feat: Add support for custom Anthropic-compatible API endpoints

This commit adds support for custom Anthropic-compatible API endpoints
that don't follow the standard /v1/messages or /v1/complete path convention.

## Changes
- Added LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX environment variable
- When set to true, prevents automatic appending of /v1/messages (for anthropic)
- When set to true, prevents automatic appending of /v1/complete (for anthropic_text)
- Added debug logging to indicate when suffix is being skipped
- Maintained full backward compatibility - existing deployments are unaffected

litellm/main.py

@@ -2169,8 +2169,18 @@ def completion(  # type: ignore # noqa: PLR0915
                 or "https://api.anthropic.com/v1/complete"
             )
 
-            if api_base is not None and not api_base.endswith("/v1/complete"):
+            # Check if we should disable automatic URL suffix appending
+            disable_url_suffix = get_secret_bool("LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX")
+            if (
+                api_base is not None
+                and not disable_url_suffix
+                and not api_base.endswith("/v1/complete")
+            ):
                 api_base += "/v1/complete"
+            elif disable_url_suffix:
+                verbose_logger.debug(
+                    "LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX is set, skipping /v1/complete suffix"
+                )
 
             response = base_llm_http_handler.completion(
                 model=model,
@@ -2206,8 +2216,18 @@ def completion(  # type: ignore # noqa: PLR0915
                 or "https://api.anthropic.com/v1/messages"
             )
 
-            if api_base is not None and not api_base.endswith("/v1/messages"):
+            # Check if we should disable automatic URL suffix appending
+            disable_url_suffix = get_secret_bool("LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX")
+            if (
+                api_base is not None
+                and not disable_url_suffix
+                and not api_base.endswith("/v1/messages")
+            ):
                 api_base += "/v1/messages"
+            elif disable_url_suffix:
+                verbose_logger.debug(
+                    "LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX is set, skipping /v1/messages suffix"
+                )
 
             response = anthropic_chat_completions.completion(
                 model=model,

tests/test_litellm/test_main.py

@@ -1121,4 +1121,124 @@ async def test_retrying() -> None:
             model="gpt-4o-mini",
             messages=[{"role": "user", "content": "Hello"}],
         )
-    assert mock_request.call_count >= 10, "Expected retrying to be used"
+
+
+def test_anthropic_disable_url_suffix_env_var():
+    """Test that LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX prevents /v1/messages suffix."""
+    from unittest.mock import patch, MagicMock
+    import os
+    from litellm import completion
+    from litellm.llms.anthropic.chat.handler import AnthropicLLM
+
+    # Test with environment variable disabled (default behavior)
+    with patch.dict(os.environ, {"ANTHROPIC_API_BASE": "https://api.example.com"}):
+        with patch.object(AnthropicLLM, "__init__", return_value=None) as mock_init:
+            with patch.object(AnthropicLLM, "completion") as mock_completion:
+                mock_completion.return_value = MagicMock()
+                
+                # Mock the initialization to capture api_base
+                actual_api_base = None
+                def capture_init(self, **kwargs):
+                    nonlocal actual_api_base
+                    actual_api_base = kwargs.get("api_base")
+                
+                with patch("litellm.main.anthropic_chat_completions") as mock_anthropic:
+                    def capture_completion(**kwargs):
+                        nonlocal actual_api_base
+                        actual_api_base = kwargs.get("api_base")
+                        return MagicMock()
+                    
+                    mock_anthropic.completion = capture_completion
+                    
+                    # This should append /v1/messages
+                    completion(
+                        model="anthropic/claude-3-sonnet",
+                        messages=[{"role": "user", "content": "test"}],
+                        api_key="test-key"
+                    )
+                    
+                    # Verify the api_base has /v1/messages appended
+                    assert actual_api_base.endswith("/v1/messages")
+                    assert actual_api_base == "https://api.example.com/v1/messages"
+
+    # Test with environment variable enabled
+    with patch.dict(os.environ, {
+        "ANTHROPIC_API_BASE": "https://api.example.com/custom/path",
+        "LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX": "true"
+    }):
+        actual_api_base = None
+        
+        with patch("litellm.main.anthropic_chat_completions") as mock_anthropic:
+            def capture_completion(**kwargs):
+                nonlocal actual_api_base
+                actual_api_base = kwargs.get("api_base")
+                return MagicMock()
+            
+            mock_anthropic.completion = capture_completion
+            
+            # This should NOT append /v1/messages
+            completion(
+                model="anthropic/claude-3-sonnet",
+                messages=[{"role": "user", "content": "test"}],
+                api_key="test-key"
+            )
+            
+            # Verify the api_base does not have /v1/messages appended
+            assert actual_api_base == "https://api.example.com/custom/path"
+            assert not actual_api_base.endswith("/v1/messages")
+
+
+def test_anthropic_text_disable_url_suffix_env_var():
+    """Test that LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX prevents /v1/complete suffix for anthropic_text."""
+    from unittest.mock import patch, MagicMock
+    import os
+    from litellm import completion
+
+    # Test with environment variable disabled (default behavior)
+    with patch.dict(os.environ, {"ANTHROPIC_API_BASE": "https://api.example.com"}):
+        actual_api_base = None
+        
+        with patch("litellm.main.base_llm_http_handler") as mock_handler:
+            def capture_completion(**kwargs):
+                nonlocal actual_api_base
+                actual_api_base = kwargs.get("api_base")
+                return MagicMock()
+            
+            mock_handler.completion = capture_completion
+            
+            # This should append /v1/complete
+            completion(
+                model="anthropic_text/claude-instant-1",
+                messages=[{"role": "user", "content": "test"}],
+                api_key="test-key"
+            )
+            
+            # Verify the api_base has /v1/complete appended
+            assert actual_api_base.endswith("/v1/complete")
+            assert actual_api_base == "https://api.example.com/v1/complete"
+
+    # Test with environment variable enabled
+    with patch.dict(os.environ, {
+        "ANTHROPIC_API_BASE": "https://api.example.com/custom/complete",
+        "LITELLM_ANTHROPIC_DISABLE_URL_SUFFIX": "true"
+    }):
+        actual_api_base = None
+        
+        with patch("litellm.main.base_llm_http_handler") as mock_handler:
+            def capture_completion(**kwargs):
+                nonlocal actual_api_base
+                actual_api_base = kwargs.get("api_base")
+                return MagicMock()
+            
+            mock_handler.completion = capture_completion
+            
+            # This should NOT append /v1/complete
+            completion(
+                model="anthropic_text/claude-instant-1",
+                messages=[{"role": "user", "content": "test"}],
+                api_key="test-key"
+            )
+            
+            # Verify the api_base does not have /v1/complete appended
+            assert actual_api_base == "https://api.example.com/custom/complete"
+            assert not actual_api_base.endswith("/v1/complete")