trading-common 0.2.0

Common utils for trading platform (Kafka/Postgres/outbox/inbox)

Trading Common

Async utilities shared across the trading platform for working with PostgreSQL inbox/outbox tables, Kafka producers/consumers, and schema validation via trading-contracts.

Components

• trading_common.db.DB — connection pool with inbox/outbox bootstrap, idempotent helpers, and convenience wrappers for writing to the outbox.
• trading_common.kafka.Producer — resilient aiokafka producer prepared for SASL/SSL clusters and controlled retry behaviour.
• trading_common.consumer_app.ConsumerApp — batching Kafka consumer that wraps message handling in PostgreSQL transactions with inbox-based deduplication.
• trading_common.outbox.OutboxProcessor — helpers for reading, acknowledging, and vacuuming the outbox table.
• trading_common.schema.ensure — thin wrapper around trading-contracts JSON schema validation.
• trading_common.settings_kafka — baseline Kafka client settings hydrated from environment variables.

Installation

# Inside a virtualenv; install runtime dependencies only
pip install -e .

# Install with development tooling (pytest, black, mypy, ...)
pip install -e ".[dev]"

Configuration

PostgreSQL

Provide a DSN string understood by asyncpg. The DB.start() method creates (or verifies) the core.inbox and core.outbox tables so that services can boot against an empty database.

from trading_common.db import DB

db = DB("postgresql://postgres:postgres@127.0.0.1:55432/trading")
await db.start()

Kafka

Baseline Kafka options come from trading_common.settings_kafka. They honour the following environment variables (defaults in parentheses):

Variable	Purpose
KAFKA_BOOTSTRAP_SERVERS (localhost:9092)	Broker connection string
KAFKA_SECURITY_PROTOCOL (SASL_SSL)	PLAINTEXT, SASL_SSL, ...
KAFKA_SASL_MECHANISM (PLAIN)	Auth mechanism
KAFKA_SASL_USERNAME / KAFKA_SASL_PASSWORD (empty)	SASL credentials
KAFKA_CLIENT_ID (market-data-service)	Default client id
KAFKA_ENABLE_IDEMPOTENCE (true)	Producer idempotence flag
KAFKA_LINGER_MS, KAFKA_BATCH_SIZE, KAFKA_COMPRESSION_TYPE, ...	Producer tuning knobs
KAFKA_ENABLE_AUTO_COMMIT (false)	Consumer offset mode
KAFKA_AUTO_OFFSET_RESET, KAFKA_MAX_POLL_INTERVAL_MS, ...	Consumer runtime tuning

Use dict(...) to clone and tweak the defaults before passing them into the Kafka helpers so that you do not mutate the shared module-level dictionaries:

from trading_common import settings_kafka

producer = Producer(
    base_cfg={**settings_kafka.KAFKA_COMMON, "client_id": "strategy-service"},
    tuning=settings_kafka.PRODUCER_TUNING,
)

Usage Examples

Producer

import asyncio
from trading_common import settings_kafka
from trading_common.kafka import Producer

async def publish() -> None:
    producer = Producer(
        base_cfg={**settings_kafka.KAFKA_COMMON, "client_id": "md-publisher"},
        tuning=settings_kafka.PRODUCER_TUNING,
    )
    await producer.start()
    try:
        await producer.send(
            topic="md.candles",
            key="BTCUSDT",
            payload={"event_id": "...", "open": 52000, "close": 52120},
        )
    finally:
        await producer.stop()

asyncio.run(publish())

Consumer with Inbox/Outbox

import asyncio
from trading_common import settings_kafka
from trading_common.consumer_app import ConsumerApp
from trading_common.db import DB
from trading_common.schema import ensure

async def handle(con, topic, key, payload):
    ensure("md.candle.closed@v1", payload)
    # Business logic ...

async def main() -> None:
    db = DB("postgresql://postgres:postgres@127.0.0.1:55432/trading")
    consumer = ConsumerApp(
        name="market-data-service",
        db=db,
        base_cfg={**settings_kafka.KAFKA_COMMON, "client_id": "market-data-service"},
        tuning=settings_kafka.CONSUMER_TUNING,
        topics=["md.candles"],
        group_id="md-service",
        handler=handle,
    )
    await consumer.start()
    try:
        await consumer.run()
    finally:
        await consumer.stop()

asyncio.run(main())

Outbox Processing

from trading_common.outbox import OutboxProcessor

processor = OutboxProcessor(db)
async with db.pool.acquire() as con:  # pool is available after db.start()
    events = await processor.get_pending_events(con, limit=100)
    for event_id, topic, key, payload in events:
        await producer.send(topic, key, payload)
        await processor.mark_published(con, event_id)
    await processor.cleanup_old_events(con, days_old=7)

Development Workflow

pytest -m "not slow"       # Run fast tests
black src tests             # Format
isort src tests             # Import sorting
mypy src                    # Static typing
pytest --cov=trading_common --cov-report=term-missing

Testing

• pytest — run the whole suite (asyncio tests use strict mode).
• pytest -m "not slow" — focus on the fast checks used in CI.
• pytest --cov=trading_common --cov-report=term-missing — optional coverage report.

These commands are mirrored in .github/workflows/test.yml, so keeping them green locally guarantees the CI job passes.

Related Projects

• trading-contracts — JSON schemas for every event type; install alongside this package to validate messages with schema.ensure.
• infra/ in the monorepo provides Kafka/PostgreSQL containers for local development (docker compose -f infra/docker-compose.yml up -d).

License

MIT License. See LICENSE for details.