galaxdb-client 0.3.0

GalaxDB Python client -- SQL, vector search, and local embeddings in one database

galaxdb-client

The Python client for GalaxDB -- an AI-native database that combines SQL, vector search, and local embeddings in a single binary.

No external API keys. No separate vector database. No data pipeline. One connection string.

Installation

pip install galaxdb-client

Requires Python 3.9+. Pre-built wheels for Linux x86_64, macOS Intel, macOS Apple Silicon, and Windows x86_64.

What GalaxDB gives you

• Full SQL -- CREATE, INSERT, UPDATE, DELETE, SELECT with WHERE filters
• Local embeddings -- text to vector conversion runs inside the process, no API key needed
• Semantic search -- SEMANTIC_MATCH(col, 'query', threshold) in any WHERE clause
• HNSW vector index -- recall@10 = 0.990 on SIFT-1M at ef=200
• Time-travel queries -- SELECT ... AT VERSION 'tag' to query historical snapshots
• Training export -- CREATE VERSION TAG ... FOR TRAINING exports a Lance dataset, zero-copy PyTorch-ready
• Near-dedup -- WHERE NOT DUPLICATE removes near-duplicate rows using MinHash LSH
• Crash safety -- WAL + checksum, 7 chaos scenarios pass in under 11 seconds
• Encryption at rest -- AES-256-GCM on every block and WAL record

Quick start -- embedded mode (no server)

import galaxdb

# Open or create a database at a local path
db = galaxdb.Database("/tmp/mydb")

# Create a table
db.execute("CREATE TABLE products (id INT PRIMARY KEY, name TEXT, price INT)")

# Insert rows
db.execute("INSERT INTO products (id, name, price) VALUES (1, 'Laptop', 1200)")
db.execute("INSERT INTO products (id, name, price) VALUES (2, 'Headphones', 150)")
db.execute("INSERT INTO products (id, name, price) VALUES (3, 'Keyboard', 80)")

# Query with filter
rows = db.execute("SELECT * FROM products WHERE price > 100")
for row in rows:
    print(row)
# {'id': '1', 'name': 'Laptop', 'price': '1200'}
# {'id': '2', 'name': 'Headphones', 'price': '150'}

# Update
db.execute("UPDATE products SET price = 1100 WHERE id = 1")

# Delete
db.execute("DELETE FROM products WHERE id = 3")

# Table info
print(db.table_exists("products"))  # True
print(db.table_count)               # 1

Semantic search with local embeddings

Start the server with the embedding sidecar to enable SEMANTIC_MATCH:

galaxdb-server \
  --data-dir ./data \
  --port 5433 \
  --sidecar /usr/local/bin/galaxdb-sidecar \
  --model sentence-transformers/all-MiniLM-L6-v2

Then connect from Python:

import galaxdb

conn = galaxdb.connect("host=localhost port=5433 dbname=galaxdb sslmode=disable")

# Create a table with an embedding column
conn.execute("""
    CREATE TABLE docs (
        id   INT PRIMARY KEY,
        body TEXT EMBEDDING MODEL 'sentence-transformers/all-MiniLM-L6-v2' DIM 384
    )
""")

# Insert rows -- embeddings are computed automatically by the local sidecar
conn.execute("INSERT INTO docs (id, body) VALUES (1, 'machine learning and neural networks')")
conn.execute("INSERT INTO docs (id, body) VALUES (2, 'rust programming language systems')")
conn.execute("INSERT INTO docs (id, body) VALUES (3, 'cooking recipes italian pasta')")
conn.execute("INSERT INTO docs (id, body) VALUES (4, 'deep learning transformers attention')")

# Semantic search -- no external API, no separate vector DB
rows = conn.execute(
    "SELECT id, body FROM docs WHERE SEMANTIC_MATCH(body, 'artificial intelligence', 0.4)"
)
for row in rows:
    print(row)
# Returns rows 1 and 4 -- the AI/ML related documents

conn.close()

Time-travel queries

# Create a named snapshot
conn.execute("CREATE VERSION TAG 'v1' FOR TRAINING WITH TRAINING PRECISION 'float32'")

# Insert more data after the snapshot
conn.execute("INSERT INTO docs (id, body) VALUES (5, 'new document added later')")

# Query the snapshot -- only sees data from before the tag
rows = conn.execute("SELECT * FROM docs AT VERSION 'v1'")
# Returns rows 1-4, not row 5

Training export

import galaxdb
import lance
import torch

db = galaxdb.Database("./data")

# Create a training snapshot
db.execute("CREATE VERSION TAG 'train-v1' FOR TRAINING WITH TRAINING PRECISION 'float32'")

# Export as a Lance dataset
path = db.training_dataset("train-v1")

# Load into PyTorch -- zero-copy, memory-mapped
dataset = lance.dataset(path).to_pytorch()
loader = torch.utils.data.DataLoader(dataset, batch_size=32)

Bulk insert

conn.execute("""
    BULK INSERT INTO products (id, name, price) VALUES
      (10, 'Monitor', 400),
      (11, 'Mouse', 30),
      (12, 'Webcam', 90)
""")

Near-duplicate deduplication

# Select only unique documents (one per near-duplicate cluster)
rows = conn.execute("SELECT * FROM docs WHERE NOT DUPLICATE")

Backup and restore

conn.execute("BACKUP TO '/path/to/backup'")
conn.execute("RESTORE FROM '/path/to/backup'")

Server mode -- connect to a running GalaxDB server

import galaxdb

# Connect using a PostgreSQL-style connection string
conn = galaxdb.connect("host=localhost port=5433 dbname=galaxdb sslmode=disable")

conn.execute("CREATE TABLE users (id INT PRIMARY KEY, name TEXT, age INT)")
conn.execute("INSERT INTO users (id, name, age) VALUES (1, 'Alice', 30)")

rows = conn.execute("SELECT * FROM users WHERE age > 25")
for row in rows:
    print(row)

conn.close()

Any PostgreSQL client works -- psycopg2, SQLAlchemy, tokio-postgres, pg (Node.js), JDBC.

Docker

docker run -d -p 5433:5433 -p 9090:9090 \
  -v /data:/data \
  -v ~/.cache/huggingface:/root/.cache/huggingface \
  harbi256/galaxdb:latest \
  --data-dir /data \
  --sidecar /usr/local/bin/galaxdb-sidecar \
  --model sentence-transformers/all-MiniLM-L6-v2

Observability

# Health check
curl http://localhost:9090/health
# {"status":"ok","version":"1.0.0-beta.1","subsystems":{"sidecar_healthy":true}}

# Prometheus metrics
curl http://localhost:9090/metrics

Links

• GitHub
• Getting Started
• Benchmarks
• Docker Hub

License

Apache 2.0