Supabase Vector Store

To use this vector store, you need a Supabase project. You can create one at supabase.com.

Installation

npm i llamaindex @llamaindex/supabase

Database Setup

Before using the vector store, you need to:

Enable the pgvector extension
Create a table for storing vectors
Create a vector similarity search function

create table documents (
id uuid primary key,
content text,
metadata jsonb,
embedding vector(1536)
);

— Create a function for similarity search with filtering support

create function match_documents (
query_embedding vector(1536),
match_count int,
filter jsonb DEFAULT '{}'
) returns table (
id uuid,
content text,
metadata jsonb,
embedding vector(1536),
similarity float
)
language plpgsql
as $$
#variable_conflict use_column
begin
return query
select
id,
content,
metadata,
embedding,
1 - (embedding <=> query_embedding) as similarity
from documents
where metadata @> filter
order by embedding <=> query_embedding
limit match_count;
end;
$$;

Importing the modules

import { Document, VectorStoreIndex } from "llamaindex";
import { SupabaseVectorStore } from "@llamaindex/supabase";

Setup Supabase

const vectorStore = new SupabaseVectorStore({
  supabaseUrl: process.env.SUPABASE_URL,
  supabaseKey: process.env.SUPABASE_KEY,
  table: "documents",
});

Setup the index

const documents = [
  new Document({
    text: "Sample document text",
    metadata: { source: "example" }
  })
];

const storageContext = await storageContextFromDefaults({ vectorStore });
const index = await VectorStoreIndex.fromDocuments(documents, {
  storageContext,
});

Query the index

const queryEngine = index.asQueryEngine();

// Basic query without filters
const response = await queryEngine.query({
  query: "What is in the document?",
});

// Output response
console.log(response.toString());

Query with filters

You can filter documents based on metadata when querying:

import { FilterOperator, MetadataFilters } from "llamaindex";

// Create a filter for documents with author = "Jane Smith"
const filters: MetadataFilters = {
  filters: [
    {
      key: "author",
      value: "Jane Smith",
      operator: FilterOperator.EQ,
    },
  ],
};

// Query with filters
const filteredResponse = await vectorStore.query({
  queryEmbedding: embedModel.getQueryEmbedding("What is vector search?"),
  similarityTopK: 5,
  filters,
});

Full code

import { Document, VectorStoreIndex, storageContextFromDefaults } from "llamaindex";
import { SupabaseVectorStore } from "@llamaindex/supabase";

async function main() {
  // Initialize the vector store
  const vectorStore = new SupabaseVectorStore({
    supabaseUrl: process.env.SUPABASE_URL,
    supabaseKey: process.env.SUPABASE_KEY,
    table: "documents",
  });

  // Create sample documents
  const documents = [
    new Document({
      text: "Vector search enables semantic similarity search",
      metadata: {
        source: "research_paper",
        author: "Jane Smith",
      },
    }),
  ];

  // Create storage context
  const storageContext = await storageContextFromDefaults({ vectorStore });

  // Create and store embeddings
  const index = await VectorStoreIndex.fromDocuments(documents, {
    storageContext,
  });

  // Query the index
  const queryEngine = index.asQueryEngine();
  const response = await queryEngine.query({
    query: "What is vector search?",
  });

  // Output response
  console.log(response.toString());
}

main().catch(console.error);

API Reference

SupabaseVectorStore