Merge ed2d5d2e00 into a5e9f275ea

README.md

@@ -399,7 +399,7 @@ Set `LITELLM_PROXY_API_BASE` and `LITELLM_PROXY_API_KEY` and every model call th
 | [Vercel AI Gateway (`vercel_ai_gateway`)](https://docs.litellm.ai/docs/providers/vercel_ai_gateway) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [VLLM (`vllm`)](https://docs.litellm.ai/docs/providers/vllm) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Volcengine (`volcengine`)](https://docs.litellm.ai/docs/providers/volcano) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
-| [Voyage AI (`voyage`)](https://docs.litellm.ai/docs/providers/voyage) |  |  |  | ✅ |  |  |  |  |  |  |
+| [VoyageAI by MongoDB (`voyage`)](https://docs.litellm.ai/docs/providers/voyage) |  |  |  | ✅ |  |  |  |  |  |  |
 | [WandB Inference (`wandb`)](https://docs.litellm.ai/docs/providers/wandb_inference) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Watsonx Text (`watsonx_text`)](https://docs.litellm.ai/docs/providers/watsonx) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [xAI (`xai`)](https://docs.litellm.ai/docs/providers/xai) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |

litellm/llms/voyage/common_utils.py

@@ -0,0 +1,35 @@
+"""
+Shared helpers for the Voyage (VoyageAI by MongoDB) provider.
+"""
+
+from typing import Final
+
+from litellm.secret_managers.main import get_secret_str
+
+VOYAGE_API_BASE: Final = "https://api.voyageai.com/v1"
+MONGODB_API_BASE: Final = "https://ai.mongodb.com/v1"
+MONGODB_API_KEY_PREFIX: Final = "al-"
+
+
+def get_voyage_api_key(api_key: str | None = None) -> str | None:
+    """Resolve the key a Voyage request will authenticate with, explicit value first."""
+    return (
+        api_key
+        or get_secret_str("VOYAGE_API_KEY")
+        or get_secret_str("VOYAGE_AI_API_KEY")
+        or get_secret_str("VOYAGE_AI_TOKEN")
+    )
+
+
+def get_default_base_url(api_key: str | None = None) -> str:
+    """
+    Pick the host that issued the key: MongoDB-issued keys (``al-`` prefix) are only
+    valid on ai.mongodb.com, every other key on api.voyageai.com.
+
+    Mirrors ``voyageai.util.get_default_base_url`` in the official SDK:
+    https://github.com/voyage-ai/voyageai-python/blob/main/voyageai/util.py
+    """
+    resolved: Final = get_voyage_api_key(api_key)
+    if resolved is not None and resolved.startswith(MONGODB_API_KEY_PREFIX):
+        return MONGODB_API_BASE
+    return VOYAGE_API_BASE

litellm/llms/voyage/embedding/transformation.py

@@ -5,7 +5,7 @@ import httpx
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
-from litellm.secret_managers.main import get_secret_str
+from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
 from litellm.types.llms.openai import AllEmbeddingInputValues, AllMessageValues
 from litellm.types.utils import EmbeddingResponse, Usage
 
@@ -49,7 +49,7 @@ class VoyageEmbeddingConfig(BaseEmbeddingConfig):
             if not api_base.endswith("/embeddings"):
                 api_base = f"{api_base}/embeddings"
             return api_base
-        return "https://api.voyageai.com/v1/embeddings"
+        return f"{get_default_base_url(api_key)}/embeddings"
 
     def get_supported_openai_params(self, model: str) -> list:
         return [
@@ -85,14 +85,8 @@ class VoyageEmbeddingConfig(BaseEmbeddingConfig):
         api_key: str | None = None,
         api_base: str | None = None,
     ) -> dict:
-        if api_key is None:
-            api_key = (
-                get_secret_str("VOYAGE_API_KEY")
-                or get_secret_str("VOYAGE_AI_API_KEY")
-                or get_secret_str("VOYAGE_AI_TOKEN")
-            )
         return {
-            "Authorization": f"Bearer {api_key}",
+            "Authorization": f"Bearer {get_voyage_api_key(api_key)}",
         }
 
     def transform_embedding_request(

litellm/llms/voyage/embedding/transformation_contextual.py

@@ -11,7 +11,7 @@ import httpx
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
-from litellm.secret_managers.main import get_secret_str
+from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
 from litellm.types.llms.openai import AllEmbeddingInputValues, AllMessageValues
 from litellm.types.utils import EmbeddingResponse, Usage
 
