How to beginner · 3 min read

How to use HuggingFace embeddings in LangChain

Q: How to use HuggingFace embeddings in LangChain

Use the HuggingFaceEmbeddings class from langchain.embeddings to load HuggingFace models as embeddings in LangChain. Initialize it with the model name, then generate embeddings for your text to use with vector stores like FAISS.

Quick answer

Use the HuggingFaceEmbeddings class from langchain.embeddings to load HuggingFace models as embeddings in LangChain. Initialize it with the model name, then generate embeddings for your text to use with vector stores like FAISS.

PREREQUISITES

Python 3.8+
pip install langchain huggingface_hub faiss-cpu
HuggingFace API token (optional for some models)

Setup

Install the required packages and set up environment variables if you plan to use private HuggingFace models.

bash

pip install langchain huggingface_hub faiss-cpu

Step by step

This example shows how to load a HuggingFace embedding model in LangChain, generate embeddings for sample texts, and store them in a FAISS vector store.

python

from langchain_openai import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS

# Initialize HuggingFace embeddings with a model
embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")

# Sample texts to embed
texts = ["Hello world", "LangChain with HuggingFace embeddings", "Vector search example"]

# Generate embeddings
embeddings = [embedding_model.embed_query(text) for text in texts]

# Create FAISS vector store from texts and embeddings
vector_store = FAISS.from_texts(texts, embedding_model)

# Query vector store
query = "How to use embeddings?"
query_embedding = embedding_model.embed_query(query)

# Search top 2 similar texts
results = vector_store.similarity_search_by_vector(query_embedding, k=2)

for i, doc in enumerate(results, 1):
    print(f"Result {i}: {doc.page_content}")

output

Result 1: LangChain with HuggingFace embeddings
Result 2: Hello world

Common variations

Use different HuggingFace models by changing model_name, e.g., "sentence-transformers/all-mpnet-base-v2".
Use async embedding calls if supported by the model wrapper.
Integrate with other vector stores like Chroma or Pinecone.

Troubleshooting

If you get authentication errors, ensure your HuggingFace API token is set in HUGGINGFACEHUB_API_TOKEN environment variable.
For slow embedding generation, check model size and consider smaller models.
If embeddings are empty or errors occur, verify model compatibility with LangChain's HuggingFaceEmbeddings class.

✅

Key Takeaways

Use HuggingFaceEmbeddings in LangChain to leverage HuggingFace models for embeddings.
Initialize with the model name and generate embeddings for texts to use with vector stores like FAISS.
Set HUGGINGFACEHUB_API_TOKEN environment variable if accessing private or gated models.
You can switch models easily by changing the model_name parameter.
Troubleshoot by verifying token setup and model compatibility.

Verified 2026-04 · sentence-transformers/all-MiniLM-L6-v2, sentence-transformers/all-mpnet-base-v2

Verify ↗