Supported in v0.6.1+

Use LiteLLMEmbeddings to embed text across 100+ providers with a single, consistent interface. All configuration is explicit -- no environment variables required.

from langchain_litellm import LiteLLMEmbeddings

embeddings = LiteLLMEmbeddings(
    model="openai/text-embedding-3-small",
    api_key="sk-...",
)

vectors = embeddings.embed_documents(["hello", "world"])
query_vector = embeddings.embed_query("hello")

Switch providers by changing model -- the interface stays the same:

# Cohere
embeddings = LiteLLMEmbeddings(
    model="cohere/embed-english-v3.0",
    api_key="...",
    document_input_type="search_document",
    query_input_type="search_query",
)

# Azure OpenAI
embeddings = LiteLLMEmbeddings(
    model="azure/my-embedding-deployment",
    api_key="...",
    api_base="https://my-resource.openai.azure.com",
    api_version="2024-02-01",
)

# Bedrock
embeddings = LiteLLMEmbeddings(
    model="bedrock/amazon.titan-embed-text-v1",
)

For load-balancing across multiple deployments of the same model, use LiteLLMEmbeddingsRouter:

from litellm import Router
from langchain_litellm import LiteLLMEmbeddingsRouter

router = Router(model_list=[
    {
        "model_name": "text-embedding-3-small",
        "litellm_params": {
            "model": "openai/text-embedding-3-small",
            "api_key": "sk-key1",
        },
    },
    {
        "model_name": "text-embedding-3-small",
        "litellm_params": {
            "model": "openai/text-embedding-3-small",
            "api_key": "sk-key2",
        },
    },
])

embeddings = LiteLLMEmbeddingsRouter(router=router)

Supported in v0.3.5+

You can use Google Search grounding with Vertex AI models (e.g., gemini-2.5-flash). Citations and metadata are returned in response_metadata (Batch) or additional_kwargs (Streaming).

Setup

import os
from langchain_litellm import ChatLiteLLM

os.environ["VERTEX_PROJECT"] = "your-project-id"
os.environ["VERTEX_LOCATION"] = "us-central1"

llm = ChatLiteLLM(model="vertex_ai/gemini-2.5-flash", temperature=0)

Batch Usage

# Invoke with Google Search tool enabled
response = llm.invoke(
    "What is the current stock price of Google?",
    tools=[{"googleSearch": {}}]
)

# Access Citations & Metadata
provider_fields = response.response_metadata.get("provider_specific_fields")
if provider_fields:
    # Vertex returns a list; the first item contains the grounding info
    print(provider_fields[0])

Streaming Usage

stream = llm.stream(
    "What is the current stock price of Google?",
    tools=[{"googleSearch": {}}]
)

for chunk in stream:
    print(chunk.content, end="", flush=True)
    # Metadata is injected into the chunk where it arrives
    if "provider_specific_fields" in chunk.additional_kwargs:
        print("\n[Metadata Found]:", chunk.additional_kwargs["provider_specific_fields"])