perf: optimal option when gemma4 model is used

Author: renesansz

src/argus/captioner.py

@@ -234,13 +234,19 @@ def caption_frame(image_path: Path, *, model: str, ollama_host: str) -> dict:
         "messages": [
             {
                 "role": "user",
-                "content": FRAME_CAPTION_PROMPT,
                 "images": [encoded],
+                "content": FRAME_CAPTION_PROMPT,
             }
         ],
         "format": "json",
         "stream": False,
     }
+    if base_model_name(model) == "gemma4":
+        payload["options"] = {
+            "temperature": 1.0,
+            "top_p": 0.95,
+            "top_k": 64,
+        }
 
     try:
         response = ollama_chat(payload, ollama_host=ollama_host)
@@ -300,6 +306,11 @@ def summarize_captions(
                 ),
             },
         ],
+        "options": {
+            "temperature": 1.0,
+            "top_p": 0.95,
+            "top_k": 64
+        },
         "format": "json",
         "stream": False,
     }

tests/test_scanner.py

@@ -8,6 +8,7 @@
 from unittest.mock import MagicMock, patch
 
 from argus.captioner import (
+    caption_frame,
     caption_output_items,
     match_ollama_model,
     normalize_clip_title,
@@ -159,6 +160,63 @@ def test_summarize_captions_rejects_missing_title(self, ollama_chat_mock) -> Non
         self.assertEqual(result["status"], "error")
         self.assertIn("required fields", result["reason"])
 
+    @patch("argus.captioner.ollama_chat")
+    def test_caption_frame_includes_options_for_gemma4(self, ollama_chat_mock) -> None:
+        ollama_chat_mock.return_value = {
+            "message": {
+                "content": json.dumps(
+                    {
+                        "short_caption": "A test frame.",
+                        "tags": ["tag"] * 40,
+                        "visible_text": [],
+                    }
+                )
+            }
+        }
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            image_path = Path(temp_dir) / "frame.jpg"
+            image_path.write_bytes(b"\xff\xd8\xff")
+            result = caption_frame(
+                image_path,
+                model="gemma4:latest",
+                ollama_host="http://localhost:11434",
+            )
+
+        self.assertEqual(result["status"], "ok")
+        payload = ollama_chat_mock.call_args[0][0]
+        self.assertIn("options", payload)
+        self.assertEqual(payload["options"]["temperature"], 1.0)
+        self.assertEqual(payload["options"]["top_p"], 0.95)
+        self.assertEqual(payload["options"]["top_k"], 64)
+
+    @patch("argus.captioner.ollama_chat")
+    def test_caption_frame_omits_options_for_non_gemma4(self, ollama_chat_mock) -> None:
+        ollama_chat_mock.return_value = {
+            "message": {
+                "content": json.dumps(
+                    {
+                        "short_caption": "A test frame.",
+                        "tags": ["tag"] * 40,
+                        "visible_text": [],
+                    }
+                )
+            }
+        }
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            image_path = Path(temp_dir) / "frame.jpg"
+            image_path.write_bytes(b"\xff\xd8\xff")
+            result = caption_frame(
+                image_path,
+                model="gemma3",
+                ollama_host="http://localhost:11434",
+            )
+
+        self.assertEqual(result["status"], "ok")
+        payload = ollama_chat_mock.call_args[0][0]
+        self.assertNotIn("options", payload)
+
     def test_normalize_clip_title_truncates_to_max_length(self) -> None:
         long_title = "word " * 40
         out = normalize_clip_title(long_title, max_len=100)