(response, "choices")
+    assert len(response.choices) > 0
+    assert hasattr(response.choices[0], "message")
+    assert hasattr(response.choices[0].message, "content")
+    # Verify usage shows exactly 1 completion token
+    assert hasattr(response, "usage")
+    assert response.usage.completion_tokens == 1