@ragpipe/plugin-supabase

@ragpipe/plugin-supabase

Supabase vector store plugin for ragpipe, powered by @supabase/supabase-js.

Install

pnpm add ragpipe @ragpipe/plugin-supabase

Usage

import { defineConfig } from "ragpipe";
import { supabaseVectorStore } from "@ragpipe/plugin-supabase";

export default defineConfig({
  // ... embedding, generation
  vectorStore: supabaseVectorStore({
    supabaseUrl: process.env.SUPABASE_URL ?? "",
    supabaseKey: process.env.SUPABASE_SERVICE_ROLE_KEY ?? "",
    tableName: "documents",   // default
    queryName: "match_documents", // default RPC function name
  }),
});

API

supabaseVectorStore(options)

Returns a VectorStorePlugin backed by Supabase using the official JS SDK.

Option	Type	Default	Description
supabaseUrl	string	—	Supabase project URL (required)
supabaseKey	string	—	Service role key (required)
tableName	string	"documents"	Table to store documents
queryName	string	"match_documents"	PostgreSQL function for vector search

Methods

Method	Description
search(vector, topK)	Calls supabase.rpc() for cosine similarity search
upsert(source, content, vector)	Inserts via supabase.from().upsert() with dedup on source,content
clear()	Deletes all rows from the documents table
disconnect()	No-op (Supabase JS client manages connections automatically)

Database Setup

Run the following in your Supabase SQL editor:

-- 1. Enable pgvector
CREATE EXTENSION IF NOT EXISTS vector;

-- 2. Create the documents table
CREATE TABLE documents (
  id BIGSERIAL PRIMARY KEY,
  source TEXT NOT NULL,
  content TEXT NOT NULL,
  vector VECTOR(3072),
  UNIQUE(source, content)
);

-- 3. Create the similarity search function
CREATE OR REPLACE FUNCTION match_documents(
  query_embedding VECTOR(3072),
  match_count INT DEFAULT 5
)
RETURNS TABLE (
  source TEXT,
  content TEXT,
  similarity FLOAT
)
LANGUAGE plpgsql
AS $$
BEGIN
  RETURN QUERY
  SELECT
    d.source,
    d.content,
    1 - (d.vector <=> query_embedding) AS similarity
  FROM documents d
  ORDER BY d.vector <=> query_embedding
  LIMIT match_count;
END;
$$;

-- 4. Create an index for faster search
CREATE INDEX ON documents
  USING ivfflat (vector vector_cosine_ops)
  WITH (lists = 100);

Adjust VECTOR(3072) to match your embedding model's dimensions.

License

MIT