diff --git a/litellm/llms/voyage/embedding/transformation_contextual.py b/litellm/llms/voyage/embedding/transformation_contextual.py index 4df2fa4ba31..2d661610eaa 100644 --- a/litellm/llms/voyage/embedding/transformation_contextual.py +++ b/litellm/llms/voyage/embedding/transformation_contextual.py @@ -106,11 +106,42 @@ class VoyageContextualEmbeddingConfig(BaseEmbeddingConfig): headers: dict, ) -> dict: return { - "inputs": input, + "inputs": self._transform_contextual_inputs(input, optional_params), "model": model, **optional_params, } + @staticmethod + def _transform_contextual_inputs( + input: Union[AllEmbeddingInputValues, List[List[str]]], + optional_params: dict, + ) -> List[List[str]]: + """ + Voyage's contextualized embeddings API expects ``inputs`` to be a + ``list[list[str]]`` (each inner list is a document made of chunks that + share context). + + It also accepts a flat ``list[str]`` - but only when + ``input_type == "query"``. In every other case a flat list must be + wrapped so each string becomes its own single-chunk document, otherwise + the API rejects the request with a 400. + + Reference: https://docs.voyageai.com/reference/contextualized-embeddings-api + """ + # Single string -> one document with a single chunk. + if isinstance(input, str): + return [[input]] + + # Flat list[str]: keep as list[str] when the API allows it + # (input_type="query"), otherwise wrap each string as its own document. + if isinstance(input, list) and all(isinstance(i, str) for i in input): + if optional_params.get("input_type") == "query": + return input # type: ignore[return-value] + return [[i] for i in input] + + # Already list[list[str]] (or another shape) -> pass through unchanged. + return input # type: ignore[return-value] + def transform_embedding_response( self, model: str, diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index c584deb683a..fc710497592 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -27345,6 +27345,38 @@ "mode": "embedding", "output_cost_per_token": 0.0 }, + "voyage/voyage-4": { + "input_cost_per_token": 6e-08, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, + "voyage/voyage-4-large": { + "input_cost_per_token": 1.2e-07, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, + "voyage/voyage-4-lite": { + "input_cost_per_token": 2e-08, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, + "voyage/voyage-4-nano": { + "input_cost_per_token": 0.0, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, "voyage/voyage-code-2": { "input_cost_per_token": 1.2e-07, "litellm_provider": "voyage", @@ -27369,6 +27401,14 @@ "mode": "embedding", "output_cost_per_token": 0.0 }, + "voyage/voyage-context-4": { + "input_cost_per_token": 1.2e-07, + "litellm_provider": "voyage", + "max_input_tokens": 120000, + "max_tokens": 120000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, "voyage/voyage-finance-2": { "input_cost_per_token": 1.2e-07, "litellm_provider": "voyage", diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index c584deb683a..fc710497592 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -27345,6 +27345,38 @@ "mode": "embedding", "output_cost_per_token": 0.0 }, + "voyage/voyage-4": { + "input_cost_per_token": 6e-08, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, + "voyage/voyage-4-large": { + "input_cost_per_token": 1.2e-07, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, + "voyage/voyage-4-lite": { + "input_cost_per_token": 2e-08, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, + "voyage/voyage-4-nano": { + "input_cost_per_token": 0.0, + "litellm_provider": "voyage", + "max_input_tokens": 32000, + "max_tokens": 32000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, "voyage/voyage-code-2": { "input_cost_per_token": 1.2e-07, "litellm_provider": "voyage", @@ -27369,6 +27401,14 @@ "mode": "embedding", "output_cost_per_token": 0.0 }, + "voyage/voyage-context-4": { + "input_cost_per_token": 1.2e-07, + "litellm_provider": "voyage", + "max_input_tokens": 120000, + "max_tokens": 120000, + "mode": "embedding", + "output_cost_per_token": 