Supabase Bridge

The Supabase bridge provides vector storage capabilities using pgvector extension through the REST API.

Unlike the Postgres Store, the Supabase Store requires manual setup of the database schema because Supabase doesn't allow arbitrary SQL execution via REST API.

Requirements

Enable pgvector extension in the relevant schema of your Supabase project for using vector column types.
Add columns for embedding (type vector) and metadata (type jsonb) to your table
Pre-configured RPC function for similarity search

See section below for detailed SQL commands.

Database Setup

Execute the following SQL commands in your Supabase SQL Editor:

Enable `pgvector` extension

        1
        CREATE EXTENSION IF NOT EXISTS vector;
    

Create the `documents` table

        1
2
3
4
5
        CREATE TABLE IF NOT EXISTS documents (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    embedding vector(768) NOT NULL,
    metadata JSONB
);
    

Create the similarity search function

        1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
        CREATE OR REPLACE FUNCTION match_documents(
    query_embedding vector(768),
    match_count int DEFAULT 10,
    match_threshold float DEFAULT 0.0
)
RETURNS TABLE (
    id UUID,
    embedding vector,
    metadata JSONB,
    score float
)
LANGUAGE sql
AS $$
    SELECT
        documents.id,
        documents.embedding,
        documents.metadata,
        1- (documents.embedding <=> query_embedding) AS score
    FROM documents
    WHERE 1- (documents.embedding <=> query_embedding) >= match_threshold
    ORDER BY documents.embedding <=> query_embedding ASC
    LIMIT match_count;
$$;
    

Create an index for better performance

        1
2
        CREATE INDEX IF NOT EXISTS documents_embedding_idx
ON documents USING ivfflat (embedding vector_cosine_ops);
    

Configuration

Basic Configuration

        1
2
3
4
5
6
7
8
9
10
11
12
        use Symfony\AI\Store\Bridge\Supabase\Store;
use Symfony\Component\HttpClient\HttpClient;

$store = new Store(
    HttpClient::create(),
    'https://your-project.supabase.co',
    'your-anon-key',
    'documents',        // table name
    'embedding',        // vector field name
    768,               // vector dimension (depending on your embedding model)
    'match_documents'   // function name
);
    

Bundle Configuration

        1
2
3
4
5
6
7
8
9
10
11
        # config/packages/ai.yaml
ai:
    store:
        supabase:
            my_supabase_store:
                url: 'https://your-project.supabase.co'
                api_key: '%env(SUPABASE_API_KEY)%'
                table: 'documents'
                vector_field: 'embedding'
                vector_dimension: 768
                function_name: 'match_documents'
    

Environment Variables

        1
2
3
        # .env.local
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_API_KEY=your-supabase-anon-key
    

Usage

Adding Documents

        1
2
3
4
5
6
7
8
9
10
11
12
        use Symfony\AI\Platform\Vector\Vector;
use Symfony\AI\Store\Document\Metadata;
use Symfony\AI\Store\Document\VectorDocument;
use Symfony\Component\Uid\Uuid;

$document = new VectorDocument(
    Uuid::v4(),
    new Vector([0.1, 0.2, 0.3, /* ... 768 dimensions */]),
    new Metadata(['title' => 'My Document', 'category' => 'example'])
);

$store->add($document);
    

Querying Documents

        1
2
3
4
5
6
7
8
9
10
11
12
        $queryVector = new Vector([0.1, 0.2, 0.3, /* ... 768 dimensions */]);

$results = $store->query($queryVector, [
    'max_items' => 10,
    'min_score' => 0.7
]);

foreach ($results as $document) {
    echo "ID: " . $document->id . "\n";
    echo "Score: " . $document->score . "\n";
    echo "Metadata: " . json_encode($document->metadata->getArrayCopy()) . "\n";
}
    