Merge 61e7ce3c4b into f814945d4c

litellm/constants.py

@@ -806,6 +806,7 @@ openai_compatible_endpoints: Final[list] = [
     "https://api.meta.ai/v1",
     "https://api.cognition.ai/v1",
     "https://api.scx.ai/v1",
+    "https://api.upstage.ai/v1",
 ]
 
 
@@ -875,6 +876,7 @@ openai_compatible_providers: Final[list] = [
     "meta",  # Meta Model API (Muse Spark) - JSON-configured provider
     "cognition",
     "scx-ai",
+    "upstage",
 ]
 openai_text_completion_compatible_providers: Final[list] = [  # providers that support `/v1/completions`
     "together_ai",

litellm/llms/openai_like/providers.json

@@ -200,5 +200,10 @@
       "temperature_max": 1.99
     },
     "supported_endpoints": ["/v1/chat/completions"]
+  },
+  "upstage": {
+    "base_url": "https://api.upstage.ai/v1",
+    "api_key_env": "UPSTAGE_API_KEY",
+    "api_base_env": "UPSTAGE_API_BASE"
   }
 }

litellm/provider_endpoints_support_backup.json

@@ -2172,6 +2172,23 @@
         "interactions": true
       }
     },
+    "upstage": {
+      "display_name": "Upstage (`upstage`)",
+      "url": "https://docs.litellm.ai/docs/providers/upstage",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": false,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false
+      }
+    },
     "v0": {
       "display_name": "V0 (`v0`)",
       "url": "https://docs.litellm.ai/docs/providers/v0",

litellm/types/utils.py

@@ -3847,6 +3847,7 @@ class LlmProviders(str, Enum):
     PINSTRIPES = "pinstripes"
     COGNITION = "cognition"
     SCX_AI = "scx-ai"
+    UPSTAGE = "upstage"
     DARKBLOOM = "darkbloom"
     META = "meta"
     LITELLM_AGENT = "litellm_agent"

provider_endpoints_support.json

@@ -2483,6 +2483,23 @@
         "interactions": true
       }
     },
+    "upstage": {
+      "display_name": "Upstage (`upstage`)",
+      "url": "https://docs.litellm.ai/docs/providers/upstage",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": false,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false
+      }
+    },
     "v0": {
       "display_name": "V0 (`v0`)",
       "url": "https://docs.litellm.ai/docs/providers/v0",