@@ -58,7 +58,7 @@ class VoyageContextualEmbeddingConfig(BaseEmbeddingConfig):
             if not api_base.endswith("/contextualizedembeddings"):
                 api_base = f"{api_base}/contextualizedembeddings"
             return api_base
-        return "https://api.voyageai.com/v1/contextualizedembeddings"
+        return f"{get_default_base_url(api_key)}/contextualizedembeddings"
 
     def get_supported_openai_params(self, model: str) -> list:  # mutable-ok: base class signature
         return ["encoding_format", "dimensions"]
@@ -91,14 +91,8 @@ class VoyageContextualEmbeddingConfig(BaseEmbeddingConfig):
         api_key: str | None = None,
         api_base: str | None = None,
     ) -> dict:
-        if api_key is None:
-            api_key = (
-                get_secret_str("VOYAGE_API_KEY")
-                or get_secret_str("VOYAGE_AI_API_KEY")
-                or get_secret_str("VOYAGE_AI_TOKEN")
-            )
         return {
-            "Authorization": f"Bearer {api_key}",
+            "Authorization": f"Bearer {get_voyage_api_key(api_key)}",
         }
 
     AUTO_CHUNK_SIZE: Final = 32000

litellm/llms/voyage/embedding/transformation_multimodal.py

@@ -13,7 +13,7 @@ import httpx
 from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
-from litellm.secret_managers.main import get_secret_str
+from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
 from litellm.types.llms.openai import AllEmbeddingInputValues, AllMessageValues
 from litellm.types.utils import EmbeddingResponse, Usage
 
@@ -58,7 +58,7 @@ class VoyageMultimodalEmbeddingConfig(BaseEmbeddingConfig):
             if not api_base.endswith("/multimodalembeddings"):
                 api_base = f"{api_base}/multimodalembeddings"
             return api_base
-        return "https://api.voyageai.com/v1/multimodalembeddings"
+        return f"{get_default_base_url(api_key)}/multimodalembeddings"
 
     def get_supported_openai_params(self, model: str) -> list:
         return ["dimensions"]
@@ -84,19 +84,14 @@ class VoyageMultimodalEmbeddingConfig(BaseEmbeddingConfig):
         api_key: str | None = None,
         api_base: str | None = None,
     ) -> dict:
-        if api_key is None:
-            api_key = (
-                get_secret_str("VOYAGE_API_KEY")
-                or get_secret_str("VOYAGE_AI_API_KEY")
-                or get_secret_str("VOYAGE_AI_TOKEN")
-            )
-        if not api_key:
+        resolved_api_key: Final = get_voyage_api_key(api_key)
+        if not resolved_api_key:
             raise ValueError(
                 "Voyage API key is required for multimodal embeddings. "
                 "Set VOYAGE_API_KEY / VOYAGE_AI_API_KEY / VOYAGE_AI_TOKEN "
                 "or pass `api_key` explicitly."
             )
-        return {"Authorization": f"Bearer {api_key}"}
+        return {"Authorization": f"Bearer {resolved_api_key}"}
 
     def _normalize_content_item(self, item: dict[str, object]) -> dict[str, object]:
         item_type: Final = item.get("type")

litellm/llms/voyage/rerank/transformation.py

@@ -12,7 +12,7 @@ import httpx
 from litellm._uuid import uuid
 from litellm.llms.base_llm.chat.transformation import LiteLLMLoggingObj
 from litellm.llms.base_llm.rerank.transformation import BaseRerankConfig
-from litellm.secret_managers.main import get_secret_str
+from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
 from litellm.types.rerank import (
     RerankBilledUnits,
     RerankResponse,
@@ -25,6 +25,16 @@ from ..embedding.transformation import VoyageError
 
 
 class VoyageRerankConfig(BaseRerankConfig):
+    """
+    ``validate_environment`` stores the credential it authenticates with so ``get_complete_url``
+    can select the host that issued it. ``ProviderConfigManager.get_provider_rerank_config``
+    builds this config per request, so that key never reaches another one.
+    """
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._api_key: str | None = None
+
     def get_supported_cohere_rerank_params(self, model: str) -> list:
         return ["query", "documents", "top_n", "return_documents"]
 
