queue-ti-client 2026.5.2

Python client for the queue-ti message queue service

queue-ti Python Client

A Python client library for queue-ti, providing high-level Producer and Consumer APIs for async and sync applications.

• Async first — Native async/await with automatic reconnect and token refresh
• Sync wrapper — Drop-in synchronous API that runs async code on a background thread
• Type-safe — Full type hints and mypy-strict compatible
• Python 3.11+ — Requires Python 3.11 or later

Installation

pip install queue-ti-client

Quick Start

Async Producer

import asyncio
from queueti import connect, ConnectOptions

async def main():
    # Connect to the server
    client = await connect("localhost:50051", options=ConnectOptions(insecure=True))
    producer = client.producer()
    
    # Publish a message
    msg_id = await producer.publish(
        topic="orders",
        payload=b'{"amount": 99.99}',
    )
    print(f"Published: {msg_id}")
    
    await client.close()

asyncio.run(main())

Async Consumer

import asyncio
from queueti import connect, ConnectOptions, ConsumerOptions

async def main():
    client = await connect("localhost:50051", options=ConnectOptions(insecure=True))
    
    # Consume messages (blocks until cancelled)
    consumer = client.consumer(
        topic="orders",
        options=ConsumerOptions(concurrency=4),
    )
    
    async def handler(msg):
        print(f"[{msg.id}] {msg.payload}")
        # Return normally to auto-ack; raise to auto-nack
    
    await consumer.consume(handler)

asyncio.run(main())

Sync Producer

from queueti import connect_sync, ConnectOptions

client = connect_sync("localhost:50051", options=ConnectOptions(insecure=True))
producer = client.producer()

msg_id = producer.publish(
    topic="orders",
    payload=b'{"amount": 99.99}',
)
print(f"Published: {msg_id}")

client.close()

Sync Consumer

from queueti import connect_sync, ConnectOptions, ConsumerOptions

client = connect_sync("localhost:50051", options=ConnectOptions(insecure=True))

consumer = client.consumer(
    topic="orders",
    options=ConsumerOptions(concurrency=4),
)

def handler(msg):
    print(f"[{msg.id}] {msg.payload}")
    # Return normally to auto-ack; raise to auto-nack

# Blocks until interrupted (Ctrl+C)
consumer.consume(handler)

Connection

Basic Connection

from queueti import connect

client = await connect("localhost:50051")

Insecure (Development)

from queueti import connect, ConnectOptions

client = await connect(
    "localhost:50051",
    options=ConnectOptions(insecure=True),
)

With Bearer Token

from queueti import connect, ConnectOptions

client = await connect(
    "localhost:50051",
    options=ConnectOptions(token="your-jwt-token"),
)

With Token Refresh

When your token expires, you can provide a refresher function to obtain a new token automatically:

from queueti import connect, ConnectOptions

async def refresh_token() -> str:
    # Fetch a new token (e.g., from your auth service)
    new_token = await fetch_fresh_token()
    return new_token

client = await connect(
    "localhost:50051",
    options=ConnectOptions(
        token="initial-token",
        token_refresher=refresh_token,
    ),
)

# Token will refresh automatically before expiry

You can also manually set a new token:

client.set_token("new-token")

ConnectOptions

All fields are optional.

Field	Type	Description
token	str | None	Bearer token for JWT authentication
token_refresher	Callable[[], Awaitable[str]] | None	Async function to refresh the token before expiry
insecure	bool	Disable TLS (for development only; default: False)

Authentication

When the server has auth.enabled = true, every RPC call requires a valid JWT. Tokens are issued by the server's HTTP API and expire after 15 minutes.

Using QueueTiAuth (recommended)

The QueueTiAuth helper automatically checks if authentication is required and handles login and token refresh:

import asyncio
import queueti

auth = queueti.QueueTiAuth.login("http://localhost:8080", "admin", "secret")

async def main():
    opts = queueti.ConnectOptions(
        token=auth.token,
        token_refresher=auth.async_refresh,
    )
    client = await queueti.connect("localhost:50051", opts)
    try:
        producer = client.producer()
        msg_id = await producer.publish("orders", b"...")
        print(f"Published: {msg_id}")
    finally:
        await client.close()

    async with queueti.AsyncAdminClient(
        "http://localhost:8080",
        queueti.AdminOptions(token=auth.token),
    ) as admin:
        configs = await admin.list_topic_configs()

asyncio.run(main())

For the synchronous client, use connect_sync and refresh() (the sync variant of async_refresh()):

import queueti

auth = queueti.QueueTiAuth.login("http://localhost:8080", "admin", "secret")

client = queueti.connect_sync("localhost:50051", queueti.ConnectOptions(
    token=auth.token,
    token_refresher=auth.async_refresh,
))
try:
    producer = client.producer()
    msg_id = producer.publish("orders", b"...")
    print(f"Published: {msg_id}")
finally:
    client.close()

