rhei-engine 2.0.0

Lightweight serverless HTAP engine — Python bindings for the Rhei Rust library

rhei-engine

Python bindings for the Rhei HTAP engine -- lightweight, serverless Hybrid Transactional/Analytical Processing.

All engine methods are async and compatible with asyncio and anyio.

Install

pip install rhei-engine
# or
uv add rhei-engine

Note: the package is published as rhei-engine on PyPI (the rhei name was already taken), but the import name is rhei:

import rhei  # not "import rhei_engine"

Build from source

cd python
uv sync
uv run maturin develop

Quick start

import anyio
import rhei

async def main():
    async with await rhei.open(oltp_path="my.db") as engine:
        # Create table and register for CDC replication
        await engine.execute("CREATE TABLE users (id INTEGER PRIMARY KEY, name TEXT, age INTEGER)")
        await engine.register_table(
            rhei.TableSchema("users", [("id", "int64"), ("name", "utf8"), ("age", "int64")], ["id"])
        )

        # Write (routed to OLTP)
        await engine.execute("INSERT INTO users VALUES (1, 'Alice', 30)")

        # Batch write (single transaction, ~35K rows/sec)
        await engine.execute_batch([
            "INSERT INTO users VALUES (2, 'Bob', 25)",
            "INSERT INTO users VALUES (3, 'Charlie', 35)",
        ])

        # Sync CDC to OLAP
        result = await engine.sync_now()
        print(result)  # SyncResult(events=3, inserted=3, ...)

        # Query (auto-routed: aggregates -> OLAP, point queries -> OLTP)
        batches = await engine.query("SELECT AVG(age) FROM users")  # list[pyarrow.RecordBatch]
        print(batches[0].to_pydict())

anyio.run(main)

API

Method	Description
rhei.open(**kwargs)	Async factory (non-blocking init)
HtapEngine(config)	Sync constructor
engine.execute(sql)	Write statement (OLTP)
engine.execute_batch(stmts)	Batch writes in single transaction
engine.query(sql)	Auto-routed query -> list[pyarrow.RecordBatch]
engine.query_olap(sql)	Force OLAP query
engine.sync_now()	Sync CDC events to OLAP
engine.register_table(schema)	Register table for replication
engine.add_column(t, c, type)	Add column + rebuild CDC triggers
engine.drop_column(t, c)	Drop column
engine.initial_sync(table)	Bulk OLTP -> OLAP load
engine.shutdown()	Cleanup (resources released on GC)

License

Apache-2.0