0.0 + }, "voyage/voyage-finance-2": { "input_cost_per_token": 1.2e-07, "litellm_provider": "voyage", diff --git a/tests/llm_translation/test_voyage_ai.py b/tests/llm_translation/test_voyage_ai.py index a0b9ee0a44b..577ee808035 100644 --- a/tests/llm_translation/test_voyage_ai.py +++ b/tests/llm_translation/test_voyage_ai.py @@ -71,6 +71,26 @@ class TestVoyageAI(BaseLLMEmbeddingTest): assert response.usage.total_tokens > 0 +@pytest.mark.parametrize( + "model, expected_input_cost", + [ + ("voyage/voyage-4", 6e-08), + ("voyage/voyage-4-large", 1.2e-07), + ("voyage/voyage-4-lite", 2e-08), + ("voyage/voyage-4-nano", 0.0), + ("voyage/voyage-context-4", 1.2e-07), + ], +) +def test_voyage_4_family_pricing_registered(model, expected_input_cost): + """The voyage-4 family and voyage-context-4 must be in the cost map.""" + os.environ["LITELLM_LOCAL_MODEL_COST_MAP"] = "True" + litellm.model_cost = litellm.get_model_cost_map(url="") + info = litellm.get_model_info(model=model) + assert info["litellm_provider"] == "voyage" + assert info["mode"] == "embedding" + assert info["input_cost_per_token"] == expected_input_cost + + def test_voyage_ai_embedding_extra_params(): """Test Voyage AI embedding with extra parameters""" try: @@ -142,6 +162,7 @@ class TestVoyageContextualEmbeddings: # Test contextual model detection assert config.is_contextualized_embeddings("voyage-context-3") is True + assert config.is_contextualized_embeddings("voyage-context-4") is True assert config.is_contextualized_embeddings("voyage-context-2") is True assert config.is_contextualized_embeddings("context-model") is True @@ -198,6 +219,70 @@ class TestVoyageContextualEmbeddings: assert transformed["model"] == "voyage-context-3" assert transformed["encoding_format"] == "float" + def test_contextual_flat_list_str_is_wrapped(self): + """Flat list[str] without input_type must be wrapped to list[list[str]].""" + from litellm.llms.voyage.embedding.transformation_contextual import ( + VoyageContextualEmbeddingConfig, + ) + + config = VoyageContextualEmbeddingConfig() + + transformed = config.transform_embedding_request( + "voyage-context-4", ["Hello", "world"], {}, {} + ) + + # Voyage rejects a flat list[str] unless input_type="query", so each + # string becomes its own single-chunk document. + assert transformed["inputs"] == [["Hello"], ["world"]] + assert transformed["model"] == "voyage-context-4" + + def test_contextual_flat_list_str_query_stays_flat(self): + """Flat list[str] with input_type='query' is sent as-is (list[str]).""" + from litellm.llms.voyage.embedding.transformation_contextual import ( + VoyageContextualEmbeddingConfig, + ) + + config = VoyageContextualEmbeddingConfig() + + transformed = config.transform_embedding_request( + "voyage-context-4", + ["Hello", "world"], + {"input_type": "query"}, + {}, + ) + + assert transformed["inputs"] == ["Hello", "world"] + assert transformed["input_type"] == "query" + + def test_contextual_single_string_is_wrapped(self): + """A single string is wrapped to a single document with a single chunk.""" + from litellm.llms.voyage.embedding.transformation_contextual import ( + VoyageContextualEmbeddingConfig, + ) + + config = VoyageContextualEmbeddingConfig() + + transformed = config.transform_embedding_request( + "voyage-context-4", "Hello", {}, {} + ) + + assert transformed["inputs"] == [["Hello"]] + + def test_contextual_nested_input_passthrough(self): + """Already-nested list[list[str]] input is passed through unchanged.""" + from litellm.llms.voyage.embedding.transformation_contextual import ( + VoyageContextualEmbeddingConfig, + ) + + config = VoyageContextualEmbeddingConfig() + + nested = [["Hello", "world"], ["Test"]] + transformed = config.transform_embedding_request( + "voyage-context-4", nested, {}, {} + ) + + assert transformed["inputs"] == nested + def test_contextual_embedding_response_transformation(self): """Test response transformation for contextual embeddings""" from litellm.llms.voyage.embedding.transformation_contextual import (