The QueueTiAuth helper:

1. Calls GET /api/auth/status to check if authentication is required
2. If auth is disabled, returns a no-op instance with a null token
3. If auth is enabled, calls POST /api/auth/login with the provided credentials
4. Exposes .token (str or None) for the current JWT, .async_refresh() for async clients, and .refresh() for sync clients

Option 1 — Obtaining a token manually

TOKEN=$(curl -s -X POST http://localhost:8080/api/auth/login \
  -H "Content-Type: application/json" \
  -d '{"username":"admin","password":"secret"}' \
  | jq -r '.token')

Option 2 — Automatic refresh with custom fetcher

When your token expires, you can provide a refresher function to obtain a new token automatically:

from queueti import connect, ConnectOptions

async def refresh_token() -> str:
    # Fetch a new token (e.g., from your auth service)
    new_token = await fetch_fresh_token()
    return new_token

client = await connect(
    "localhost:50051",
    options=ConnectOptions(
        token="initial-token",
        token_refresher=refresh_token,
    ),
)

# Token will refresh automatically before expiry

Option 3 — Manual update

Call client.set_token() to swap the token on a live connection. The new token takes effect on the very next RPC call; no reconnection is needed.

client = await connect(
    "localhost:50051",
    options=ConnectOptions(token="initial-token"),
)

# Later, when you have a fresh token:
client.set_token("new-token")

This is useful when token lifecycle is managed externally (e.g. a shared token store, a sidecar, or an existing refresh loop in your application).

Option 4 — Static token (short-lived processes)

For scripts or batch jobs that complete within the 15-minute token window, a static token is sufficient:

from queueti import connect, ConnectOptions
import os

client = await connect(
    "localhost:50051",
    options=ConnectOptions(token=os.getenv("QUEUETI_TOKEN")),
)

Producer API

AsyncProducer.publish()

msg_id: str = await producer.publish(
    topic: str,
    payload: bytes,
    options: PublishOptions | None = None,
) -> str

Publish a message and return its assigned ID.

Parameters:

• topic (str) — Topic name
• payload (bytes) — Message content
• options (PublishOptions | None) — Optional metadata and publishing settings

Returns: Message ID (str)

Raises: PublishError if the RPC fails

Example:

msg_id = await producer.publish(
    topic="orders",
    payload=b'{"sku": "ABC123", "qty": 5}',
    options=PublishOptions(metadata={"source": "api"}),
)

Producer.publish() (Sync)

Identical to AsyncProducer.publish() but blocks instead of awaiting.

PublishOptions

Field	Type	Description
metadata	dict[str, str]	Optional metadata key-value pairs (default: {})

Consumer API

AsyncConsumer.consume()

await consumer.consume(
    handler: Callable[[Message], Awaitable[None]],
) -> None

Stream messages from the topic, calling the handler for each message. Runs until cancelled. Auto-acks on success; auto-nacks on exception.

Parameters:

• handler — Async function called for each message. Raise an exception to nack.

Behavior:

• Reconnects with exponential backoff on stream errors
• Concurrency controlled via ConsumerOptions.concurrency
• Visibility timeout overridable per-call via ConsumerOptions.visibility_timeout_seconds

Example:

async def process_order(msg: Message):
    order = json.loads(msg.payload)
    print(f"Processing order {order['id']} (retry #{msg.retry_count})")
    if order["amount"] < 0:
        raise ValueError("invalid amount")

await consumer.consume(process_order)

AsyncConsumer.consume_batch()

await consumer.consume_batch(
    options: BatchOptions,
    handler: Callable[[list[Message]], Awaitable[None]],
) -> None

Poll batches from the topic, calling the handler with all messages in the batch. Runs until cancelled. Handler is responsible for acking/nacking each message.

Parameters:

• options (BatchOptions) — Batch size and optional visibility timeout override
• handler — Async function called with a list of Message objects

Behavior:

• Polls with exponential backoff when the queue is empty
• Each message is individually locked and can be acked/nacked independently
• Handler errors do not prevent ack/nack of individual messages

Example:

from queueti import BatchOptions

async def handle_batch(messages: list[Message]):
    for msg in messages:
        try:
            order = json.loads(msg.payload)
            await process_order(order)
            await msg.ack()
        except Exception as e:
            await msg.nack(f"processing failed: {e}")

await consumer.consume_batch(
    options=BatchOptions(batch_size=10, visibility_timeout_seconds=60),
    handler=handle_batch,
)

Consumer.consume() (Sync)

Blocks on the calling thread and processes messages one at a time. Identical behavior to async version.

def handler(msg: SyncMessage):
    # Process message; raise to nack
    pass

consumer.consume(handler)  # Blocks until interrupted

Consumer.consume_batch() (Sync)

Blocks on the calling thread and processes message batches.

def handler(messages: list[SyncMessage]):
    for msg in messages:
        try:
            # Process...
            msg.ack()
        except Exception as e:
            msg.nack(f"error: {e}")

consumer.consume_batch(
    options=BatchOptions(batch_size=10),
    handler=handler,
)

ConsumerOptions

All fields are optional.

