Qdrant v0.0.8

View Source Qdrant Elixir Client

this-library-is-under-active-development-and-is-subject-to-change-please-use-the-latest-version-from-github

 ⚠️ This library is under active development and is subject to change. Please use the latest version from GitHub ⚠️

An Elixir client for the Qdrant vector similarity search engine. This library provides a convenient way to interact with the Qdrant API, offering functionality to create collections, insert vectors, search, delete data, and more.

Hex.pm Hex.pm Hex.pm

installation

 Installation

It's available in Hex, the package can be installed by adding qdrant to your list of dependencies in mix.exs:

def deps do
  [
    {:qdrant, "~> 0.8.0"}
    # Or use the latest version from GitHub | Recommended during development phase
    {:qdrant, git: "git@github.com:marinac-dev/qdrant.git"},
  ]
end

config

 Config 

config :qdrant,
  port: 6333,
  interface: "rest", # gRPC not yet supported
  database_url: System.get_env("QDRANT_DATABASE_URL"),
  # If you are using cloud version of Qdrant, add API key
  api_key: System.get_env("QDRANT_API_KEY")

usage

 Usage

The Qdrant Elixir Client provides a simple interface for interacting with the Qdrant API. For example, you can create a new collection, insert vectors, search, and delete data using the provided functions.

collection_name = "my-collection"

# Create a new collection
# The vectors are 1536-dimensional (because of OpenAi embedding) and use the Cosine distance metric
Qdrant.create_collection(collection_name, %{vectors: %{size: 1536, distance: "Cosine"}})

# Create embeddings for some text
vector1 = OpenAi.embed_text("Hello world")
vector2 = OpenAi.embed_text("This is OpenAI")

# Now we can insert the vectors with batch
Qdrant.upsert_points(collection_name, %{batch: %{ids: [1,2], vectors: [vector1, vector2]}})
# Or one by one
Qdrant.upsert_point(collection_name, %{points: [%{id: 1, vector: vector1}, %{id: 2, vector: vector2}]})

# Search for similar vectors
vector3 = OpenAi.embed_text("Hello world!")
Qdrant.search(collection_name, %{vector: vector3, limit: 3})

contributing

 Contributing

  • Fork the repository
  • Create a branch for your changes
  • Make your changes
  • Run mix format to format your code

change-log

 Change Log

Generate change log with git-chglog -o CHANGELOG.md

license

 License

This project is licensed under the MIT License - see the LICENSE file for details