@@ -60,7 +70,7 @@ class VoyageRerankConfig(BaseRerankConfig):
         optional_params: dict | None = None,
     ) -> str:
         if api_base is None:
-            return "https://api.voyageai.com/v1/rerank"
+            return f"{get_default_base_url(self._api_key)}/rerank"
         api_base = api_base.rstrip("/")
         if not api_base.endswith("/v1/rerank"):
             if api_base.endswith("/v1"):
@@ -141,12 +151,12 @@ class VoyageRerankConfig(BaseRerankConfig):
         optional_params: dict | None = None,
         litellm_params: Mapping[str, object] | None = None,
     ) -> dict:
-        if api_key is None:
-            api_key = get_secret_str("VOYAGE_API_KEY") or get_secret_str("VOYAGE_AI_API_KEY")
-        if api_key is None:
+        resolved_api_key: Final = get_voyage_api_key(api_key)
+        if resolved_api_key is None:
             raise ValueError("Voyage AI API key is required. Set via `api_key` parameter or `VOYAGE_API_KEY` env var.")
+        self._api_key = resolved_api_key
         return {
-            "Authorization": f"Bearer {api_key}",
+            "Authorization": f"Bearer {resolved_api_key}",
             "content-type": "application/json",
         }
 

litellm/provider_endpoints_support_backup.json

@@ -2336,7 +2336,7 @@
       }
     },
     "voyage": {
-      "display_name": "Voyage AI (`voyage`)",
+      "display_name": "VoyageAI by MongoDB (`voyage`)",
       "url": "https://docs.litellm.ai/docs/providers/voyage",
       "endpoints": {
         "chat_completions": false,

litellm/proxy/public_endpoints/provider_create_fields.json

@@ -3499,7 +3499,7 @@
   },
   {
     "provider": "Voyage",
-    "provider_display_name": "Voyage AI",
+    "provider_display_name": "VoyageAI by MongoDB",
     "litellm_provider": "voyage",
     "credential_fields": [
       {

provider_endpoints_support.json

@@ -2697,7 +2697,7 @@
       }
     },
     "voyage": {
-      "display_name": "Voyage AI (`voyage`)",
+      "display_name": "VoyageAI by MongoDB (`voyage`)",
       "url": "https://docs.litellm.ai/docs/providers/voyage",
       "endpoints": {
         "chat_completions": false,

tests/unit/llms/voyage/rerank/test_voyage_rerank_transformation.py

@@ -294,11 +294,10 @@ class TestVoyageRerankTransform:
         assert "top_n" in supported_params
         assert "return_documents" in supported_params
 
-    @patch("litellm.llms.voyage.rerank.transformation.get_secret_str")
-    def test_validate_environment_missing_api_key(self, mock_get_secret_str):
+    def test_validate_environment_missing_api_key(self, monkeypatch):
         """Test that validate_environment raises error when API key is missing."""
-        # Mock get_secret_str to return None for both environment variables
-        mock_get_secret_str.return_value = None
+        for env_var in ("VOYAGE_API_KEY", "VOYAGE_AI_API_KEY", "VOYAGE_AI_TOKEN"):
+            monkeypatch.delenv(env_var, raising=False)
         with pytest.raises(ValueError, match="Voyage AI API key is required"):
             self.config.validate_environment(
                 headers={},

