Implementation:Langchain ai Langchain OllamaEmbeddings

Knowledge Sources	Langchain_ai_Langchain
Domains	Embeddings, Ollama, Local Inference
Last Updated	2026-02-11 00:00 GMT

Overview

OllamaEmbeddings is a LangChain embedding model integration that generates text embeddings using a locally running Ollama server.

Description

The OllamaEmbeddings class, defined in the langchain-ollama partner package, extends both BaseModel (Pydantic) and Embeddings (LangChain core). It connects to a local Ollama server via the official ollama Python client to generate embeddings using locally hosted models. The class supports both synchronous and asynchronous embedding operations, configurable sampling parameters (temperature, top_k, top_p, mirostat, etc.), custom httpx client kwargs, and URL-based authentication for proxied Ollama servers. It optionally validates that the specified model exists locally on initialization.

Usage

Import this class when you need to generate text embeddings using models hosted locally through Ollama, such as for semantic search or RAG pipelines in air-gapped or privacy-sensitive environments.

Code Reference

Source Location

• Repository: Langchain_ai_Langchain
• File: libs/partners/ollama/langchain_ollama/embeddings.py
• Lines: 1-332

Signature

class OllamaEmbeddings(BaseModel, Embeddings):
    model: str
    validate_model_on_init: bool = False
    base_url: str | None = None
    client_kwargs: dict | None = {}
    async_client_kwargs: dict | None = {}
    sync_client_kwargs: dict | None = {}
    mirostat: int | None = None
    mirostat_eta: float | None = None
    mirostat_tau: float | None = None
    num_ctx: int | None = None
    num_gpu: int | None = None
    keep_alive: int | None = None
    num_thread: int | None = None
    repeat_last_n: int | None = None
    repeat_penalty: float | None = None
    temperature: float | None = None
    stop: list[str] | None = None
    tfs_z: float | None = None
    top_k: int | None = None
    top_p: float | None = None

Import

from langchain_ollama import OllamaEmbeddings

I/O Contract

Inputs

Name	Type	Required	Description
model	str	Yes	Name of the Ollama model to use for generating embeddings.
base_url	str or None	No	Base URL where the Ollama server is hosted. Defaults to the Ollama client default. Supports userinfo auth in the URL.
validate_model_on_init	bool	No	Whether to validate the model exists locally on initialization. Defaults to False.
client_kwargs	dict or None	No	Additional kwargs passed to both sync and async httpx clients.
temperature	float or None	No	Sampling temperature. Higher values produce more creative output.
num_ctx	int or None	No	Context window size for token generation. Defaults to 2048.
keep_alive	int or None	No	Duration in seconds the model stays loaded in memory after a request.
top_k	int or None	No	Limits next-token selection to K most probable tokens. Defaults to 40.
top_p	float or None	No	Nucleus sampling parameter. Defaults to 0.9.

Outputs

Name	Type	Description
embed_documents	list[list[float]]	Returns a list of embedding vectors, one per input document.
embed_query	list[float]	Returns a single embedding vector for a query text.

Usage Examples

Basic Usage

from langchain_ollama import OllamaEmbeddings

embed = OllamaEmbeddings(model="llama3")

# Embed a single query
vector = embed.embed_query("The meaning of life is 42")
print(vector[:3])

# Embed multiple documents
vectors = embed.embed_documents(["Document 1...", "Document 2..."])
print(len(vectors))

Async Usage

from langchain_ollama import OllamaEmbeddings

embed = OllamaEmbeddings(model="llama3")

vector = await embed.aembed_query("The meaning of life is 42")
print(vector[:3])