Fix PR review issues: gpt-4-0314 prompt caching, case-insensitive data URL check, test I/O mocking

- Remove incorrect supports_prompt_caching from gpt-4-0314 (predates the feature)
- Make data-URL detection case-insensitive in Gemini tool call result conversion
- Mock show_banner/generate_feedback_box in max_budget tests to prevent real I/O

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

litellm/litellm_core_utils/prompt_templates/factory.py

@@ -1506,7 +1506,7 @@ def convert_to_gemini_tool_call_result(  # noqa: PLR0915
             # Detect data-URL images (e.g. from Anthropic tool_result with a single image block
             # that was serialised as a plain string by translate_anthropic_messages_to_openai)
             # and promote them to inline_data so Gemini receives actual image bytes.
-            if content_str.startswith("data:") and ";base64," in content_str:
+            if content_str[:5].lower() == "data:" and ";base64," in content_str:
                 try:
                     mime_rest = content_str[5:].split(";base64,", 1)
                     if len(mime_rest) == 2 and mime_rest[0].startswith("image/"):

litellm/model_prices_and_context_window_backup.json

@@ -6152,7 +6152,8 @@
         "max_query_tokens": 4096,
         "max_tokens": 32768,
         "mode": "rerank",
-        "output_cost_per_token": 0.0
+        "output_cost_per_token": 0.0,
+        "source": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/introducing-cohere-rerank-4-0-in-microsoft-foundry/4477076"
     },
     "azure_ai/cohere-rerank-v4.0-fast": {
         "input_cost_per_query": 0.002,
@@ -6163,7 +6164,8 @@
         "max_query_tokens": 4096,
         "max_tokens": 32768,
         "mode": "rerank",
-        "output_cost_per_token": 0.0
+        "output_cost_per_token": 0.0,
+        "source": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/introducing-cohere-rerank-4-0-in-microsoft-foundry/4477076"
     },
     "azure_ai/deepseek-v3.2": {
         "input_cost_per_token": 5.8e-07,
@@ -6173,6 +6175,7 @@
         "max_tokens": 163840,
         "mode": "chat",
         "output_cost_per_token": 1.68e-06,
+        "source": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/introducing-deepseek-v3-2-and-deepseek-v3-2-speciale-in-microsoft-foundry/4477549",
         "supports_assistant_prefill": true,
         "supports_function_calling": true,
         "supports_prompt_caching": true,
@@ -6187,6 +6190,7 @@
         "max_tokens": 163840,
         "mode": "chat",
         "output_cost_per_token": 1.68e-06,
+        "source": "https://techcommunity.microsoft.com/blog/azure-ai-foundry-blog/introducing-deepseek-v3-2-and-deepseek-v3-2-speciale-in-microsoft-foundry/4477549",
         "supports_assistant_prefill": true,
         "supports_function_calling": true,
         "supports_prompt_caching": true,
@@ -16936,6 +16940,18 @@
         "supports_system_messages": true,
         "supports_tool_choice": true
     },
+    "gpt-4-0314": {
+        "deprecation_date": "2026-03-26",
+        "input_cost_per_token": 3e-05,
+        "litellm_provider": "openai",
+        "max_input_tokens": 8192,
+        "max_output_tokens": 4096,
+        "max_tokens": 4096,
+        "mode": "chat",
+        "output_cost_per_token": 6e-05,
+        "supports_system_messages": true,
+        "supports_tool_choice": true
+    },
     "gpt-4-0613": {
         "deprecation_date": "2025-06-06",
         "input_cost_per_token": 3e-05,
@@ -30506,7 +30522,7 @@
         "output_cost_per_token": 5.4e-06,
         "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#partner-models",
         "supported_regions": [
-            "us-west2"
+            "us-central1"
         ],
         "supports_assistant_prefill": true,
         "supports_function_calling": true,
@@ -30526,7 +30542,7 @@
         "output_cost_per_token_batches": 8.4e-07,
         "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#partner-models",
         "supported_regions": [
-            "us-west2"
+            "global"
         ],
         "supports_assistant_prefill": true,
         "supports_function_calling": true,
@@ -30543,6 +30559,9 @@
         "mode": "chat",
         "output_cost_per_token": 5.4e-06,
         "source": "https://cloud.google.com/vertex-ai/generative-ai/pricing#partner-models",
+        "supported_regions": [
+            "us-central1"
+        ],
         "supports_assistant_prefill": true,
         "supports_function_calling": true,
         "supports_prompt_caching": true,
@@ -31167,7 +31186,10 @@
         "input_cost_per_token": 3e-07,
         "output_cost_per_token": 1.2e-06,
         "ocr_cost_per_page": 0.0003,
-        "source": "https://cloud.google.com/vertex-ai/pricing"
+        "source": "https://cloud.google.com/vertex-ai/pricing",
+        "supported_regions": [
+            "us-central1"
+        ]
     },
     "vertex_ai/openai/gpt-oss-120b-maas": {
         "input_cost_per_token": 1.5e-07,

model_prices_and_context_window.json

@@ -16949,7 +16949,6 @@
         "max_tokens": 4096,
         "mode": "chat",
         "output_cost_per_token": 6e-05,
-        "supports_prompt_caching": true,
         "supports_system_messages": true,
         "supports_tool_choice": true
     },

tests/test_litellm/proxy/test_max_budget_env_var.py

@@ -4,10 +4,11 @@ converted to float.
 GitHub Issue: #23843
 """
 
+from unittest.mock import patch
+
 import pytest
 
 import litellm
-from litellm.proxy.proxy_server import initialize
 
 
 @pytest.mark.asyncio
@@ -17,22 +18,32 @@ async def test_max_budget_string_converted_to_float():
     string. initialize() should convert it to float so the comparison
     `litellm.max_budget > 0` doesn't raise TypeError.
     """
-    original = litellm.max_budget
-    try:
-        await initialize(max_budget="100.5")
-        assert isinstance(litellm.max_budget, float)
-        assert litellm.max_budget == 100.5
-    finally:
-        litellm.max_budget = original
+    with patch("litellm.proxy.common_utils.banner.show_banner"), patch(
+        "litellm.proxy.proxy_server.generate_feedback_box"
+    ):
+        from litellm.proxy.proxy_server import initialize
+
+        original = litellm.max_budget
+        try:
+            await initialize(max_budget="100.5")
+            assert isinstance(litellm.max_budget, float)
+            assert litellm.max_budget == 100.5
+        finally:
+            litellm.max_budget = original
 
 
 @pytest.mark.asyncio
 async def test_max_budget_float_stays_float():
     """max_budget as float should still work."""
-    original = litellm.max_budget
-    try:
-        await initialize(max_budget=200.0)
-        assert isinstance(litellm.max_budget, float)
-        assert litellm.max_budget == 200.0
-    finally:
-        litellm.max_budget = original
+    with patch("litellm.proxy.common_utils.banner.show_banner"), patch(
+        "litellm.proxy.proxy_server.generate_feedback_box"
+    ):
+        from litellm.proxy.proxy_server import initialize
+
+        original = litellm.max_budget
+        try:
+            await initialize(max_budget=200.0)
+            assert isinstance(litellm.max_budget, float)
+            assert litellm.max_budget == 200.0
+        finally:
+            litellm.max_budget = original