Field	Type	Default	Description
concurrency	int	1	Number of messages to process in parallel (must be ≥ 1)
visibility_timeout_seconds	int | None	None	Override server default visibility timeout for this consumer (seconds)
consumer_group	str | None	None	Consumer group name for independent message consumption; see Consumer Groups

BatchOptions

Field	Type	Description
batch_size	int	Maximum messages to dequeue in one call (must be ≥ 1)
visibility_timeout_seconds	int | None	Optional visibility timeout override (seconds)
consumer_group	str | None	Consumer group name for independent message consumption

Message

Fields

Received from consume() or consume_batch().

Field	Type	Description
id	str	Unique message identifier
topic	str	Topic the message belongs to
payload	bytes	Message content
metadata	dict[str, str]	User-supplied metadata
created_at	datetime	Enqueue timestamp (UTC)
retry_count	int	Current retry count (0 = first attempt)

Methods

await msg.ack() -> None — Acknowledge the message (removes it from the queue).

Raises AckError if the RPC fails.

await msg.nack(reason: str = "") -> None — Return the message to the queue for retry (or to DLQ if max retries exceeded).

Raises NackError if the RPC fails.

Note: When using consume(), ack/nack are called automatically. Only call them directly with consume_batch().

SyncMessage

Identical to Message, but with synchronous ack() and nack() methods. Used with Consumer.consume() and Consumer.consume_batch().

Consumer Groups

Consumer groups enable independent consumption of the same messages by multiple systems. Each group tracks its own delivery state, allowing parallel processing of the same message by different applications without interference.

When a consumer group is specified, the client sends all RPCs scoped to that group and receives all messages enqueued to the topic. Each message is delivered independently to each group. A message is only deleted from the queue when all registered groups have acknowledged it.

Registering a Consumer Group

Consumer groups must be registered on the server before use:

curl -X POST http://localhost:8080/api/topics/orders/consumer-groups \
  -H "Content-Type: application/json" \
  -d '{"consumer_group": "warehouse"}'

Once registered, the group automatically receives all pending messages enqueued before registration (backfill), plus all future messages.

Async Consumer

import asyncio
from queueti import connect, ConnectOptions, ConsumerOptions

async def main():
    client = await connect("localhost:50051", options=ConnectOptions(insecure=True))
    consumer = client.consumer(
        "orders",
        options=ConsumerOptions(consumer_group="warehouse", concurrency=4),
    )
    
    async def handler(msg):
        print(f"[warehouse] processing {msg.id}")
        # Return normally to Ack; raise to Nack
    
    await consumer.consume(handler)

asyncio.run(main())

Async Batch Consumer

from queueti import BatchOptions

async def handle_batch(messages):
    for msg in messages:
        try:
            # Process...
            await msg.ack()
        except Exception as e:
            await msg.nack(f"error: {e}")

await consumer.consume_batch(
    options=BatchOptions(batch_size=50, consumer_group="warehouse"),
    handler=handle_batch,
)

Sync Consumer

from queueti import connect_sync, ConnectOptions, ConsumerOptions

client = connect_sync("localhost:50051", options=ConnectOptions(insecure=True))
consumer = client.consumer(
    "orders",
    options=ConsumerOptions(consumer_group="warehouse", concurrency=4),
)

def handler(msg):
    print(f"[warehouse] processing {msg.id}")
    # Return normally to Ack; raise to Nack

consumer.consume(handler)  # Blocks until interrupted

Sync Batch Consumer

from queueti import BatchOptions

def handle_batch(messages):
    for msg in messages:
        try:
            # Process...
            msg.ack()
        except Exception as e:
            msg.nack(f"error: {e}")

consumer.consume_batch(
    options=BatchOptions(batch_size=50, consumer_group="warehouse"),
    handler=handle_batch,
)

Error Handling

All exceptions inherit from QueueTiError.