tests/test_litellm/llms/openai_like/test_upstage_provider.py

@@ -0,0 +1,223 @@
+"""
+Tests for the Upstage (Solar) provider configuration and integration.
+"""
+
+import json
+from pathlib import Path
+
+import httpx
+import pytest
+
+import litellm
+from litellm import completion
+
+CHAT_COMPLETION_RESPONSE = {
+    "id": "chatcmpl-upstage-test",
+    "object": "chat.completion",
+    "created": 1787686209,
+    "model": "solar-pro4-260806",
+    "choices": [
+        {
+            "index": 0,
+            "message": {"role": "assistant", "content": "Hello from Solar"},
+            "finish_reason": "stop",
+        }
+    ],
+    "usage": {"prompt_tokens": 11, "completion_tokens": 24, "total_tokens": 35},
+}
+
+STREAMING_RESPONSE = (
+    'data: {"id":"chatcmpl-upstage-stream","object":"chat.completion.chunk","created":1787686266,'
+    '"model":"solar-pro4-260806","choices":[{"index":0,"delta":{"role":"assistant","content":"Hello"},'
+    '"finish_reason":null}]}\n\n'
+    'data: {"id":"chatcmpl-upstage-stream","object":"chat.completion.chunk","created":1787686266,'
+    '"model":"solar-pro4-260806","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}\n\n'
+    "data: [DONE]\n\n"
+)
+
+
+class TestUpstageProviderIdentity:
+    def test_upstage_is_a_registered_provider(self):
+        from litellm import LlmProviders
+
+        assert LlmProviders.UPSTAGE.value == "upstage"
+        assert "upstage" in litellm.provider_list
+
+    def test_upstage_json_config(self):
+        from litellm.llms.openai_like.json_loader import JSONProviderRegistry
+
+        upstage = JSONProviderRegistry.get("upstage")
+        assert upstage is not None
+        assert upstage.base_url == "https://api.upstage.ai/v1"
+        assert upstage.api_key_env == "UPSTAGE_API_KEY"
+        assert upstage.api_base_env == "UPSTAGE_API_BASE"
+
+    def test_upstage_in_openai_compatible_providers(self):
+        from litellm.constants import openai_compatible_providers
+
+        assert "upstage" in openai_compatible_providers
+
+    def test_prefixed_model_resolves_to_upstage_not_openai(self):
+        from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+        model, provider, _, api_base = get_llm_provider(
+            model="upstage/solar-pro4",
+            custom_llm_provider=None,
+            api_base=None,
+            api_key=None,
+        )
+
+        assert model == "solar-pro4"
+        assert provider == "upstage"
+        assert api_base == "https://api.upstage.ai/v1"
+
+    def test_explicit_api_base_and_key_win(self):
+        from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+        _, provider, api_key, api_base = get_llm_provider(
+            model="upstage/solar-pro4",
+            custom_llm_provider=None,
+            api_base="https://upstage.internal.example/v1",
+            api_key="sk-test",
+        )
+
+        assert provider == "upstage"
+        assert api_base == "https://upstage.internal.example/v1"
+        assert api_key == "sk-test"
+
+    def test_api_base_autodetects_upstage(self, monkeypatch: pytest.MonkeyPatch):
+        from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+        monkeypatch.setenv("UPSTAGE_API_KEY", "up-env-key")
+
+        _, provider, api_key, api_base = get_llm_provider(
+            model="solar-pro4",
+            custom_llm_provider=None,
+            api_base="https://api.upstage.ai/v1",
+            api_key=None,
+        )
+
+        assert provider == "upstage"
+        assert api_base == "https://api.upstage.ai/v1"
+        assert api_key == "up-env-key"
+
+    def test_autodetected_api_base_keeps_the_caller_api_key(self, monkeypatch: pytest.MonkeyPatch):
+        from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+        monkeypatch.setenv("UPSTAGE_API_KEY", "up-env-key")
+
+        _, provider, api_key, _ = get_llm_provider(
+            model="solar-pro4",
+            custom_llm_provider=None,
+            api_base="https://api.upstage.ai/v1",
+            api_key="up-caller-key",
+        )
+
+        assert provider == "upstage"
+        assert api_key == "up-caller-key"
+
+    def test_env_api_key_is_read_from_upstage_variable(self, monkeypatch: pytest.MonkeyPatch):
+        from litellm.llms.openai_like.dynamic_config import create_config_class
+        from litellm.llms.openai_like.json_loader import JSONProviderRegistry
+
+        monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+        monkeypatch.setenv("UPSTAGE_API_KEY", "up-env-key")
+
+        provider = JSONProviderRegistry.get("upstage")
+        assert provider is not None
+
+        api_base, api_key = create_config_class(provider)()._get_openai_compatible_provider_info(None, None)
+        assert api_base == "https://api.upstage.ai/v1"
+        assert api_key == "up-env-key"
+
+
+class TestUpstageRequestShape:
+    """A wrong host, a wrong env var, or a leaked `upstage/` prefix fails here and nowhere else."""
+
+    @pytest.mark.respx()
+    def test_chat_completion_targets_the_upstage_endpoint(self, respx_mock, monkeypatch: pytest.MonkeyPatch):
+        monkeypatch.setenv("UPSTAGE_API_KEY", "up-env-key")
+
+        respx_mock.post("https://api.upstage.ai/v1/chat/completions").respond(
+            json=CHAT_COMPLETION_RESPONSE, status_code=200
+        )
+
+        response = completion(
+            model="upstage/solar-pro4",
+            messages=[{"role": "user", "content": "Hello!"}],
+        )
+
+        assert len(respx_mock.calls) == 1
+        request = respx_mock.calls[0].request
+
+        assert str(request.url) == "https://api.upstage.ai/v1/chat/completions"
+        assert request.headers["Authorization"] == "Bearer up-env-key"
+
+        body = json.loads(request.content)
+        assert body["model"] == "solar-pro4"
+        assert body["messages"] == [{"role": "user", "content": "Hello!"}]
+
+        assert response.choices[0].message.content == "Hello from Solar"
+
+    @pytest.mark.respx()
+    def test_streaming_reaches_upstage_and_reassembles(self, respx_mock, monkeypatch: pytest.MonkeyPatch):
+        monkeypatch.setenv("UPSTAGE_API_KEY", "up-env-key")
+
+        respx_mock.post("https://api.upstage.ai/v1/chat/completions").mock(
+            return_value=httpx.Response(
+                200,
+                headers={"Content-Type": "text/event-stream"},
+                content=STREAMING_RESPONSE.encode(),
+            )
+        )
+
+        chunks = list(
+            completion(
+                model="upstage/solar-pro4",
+                messages=[{"role": "user", "content": "Hello!"}],
+                stream=True,
+            )
+        )
+
+        request = respx_mock.calls[0].request
+        assert str(request.url) == "https://api.upstage.ai/v1/chat/completions"
+        assert json.loads(request.content)["stream"] is True
+
+        assert "".join(chunk.choices[0].delta.content or "" for chunk in chunks) == "Hello"
+
+
+class TestUpstageParamHandling:
+    def test_max_completion_tokens_is_not_rewritten(self):
+        """Upstage takes `max_completion_tokens` natively, so the usual rewrite to `max_tokens` would break it."""
+        from litellm.llms.openai_like.dynamic_config import create_config_class
+        from litellm.llms.openai_like.json_loader import JSONProviderRegistry
+
+        provider = JSONProviderRegistry.get("upstage")
+        assert provider is not None
+
+        optional_params = create_config_class(provider)().map_openai_params(
+            non_default_params={"max_completion_tokens": 256},
+            optional_params={},
+            model="solar-pro4",
+            drop_params=False,
+        )
+
+        assert optional_params["max_completion_tokens"] == 256
+        assert "max_tokens" not in optional_params
+
+
+class TestUpstageEndpointSupportMatrix:
+    @pytest.mark.parametrize(
+        "matrix_path",
+        [
+            Path(__file__).parents[4] / "provider_endpoints_support.json",
+            Path(litellm.__file__).parent / "provider_endpoints_support_backup.json",
+        ],
+    )
+    def test_matrix_declares_chat_completions_only(self, matrix_path: Path):
+        endpoints = json.loads(matrix_path.read_text())["providers"]["upstage"]["endpoints"]
+
+        assert endpoints["chat_completions"] is True
+        assert endpoints["responses"] is False
+        assert endpoints["messages"] is False
+        assert endpoints["embeddings"] is False