From c82e486fdbfc57378c972d4a433fc657714dbd83 Mon Sep 17 00:00:00 2001
From: "Grisman, Jorge" <jorge.grisman@grainger.com>
Date: Tue, 17 Feb 2026 09:24:22 -0600
Subject: [PATCH] test(server): add multi-image and no-image vision API tests

Add three new test cases to test_vision_api.py that address the TODO
for testing with multiple images and no images:

- test_vision_chat_completion_multiple_images: verifies the server
  handles multiple image_url content parts in a single request
- test_vision_chat_completion_no_image: verifies text-only messages
  work correctly on a multimodal model
- test_vision_chat_completion_no_image_content_parts: verifies
  content parts with only text type (no image_url) work correctly

The audio test TODO is narrowed to note it needs a model with audio
input support, which the current tinygemma3 test model lacks.
---
 tools/server/tests/unit/test_vision_api.py | 65 +++++++++++++++++++++-
 1 file changed, 64 insertions(+), 1 deletion(-)

diff --git a/tools/server/tests/unit/test_vision_api.py b/tools/server/tests/unit/test_vision_api.py
index 9408116d1c..29065acdd4 100644
--- a/tools/server/tests/unit/test_vision_api.py
+++ b/tools/server/tests/unit/test_vision_api.py
@@ -70,7 +70,7 @@ def test_v1_models_supports_multimodal_capability():
         ("What is this:\n", "malformed",              False, None),
         ("What is this:\n", "https://google.com/404", False, None), # non-existent image
         ("What is this:\n", "https://ggml.ai",        False, None), # non-image data
-        # TODO @ngxson : test with multiple images, no images and with audio
+        # TODO @ngxson : test with audio (needs a model that supports audio input)
     ]
 )
 def test_vision_chat_completion(prompt, image_url, success, re_content):
@@ -97,6 +97,69 @@ def test_vision_chat_completion(prompt, image_url, success, re_content):
         assert res.status_code != 200
 
 
+def test_vision_chat_completion_multiple_images():
+    """Test sending multiple images in a single chat completion request."""
+    global server
+    server.n_ctx = 2048
+    server.n_slots = 1
+    server.start()
+    res = server.make_request("POST", "/chat/completions", data={
+        "temperature": 0.0,
+        "top_k": 1,
+        "messages": [
+            {"role": "user", "content": [
+                {"type": "text", "text": "What are these:\n"},
+                {"type": "image_url", "image_url": {
+                    "url": get_img_url("IMG_URL_0"),
+                }},
+                {"type": "image_url", "image_url": {
+                    "url": get_img_url("IMG_URL_1"),
+                }},
+            ]},
+        ],
+    })
+    assert res.status_code == 200
+    choice = res.body["choices"][0]
+    assert "assistant" == choice["message"]["role"]
+    assert len(choice["message"]["content"]) > 0
+
+
+def test_vision_chat_completion_no_image():
+    """Test sending a text-only message to a multimodal model (no images)."""
+    global server
+    server.start()
+    res = server.make_request("POST", "/chat/completions", data={
+        "temperature": 0.0,
+        "top_k": 1,
+        "messages": [
+            {"role": "user", "content": "Hello, how are you?"},
+        ],
+    })
+    assert res.status_code == 200
+    choice = res.body["choices"][0]
+    assert "assistant" == choice["message"]["role"]
+    assert len(choice["message"]["content"]) > 0
+
+
+def test_vision_chat_completion_no_image_content_parts():
+    """Test sending content parts with only text (no image_url parts)."""
+    global server
+    server.start()
+    res = server.make_request("POST", "/chat/completions", data={
+        "temperature": 0.0,
+        "top_k": 1,
+        "messages": [
+            {"role": "user", "content": [
+                {"type": "text", "text": "Hello, how are you?"},
+            ]},
+        ],
+    })
+    assert res.status_code == 200
+    choice = res.body["choices"][0]
+    assert "assistant" == choice["message"]["role"]
+    assert len(choice["message"]["content"]) > 0
+
+
 @pytest.mark.parametrize(
     "prompt, image_data, success, re_content",
     [