QueueTiError

Base exception for all queue-ti client errors.

from queueti import QueueTiError

try:
    await consumer.consume(handler)
except QueueTiError as e:
    print(f"Queue operation failed: {e}")

PublishError

Raised when a message cannot be published.

from queueti import PublishError

try:
    msg_id = await producer.publish("orders", payload)
except PublishError as e:
    print(f"Failed to publish: {e}")

AckError

Raised when acknowledging a message fails.

from queueti import AckError

try:
    await msg.ack()
except AckError as e:
    print(f"Failed to ack message {msg.id}: {e}")

NackError

Raised when nacking a message fails.

from queueti import NackError

try:
    await msg.nack("processing failed")
except NackError as e:
    print(f"Failed to nack message {msg.id}: {e}")

Examples

Robust Async Consumer with Exponential Backoff

import asyncio
from queueti import connect, ConnectOptions, ConsumerOptions, Message

async def consume_with_backoff():
    client = await connect(
        "localhost:50051",
        options=ConnectOptions(insecure=True),
    )
    
    consumer = client.consumer(
        topic="emails",
        options=ConsumerOptions(concurrency=8),
    )
    
    async def send_email(msg: Message):
        payload = json.loads(msg.payload)
        try:
            await send_smtp(payload["to"], payload["body"])
        except TemporaryFailure:
            raise  # Nack; will retry after visibility timeout
        except PermanentFailure:
            # Don't raise; let it go to DLQ if max retries exceeded
            await msg.nack("permanent failure, skipping")
    
    try:
        await consumer.consume(send_email)
    except KeyboardInterrupt:
        print("Shutting down...")
    finally:
        await client.close()

asyncio.run(consume_with_backoff())

Batch Processing with Manual Ack/Nack

import asyncio
import json
from queueti import connect, ConnectOptions, BatchOptions, Message

async def batch_processor():
    client = await connect("localhost:50051", options=ConnectOptions(insecure=True))
    consumer = client.consumer("events")
    
    async def process_batch(messages: list[Message]):
        # Process all messages; commit to DB once
        rows = []
        for msg in messages:
            event = json.loads(msg.payload)
            rows.append(event)
        
        try:
            async with db_pool.acquire() as conn:
                await conn.executemany(
                    "INSERT INTO events (...) VALUES (...)",
                    rows,
                )
            # Commit succeeded; ack all
            for msg in messages:
                await msg.ack()
        except Exception as e:
            # Commit failed; nack all
            for msg in messages:
                await msg.nack(f"db error: {e}")
    
    await consumer.consume_batch(
        options=BatchOptions(batch_size=100),
        handler=process_batch,
    )

asyncio.run(batch_processor())

Sync Consumer in a Worker Thread

import threading
import json
from queueti import connect_sync, ConnectOptions, SyncMessage

def worker():
    client = connect_sync("localhost:50051", options=ConnectOptions(insecure=True))
    consumer = client.consumer("webhooks")
    
    def handle_webhook(msg: SyncMessage):
        payload = json.loads(msg.payload)
        requests.post(payload["url"], json=payload["data"])
    
    try:
        consumer.consume(handle_webhook)
    finally:
        client.close()

# Run in a separate thread
thread = threading.Thread(target=worker, daemon=True)
thread.start()
thread.join()

Development setup

macOS and some Linux distributions ship an externally-managed Python that blocks pip install at the system level. Use a virtual environment:

# From the repo root — creates .venv and installs all dev dependencies
make setup-python

# Run the test suite
make test-python

Or manually:

cd clients/python
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Testing

With the virtual environment active:

# Run all tests
pytest

# Run specific test file
pytest tests/test_consumer.py

# Run with verbose output
pytest -v

# Run mypy
mypy queueti/

Logging

The library uses Python's standard logging module. To see internal debug logs:

import logging

logging.basicConfig(level=logging.DEBUG)
logger = logging.getLogger("queueti")
logger.setLevel(logging.DEBUG)

Troubleshooting

Connection refused

Ensure the queue-ti server is running on the correct host and port:

# Development (insecure, local)
client = await connect("localhost:50051", options=ConnectOptions(insecure=True))

# Production (TLS required)
client = await connect("queue-ti.example.com:50051")

Token refresh not working

Ensure your token_refresher function returns a valid JWT string and handles errors:

async def refresh_token() -> str:
    try:
        response = await auth_service.refresh()
        return response.token
    except Exception as e:
        logger.error(f"Token refresh failed: {e}")
        raise  # Exponential backoff will apply

Messages not being processed

Check that:

1. Messages are being published (publish() succeeded)
2. Consumer handler is not raising unexpected exceptions
3. Visibility timeout is long enough for your processing (increase via ConsumerOptions.visibility_timeout_seconds)
4. Topic exists and has messages (use admin UI or check logs)

License

MIT