Merge 6a910814bb into 08f2a98250

litellm/llms/voyage/common_utils.py

@@ -33,3 +33,8 @@ def get_default_base_url(api_key: str | None = None) -> str:
     if resolved is not None and resolved.startswith(MONGODB_API_KEY_PREFIX):
         return MONGODB_API_BASE
     return VOYAGE_API_BASE
+
+
+def get_voyage_api_base(api_base: str | None = None, api_key: str | None = None) -> str:
+    """Resolve the host to call, explicit value first, then ``VOYAGE_API_BASE``, then the key's issuer."""
+    return api_base or get_secret_str("VOYAGE_API_BASE") or get_default_base_url(api_key)

litellm/llms/voyage/embedding/transformation.py

@@ -8,7 +8,7 @@ from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLogging
 from litellm.litellm_core_utils.url_utils import ensure_api_route
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
-from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
+from litellm.llms.voyage.common_utils import get_voyage_api_base, get_voyage_api_key
 from litellm.types.llms.openai import AllEmbeddingInputValues, AllMessageValues
 from litellm.types.utils import EmbeddingResponse, Usage
 
@@ -51,7 +51,7 @@ class VoyageEmbeddingConfig(BaseEmbeddingConfig):
         litellm_params: dict,
         stream: bool | None = None,
     ) -> str:
-        return ensure_api_route(api_base or get_default_base_url(api_key), "embeddings")
+        return ensure_api_route(get_voyage_api_base(api_base, api_key), "embeddings")
 
     def get_supported_openai_params(self, model: str) -> list:
         return [

litellm/llms/voyage/embedding/transformation_contextual.py

@@ -13,7 +13,7 @@ from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLogging
 from litellm.litellm_core_utils.url_utils import ensure_api_route
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
-from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
+from litellm.llms.voyage.common_utils import get_voyage_api_base, get_voyage_api_key
 from litellm.types.llms.openai import AllEmbeddingInputValues, AllMessageValues
 from litellm.types.utils import EmbeddingResponse, Usage
 
@@ -59,7 +59,7 @@ class VoyageContextualEmbeddingConfig(BaseEmbeddingConfig):
         litellm_params: dict,
         stream: bool | None = None,
     ) -> str:
-        return ensure_api_route(api_base or get_default_base_url(api_key), "contextualizedembeddings")
+        return ensure_api_route(get_voyage_api_base(api_base, api_key), "contextualizedembeddings")
 
     def get_supported_openai_params(self, model: str) -> list:  # mutable-ok: base class signature
         return ["encoding_format", "dimensions"]

litellm/llms/voyage/embedding/transformation_multimodal.py

@@ -16,7 +16,7 @@ from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLogging
 from litellm.litellm_core_utils.url_utils import ensure_api_route
 from litellm.llms.base_llm.chat.transformation import BaseLLMException
 from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
-from litellm.llms.voyage.common_utils import get_default_base_url, get_voyage_api_key
+from litellm.llms.voyage.common_utils import get_voyage_api_base, get_voyage_api_key
 from litellm.types.llms.openai import AllEmbeddingInputValues, AllMessageValues
 from litellm.types.utils import EmbeddingResponse, Usage
 
@@ -60,7 +60,7 @@ class VoyageMultimodalEmbeddingConfig(BaseEmbeddingConfig):
         litellm_params: dict,
         stream: bool | None = None,
     ) -> str:
-        return ensure_api_route(api_base or get_default_base_url(api_key), "multimodalembeddings")
+        return ensure_api_route(get_voyage_api_base(api_base, api_key), "multimodalembeddings")
 
     def get_supported_openai_params(self, model: str) -> list:
         return ["dimensions"]

tests/unit/llms/voyage/test_common_utils.py

@@ -4,6 +4,7 @@ from litellm.llms.voyage.common_utils import (
     MONGODB_API_BASE,
     VOYAGE_API_BASE,
     get_default_base_url,
+    get_voyage_api_base,
     get_voyage_api_key,
 )
 from litellm.llms.voyage.embedding.transformation import VoyageEmbeddingConfig
@@ -20,7 +21,7 @@ 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:
+    for env_var in (*VOYAGE_KEY_ENV_VARS, "VOYAGE_API_BASE"):
         monkeypatch.delenv(env_var, raising=False)
 
 
@@ -72,6 +73,50 @@ def test_explicit_api_base_overrides_key_routing(config, endpoint):
     assert url == f"https://gateway.internal/v1/{endpoint}"
 
 
+@pytest.mark.parametrize(
+    "config, endpoint",
+    [
+        (VoyageEmbeddingConfig(), "embeddings"),
+        (VoyageContextualEmbeddingConfig(), "contextualizedembeddings"),
+        (VoyageMultimodalEmbeddingConfig(), "multimodalembeddings"),
+    ],
+)
+@pytest.mark.parametrize("env_value", ["https://gateway.internal/v1", "https://gateway.internal/v1/"])
+def test_env_api_base_overrides_key_routing(monkeypatch, config, endpoint, env_value):
+    """A caller who only gets to pass a model and a key reaches a custom host through the env var"""
+    monkeypatch.setenv("VOYAGE_API_BASE", env_value)
+
+    url = config.get_complete_url(None, "al-key", "voyage-3", {}, {})
+
+    assert url == f"https://gateway.internal/v1/{endpoint}"
+
+
+@pytest.mark.parametrize(
+    "config, endpoint",
+    [
+        (VoyageEmbeddingConfig(), "embeddings"),
+        (VoyageContextualEmbeddingConfig(), "contextualizedembeddings"),
+        (VoyageMultimodalEmbeddingConfig(), "multimodalembeddings"),
+    ],
+)
+def test_explicit_api_base_beats_env_api_base(monkeypatch, config, endpoint):
+    monkeypatch.setenv("VOYAGE_API_BASE", "https://from-env.internal/v1")
+
+    url = config.get_complete_url("https://on-request.internal/v1", "pa-key", "voyage-3", {}, {})
+
+    assert url == f"https://on-request.internal/v1/{endpoint}"
+
+
+def test_get_voyage_api_base_prefers_request_then_env_then_key_issuer(monkeypatch):
+    assert get_voyage_api_base("https://on-request.internal/v1", "al-key") == "https://on-request.internal/v1"
+    assert get_voyage_api_base(None, "al-key") == MONGODB_API_BASE
+
+    monkeypatch.setenv("VOYAGE_API_BASE", "https://from-env.internal/v1")
+
+    assert get_voyage_api_base(None, "al-key") == "https://from-env.internal/v1"
+    assert get_voyage_api_base("https://on-request.internal/v1", "al-key") == "https://on-request.internal/v1"
+
+
 @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()