From 156d97b219d9fa034192e5a1e5308c4add2a4aa5 Mon Sep 17 00:00:00 2001
From: Harish <harishrsk11@gmail.com>
Date: Wed, 7 Jan 2026 15:49:02 -0800
Subject: [PATCH] Fix KeyError when logprobs=false in completions endpoint
 (#16095)

---
 .../entrypoints/openai/serving_completions.py | 16 ++++++------
 .../basic/test_serving_completions.py         | 25 +++++++++++++++++++
 2 files changed, 34 insertions(+), 7 deletions(-)

diff --git a/python/sglang/srt/entrypoints/openai/serving_completions.py b/python/sglang/srt/entrypoints/openai/serving_completions.py
index c6bd4bc9f..9fbfb8841 100644
--- a/python/sglang/srt/entrypoints/openai/serving_completions.py
+++ b/python/sglang/srt/entrypoints/openai/serving_completions.py
@@ -242,9 +242,9 @@ class OpenAIServingCompletion(OpenAIServingBase):
                         output_token_logprobs=content["meta_info"][
                             "output_token_logprobs"
                         ][n_prev_token:],
-                        output_top_logprobs=content["meta_info"]["output_top_logprobs"][
-                            n_prev_token:
-                        ],
+                        output_top_logprobs=content["meta_info"].get(
+                            "output_top_logprobs", []
+                        )[n_prev_token:],
                     )
                     n_prev_tokens[index] = len(
                         content["meta_info"]["output_token_logprobs"]
@@ -398,10 +398,12 @@ class OpenAIServingCompletion(OpenAIServingBase):
                 logprobs = to_openai_style_logprobs(
                     input_token_logprobs=input_token_logprobs,
                     input_top_logprobs=input_top_logprobs,
-                    output_token_logprobs=ret_item["meta_info"][
-                        "output_token_logprobs"
-                    ],
-                    output_top_logprobs=ret_item["meta_info"]["output_top_logprobs"],
+                    output_token_logprobs=ret_item["meta_info"].get(
+                        "output_token_logprobs", []
+                    ),
+                    output_top_logprobs=ret_item["meta_info"].get(
+                        "output_top_logprobs", []
+                    ),
                 )
 
             # Handle hidden states
diff --git a/test/registered/openai_server/basic/test_serving_completions.py b/test/registered/openai_server/basic/test_serving_completions.py
index ea58ec1c7..504ee2358 100644
--- a/test/registered/openai_server/basic/test_serving_completions.py
+++ b/test/registered/openai_server/basic/test_serving_completions.py
@@ -156,6 +156,31 @@ class ServingCompletionTestCase(unittest.TestCase):
         # (but might have json_schema from the legacy json_schema field)
         self.assertIsNone(sampling_params.get("structural_tag"))
 
+    def test_logprobs_false_non_streaming(self):
+        """Test that logprobs=False doesn't cause KeyError in non-streaming response."""
+        req = CompletionRequest(
+            model="x", prompt="Hello", max_tokens=10, logprobs=False
+        )
+
+        mock_ret = [
+            {
+                "text": " world",
+                "meta_info": {
+                    "id": "test-id",
+                    "prompt_tokens": 1,
+                    "completion_tokens": 2,
+                    "finish_reason": {"type": "stop"},
+                    "weight_version": "v1",
+                },
+            }
+        ]
+
+        response = self.sc._build_completion_response(req, mock_ret, 1234567890)
+
+        self.assertEqual(len(response.choices), 1)
+        self.assertEqual(response.choices[0].text, " world")
+        self.assertEqual(len(response.choices[0].logprobs.top_logprobs), 0)
+
 
 if __name__ == "__main__":
     unittest.main(verbosity=2)