tests/unit/llms/voyage/test_common_utils.py

@@ -0,0 +1,164 @@
+import pytest
+
+from litellm.llms.voyage.common_utils import (
+    MONGODB_API_BASE,
+    VOYAGE_API_BASE,
+    get_default_base_url,
+    get_voyage_api_key,
+)
+from litellm.llms.voyage.embedding.transformation import VoyageEmbeddingConfig
+from litellm.llms.voyage.embedding.transformation_contextual import (
+    VoyageContextualEmbeddingConfig,
+)
+from litellm.llms.voyage.embedding.transformation_multimodal import (
+    VoyageMultimodalEmbeddingConfig,
+)
+from litellm.llms.voyage.rerank.transformation import VoyageRerankConfig
+
+VOYAGE_KEY_ENV_VARS = ("VOYAGE_API_KEY", "VOYAGE_AI_API_KEY", "VOYAGE_AI_TOKEN")
+
+
+@pytest.fixture(autouse=True)
+def clear_voyage_env(monkeypatch):
+    for env_var in VOYAGE_KEY_ENV_VARS:
+        monkeypatch.delenv(env_var, raising=False)
+
+
+def test_mongodb_key_routes_to_mongodb_host():
+    """MongoDB-issued keys carry the `al-` prefix and are only valid on ai.mongodb.com"""
+    assert get_default_base_url("al-1234567890") == MONGODB_API_BASE
+
+
+@pytest.mark.parametrize("api_key", ["pa-1234567890", "sk-1234567890", "al", "", None])
+def test_non_mongodb_key_routes_to_voyage_host(api_key):
+    assert get_default_base_url(api_key) == VOYAGE_API_BASE
+
+
+@pytest.mark.parametrize("env_var", VOYAGE_KEY_ENV_VARS)
+def test_mongodb_key_from_any_supported_env_var_routes_to_mongodb_host(monkeypatch, env_var):
+    monkeypatch.setenv(env_var, "al-from-env")
+    assert get_default_base_url() == MONGODB_API_BASE
+
+
+def test_explicit_key_wins_over_env_for_routing(monkeypatch):
+    monkeypatch.setenv("VOYAGE_API_KEY", "al-from-env")
+    assert get_default_base_url("pa-explicit") == VOYAGE_API_BASE
+
+
+@pytest.mark.parametrize(
+    "config, endpoint",
+    [
+        (VoyageEmbeddingConfig(), "embeddings"),
+        (VoyageContextualEmbeddingConfig(), "contextualizedembeddings"),
+        (VoyageMultimodalEmbeddingConfig(), "multimodalembeddings"),
+    ],
+)
+@pytest.mark.parametrize("api_key, expected_host", [("al-key", MONGODB_API_BASE), ("pa-key", VOYAGE_API_BASE)])
+def test_embedding_configs_route_by_key_prefix(config, endpoint, api_key, expected_host):
+    url = config.get_complete_url(None, api_key, "voyage-3", {}, {})
+    assert url == f"{expected_host}/{endpoint}"
+
+
+@pytest.mark.parametrize(
+    "config, endpoint",
+    [
+        (VoyageEmbeddingConfig(), "embeddings"),
+        (VoyageContextualEmbeddingConfig(), "contextualizedembeddings"),
+        (VoyageMultimodalEmbeddingConfig(), "multimodalembeddings"),
+    ],
+)
+def test_explicit_api_base_overrides_key_routing(config, endpoint):
+    url = config.get_complete_url("https://gateway.internal/v1", "al-key", "voyage-3", {}, {})
+    assert url == f"https://gateway.internal/v1/{endpoint}"
+
+
+@pytest.mark.parametrize("api_key, expected_host", [("al-key", MONGODB_API_BASE), ("pa-key", VOYAGE_API_BASE)])
+def test_rerank_routes_by_request_key_prefix(api_key, expected_host):
+    config = VoyageRerankConfig()
+    config.validate_environment({}, "rerank-2.5", api_key=api_key)
+
+    assert config.get_complete_url(None, "rerank-2.5") == f"{expected_host}/rerank"
+
+
+@pytest.mark.parametrize("api_key, expected_host", [("al-key", MONGODB_API_BASE), ("pa-key", VOYAGE_API_BASE)])
+def test_rerank_routes_by_env_key_prefix(monkeypatch, api_key, expected_host):
+    monkeypatch.setenv("VOYAGE_API_KEY", api_key)
+    config = VoyageRerankConfig()
+    config.validate_environment({}, "rerank-2.5")
+
+    assert config.get_complete_url(None, "rerank-2.5") == f"{expected_host}/rerank"
+
+
+def test_rerank_request_key_beats_env_key_for_routing(monkeypatch):
+    """A MongoDB key on the request must not be posted to the Voyage host the env key names"""
+    monkeypatch.setenv("VOYAGE_API_KEY", "pa-from-env")
+    config = VoyageRerankConfig()
+
+    headers = config.validate_environment({}, "rerank-2.5", api_key="al-on-request")
+
+    assert headers["Authorization"] == "Bearer al-on-request"
+    assert config.get_complete_url(None, "rerank-2.5") == f"{MONGODB_API_BASE}/rerank"
+
+
+def test_rerank_config_is_built_per_request_so_keys_cannot_leak(monkeypatch):
+    """get_complete_url reads a key off the instance, so each request must get its own instance"""
+    import litellm
+    from litellm.types.utils import LlmProviders
+    from litellm.utils import ProviderConfigManager
+
+    monkeypatch.delenv("VOYAGE_API_KEY", raising=False)
+    first = ProviderConfigManager.get_provider_rerank_config(
+        model="rerank-2.5", provider=LlmProviders.VOYAGE, api_base=None, present_version_params=[]
+    )
+    second = ProviderConfigManager.get_provider_rerank_config(
+        model="rerank-2.5", provider=LlmProviders.VOYAGE, api_base=None, present_version_params=[]
+    )
+    assert isinstance(first, litellm.VoyageRerankConfig) and first is not second
+
+    first.validate_environment({}, "rerank-2.5", api_key="al-first-request")
+    second.validate_environment({}, "rerank-2.5", api_key="pa-second-request")
+
+    assert first.get_complete_url(None, "rerank-2.5") == f"{MONGODB_API_BASE}/rerank"
+    assert second.get_complete_url(None, "rerank-2.5") == f"{VOYAGE_API_BASE}/rerank"
+
+
+def test_rerank_falls_back_to_env_when_validate_environment_did_not_run(monkeypatch):
+    """A caller that skips validate_environment keeps the pre-existing env-only behaviour"""
+    monkeypatch.setenv("VOYAGE_API_KEY", "al-from-env")
+
+    assert VoyageRerankConfig().get_complete_url(None, "rerank-2.5") == f"{MONGODB_API_BASE}/rerank"
+
+
+@pytest.mark.parametrize(
+    "config",
+    [VoyageEmbeddingConfig(), VoyageContextualEmbeddingConfig(), VoyageMultimodalEmbeddingConfig()],
+)
+def test_auth_header_uses_the_key_the_url_was_routed_on(monkeypatch, config):
+    """The host is picked from a key, so the Authorization header has to carry that same key"""
+    monkeypatch.setenv("VOYAGE_AI_TOKEN", "al-from-env")
+
+    headers = config.validate_environment({}, "voyage-3", [], {}, {})
+    url = config.get_complete_url(None, None, "voyage-3", {}, {})
+
+    assert headers["Authorization"] == "Bearer al-from-env"
+    assert url.startswith(MONGODB_API_BASE)
+
+
+def test_rerank_auth_header_uses_the_key_the_url_was_routed_on(monkeypatch):
+    monkeypatch.setenv("VOYAGE_AI_TOKEN", "al-from-env")
+    config = VoyageRerankConfig()
+
+    headers = config.validate_environment({}, "rerank-2.5")
+
+    assert headers["Authorization"] == "Bearer al-from-env"
+    assert config.get_complete_url(None, "rerank-2.5").startswith(MONGODB_API_BASE)
+
+
+def test_get_voyage_api_key_prefers_env_vars_in_documented_order(monkeypatch):
+    monkeypatch.setenv("VOYAGE_AI_API_KEY", "second")
+    monkeypatch.setenv("VOYAGE_AI_TOKEN", "third")
+    assert get_voyage_api_key() == "second"
+
+    monkeypatch.setenv("VOYAGE_API_KEY", "first")
+    assert get_voyage_api_key() == "first"
+    assert get_voyage_api_key("explicit") == "explicit"

