# pgvector-elixir
[pgvector](https://github.com/pgvector/pgvector) support for Elixir
Supports [Ecto](https://github.com/elixir-ecto/ecto) and [Postgrex](https://github.com/elixir-ecto/postgrex)
[](https://github.com/pgvector/pgvector-elixir/actions)
## Installation
Add this line to your application’s `mix.exs` under `deps`:
```elixir
{:pgvector, "~> 0.2.0"}
```
And follow the instructions for your database library:
- [Ecto](#ecto)
- [Postgrex](#postgrex)
Or check out an example:
- [Embeddings](examples/openai/example.exs) with OpenAI
## Ecto
Create `lib/postgrex_types.ex` with:
```elixir
Postgrex.Types.define(MyApp.PostgrexTypes, [Pgvector.Extensions.Vector] ++ Ecto.Adapters.Postgres.extensions(), [])
```
And add to `config/config.exs`:
```elixir
config :my_app, MyApp.Repo, types: MyApp.PostgrexTypes
```
Create a migration
```sh
mix ecto.gen.migration create_vector_extension
```
with:
```elixir
defmodule MyApp.Repo.Migrations.CreateVectorExtension do
use Ecto.Migration
def up do
execute "CREATE EXTENSION IF NOT EXISTS vector"
end
def down do
execute "DROP EXTENSION vector"
end
end
```
Run the migration
```sh
mix ecto.migrate
```
You can now use the `vector` type in future migrations
```elixir
create table(:items) do
add :embedding, :vector, size: 3
end
```
Update the model
```elixir
schema "items" do
field :embedding, Pgvector.Ecto.Vector
end
```
Insert a vector
```elixir
alias MyApp.{Repo, Item}
Repo.insert(%Item{embedding: [1, 2, 3]})
```
Get the nearest neighbors
```elixir
import Ecto.Query
import Pgvector.Ecto.Query
Repo.all(from i in Item, order_by: l2_distance(i.embedding, [1, 2, 3]), limit: 5)
```
Also supports `max_inner_product` and `cosine_distance`
Convert a vector to a list or Nx tensor
```elixir
item.embedding |> Pgvector.to_list()
item.embedding |> Pgvector.to_tensor()
```
Add an approximate index in a migration
```elixir
create index("items", ["embedding vector_l2_ops"], using: :ivfflat)
# or
create index("items", ["embedding vector_l2_ops"], using: :hnsw)
```
Use `vector_ip_ops` for inner product and `vector_cosine_ops` for cosine distance
## Postgrex
[Register](https://github.com/elixir-ecto/postgrex#extensions) the extension
```elixir
Postgrex.Types.define(MyApp.PostgrexTypes, [Pgvector.Extensions.Vector], [])
```
And pass it to `start_link`
```elixir
{:ok, pid} = Postgrex.start_link(types: MyApp.PostgrexTypes)
```
Enable the extension
```elixir
Postgrex.query!(pid, "CREATE EXTENSION IF NOT EXISTS vector", [])
```
Create a table
```elixir
Postgrex.query!(pid, "CREATE TABLE items (embedding vector(3))", [])
```
Insert a vector
```elixir
Postgrex.query!(pid, "INSERT INTO items (embedding) VALUES ($1)", [[1, 2, 3]])
```
Get the nearest neighbors
```elixir
Postgrex.query!(pid, "SELECT * FROM items ORDER BY embedding <-> $1 LIMIT 5", [[1, 2, 3]])
```
Convert a vector to a list or Nx tensor
```elixir
vector |> Pgvector.to_list()
vector |> Pgvector.to_tensor()
```
Add an approximate index
```elixir
Postgrex.query!(pid, "CREATE INDEX ON items USING ivfflat (embedding vector_l2_ops) WITH (lists = 100)", [])
# or
Postgrex.query!(pid, "CREATE INDEX ON items USING hnsw (embedding vector_l2_ops)", [])
```
Use `vector_ip_ops` for inner product and `vector_cosine_ops` for cosine distance
## Upgrading
### 0.2.0
Vectors are now returned as `Pgvector` structs instead of lists. Get a list with:
```elixir
vector |> Pgvector.to_list()
```
or an Nx tensor with:
```elixir
vector |> Pgvector.to_tensor()
```
## History
View the [changelog](https://github.com/pgvector/pgvector-elixir/blob/master/CHANGELOG.md)
## Contributing
Everyone is encouraged to help improve this project. Here are a few ways you can help:
- [Report bugs](https://github.com/pgvector/pgvector-elixir/issues)
- Fix bugs and [submit pull requests](https://github.com/pgvector/pgvector-elixir/pulls)
- Write, clarify, or fix documentation
- Suggest or add new features
To get started with development:
```sh
git clone https://github.com/pgvector/pgvector-elixir.git
cd pgvector-elixir
mix deps.get
createdb pgvector_elixir_test
mix test
```
