a simple and blazing fast vector database

there is no database you need to set up, no server to run, and no complex configuration. just install vlite and start using it. take the CTX file with you wherever you go. its like a browser cookie but with embeddings.

• 🔥 Fastest vector db retrieval with binary embeddings, less than 1.1s to search 500k documents
• 🔋 Made for RAG -- with embedding generation with mixedbread embed-large baked in
• 🍪 CTX (context) file format, a novel abstraction for storing user context similar to browser cookies
• Ingest text, PDF, CSV, PPTX, and webpages
• Chunking, metadata filtering, PDF OCR support for extracting text from scanned PDFs
• >77.95% faster than Chroma on indexing, >422% faster on retrieval, and >3.6x smaller on disk
• 🦜 Available in LangChain since vlite v0.2.2

pip install vlite

To enable PDF OCR support (with surya), install the vlite[ocr] extra:

pip install vlite[ocr]

from vlite import VLite
from vlite.utils import process_pdf

vdb = VLite()
vdb.add("hello world", metadata={"artist": "adele"})
vdb.add(process_pdf("attention-is-all-you-need.pdf", use_ocr=True))

results = vdb.retrieve("how do transformers work?")
print(results)

from langchain.document_loaders import TextLoader
from langchain.text_splitter import CharacterTextSplitter
from langchain.vectorstores import VLite

# Load the document and split it into chunks
loader = TextLoader("path/to/document.txt")
documents = loader.load()

# Create a VLite instance
vlite = VLite(collection="my_collection")

# Add documents to the VLite vector database
vlite.add_documents(documents)

# Perform a similarity search
query = "What is the main topic of the document?"
docs = vlite.similarity_search(query)

# Print the most relevant document
print(docs[0].page_content)

vlite is a vector database built for agents, ChatGPT Plugins, and other AI apps that need a fast and simple database to store vectors. It was developed to support the billions of embeddings generated, indexed, and sorted with ChatWith+ ChatGPT Plugins, which run for millions of users. Most vector databases either repeatedly crashed on a daily basis or were too expensive for the high throughput required.

vlite introduces the CTX file format, which acts like a browser cookie for user embeddings, providing efficient storage, retrieval of embeddings, composability, portability, and user context.

AGPL-3.0 License

Thanks to Claude and Ray for their contributions to vlite. If you'd like to contribute, please open an issue or a pull request.