tests/unit/llms/voyage/test_voyage_multimodal_embedding.py

@@ -172,15 +172,12 @@ class TestVoyageMultimodalEmbeddings:
         assert headers == {"Authorization": "Bearer test-key"}
 
     def test_validate_environment_uses_secret_fallback(self, monkeypatch):
-        import litellm.llms.voyage.embedding.transformation_multimodal as module
         from litellm.llms.voyage.embedding.transformation_multimodal import (
             VoyageMultimodalEmbeddingConfig,
         )
 
-        def fake_get_secret(name):
-            return "secret-key" if name == "VOYAGE_AI_API_KEY" else None
-
-        monkeypatch.setattr(module, "get_secret_str", fake_get_secret)
+        monkeypatch.delenv("VOYAGE_API_KEY", raising=False)
+        monkeypatch.setenv("VOYAGE_AI_API_KEY", "secret-key")
         config = VoyageMultimodalEmbeddingConfig()
         headers = config.validate_environment(
             {}, "voyage-multimodal-3.5", [], {}, {}, api_key=None
@@ -188,12 +185,12 @@ class TestVoyageMultimodalEmbeddings:
         assert headers == {"Authorization": "Bearer secret-key"}
 
     def test_validate_environment_raises_without_api_key(self, monkeypatch):
-        import litellm.llms.voyage.embedding.transformation_multimodal as module
         from litellm.llms.voyage.embedding.transformation_multimodal import (
             VoyageMultimodalEmbeddingConfig,
         )
 
-        monkeypatch.setattr(module, "get_secret_str", lambda name: None)
+        for env_var in ("VOYAGE_API_KEY", "VOYAGE_AI_API_KEY", "VOYAGE_AI_TOKEN"):
+            monkeypatch.delenv(env_var, raising=False)
         config = VoyageMultimodalEmbeddingConfig()
         with pytest.raises(ValueError, match='Voyage API key is required for multimodal embeddings\\. Set') as exc_info:
             config.validate_environment(

ui/litellm-dashboard/src/components/provider_info_helpers.tsx

@@ -178,7 +178,7 @@ export enum Providers {
   VERTEX_AI_BETA = "Vertex Ai Beta",
   VLLM = "Local vLLM",
   VolcEngine = "VolcEngine",
-  Voyage = "Voyage AI",
+  Voyage = "VoyageAI by MongoDB",
   WANDB = "Wandb",
   WATSONX = "Watsonx",
   WATSONX_TEXT = "Watsonx Text",