pgvector

# pgvector Skill

PostgreSQL + pgvector extension for vector similarity search.

## Quick Connect

```bash
# Connect to pgvector database (default port 5433)
psql -h localhost -p 5433 -U damien -d postgres

# Or use environment variables
export PGHOST=localhost
export PGPORT=5433
export PGUSER=damien
export PGPASSWORD=''
export PGDATABASE=postgres
```

## Environment

- **Host**: localhost
- **Port**: 5433
- **User**: damien
- **Password**: (empty)
- **Database**: postgres

## Core Capabilities

### 1. Create Vector Table

```sql
-- Basic vector table (1536 dimensions for OpenAI embeddings)
CREATE TABLE IF NOT EXISTS documents (
    id BIGSERIAL PRIMARY KEY,
    content TEXT NOT NULL,
    embedding vector(1536) NOT NULL,
    metadata JSONB,
    created_at TIMESTAMPTZ DEFAULT NOW()
);

-- Create HNSW index for fast similarity search
CREATE INDEX ON documents USING hnsw (embedding vector_cosine_ops)
WITH (m = 16, ef_construction = 64);

-- Or use IVFFlat index (faster build, slower search)
CREATE INDEX ON documents USING ivfflat (embedding vector_cosine_ops)
WITH (lists = 100);
```

### 2. Insert Embeddings

```sql
-- Manual insert (replace with actual embedding)
INSERT INTO documents (content, embedding)
VALUES ('Your text here', '[0.1, 0.2, ..., 0.1536]');

-- With metadata
INSERT INTO documents (content, embedding, metadata)
VALUES (
    'AI is transforming technology',
    '[0.1, 0.3, ..., 0.5]',
    '{"source": "article", "author": "John"}'::jsonb
);
```

### 3. Vector Similarity Search

```sql
-- Cosine similarity (most common)
SELECT id, content, (1 - (embedding <=> '[query_embedding]')) AS similarity
FROM documents
ORDER BY embedding <=> '[query_embedding]'
LIMIT 5;

-- Euclidean distance
SELECT id, content, (embedding <-> '[query_embedding]') AS distance
FROM documents
ORDER BY embedding <-> '[query_embedding]'
LIMIT 5;

-- Inner product (for normalized vectors)
SELECT id, content, (embedding <#> '[query_embedding]') AS similarity
FROM documents
ORDER BY embe...