Concurry¶
Welcome to Concurry — the Python concurrency library designed to be delightful.
Concurry provides a unified, framework-agnostic interface for parallel and asynchronous programming. Whether you're using threading, multiprocessing, asyncio, or Ray, Concurry gives you a single, consistent API that just works.
Concept Map: What do you want to build?¶
| I want to... | Use... | Why? |
|---|---|---|
| Run API calls in parallel | Workers (Thread mode) | I/O bound operations don't block the GIL. |
| Process data (heavy CPU) | Workers (Process mode) | Bypasses the GIL to use all CPU cores. |
| Scale across machines | Workers (Ray mode) | Seamlessly moves your local code to a cluster. |
| Process 1000s of items | Worker Pools | Automatically load-balances tasks across workers. |
| Respect API Rate Limits | Limits | Thread-safe, distributed rate limiting built-in. |
| Handle Flaky APIs | Retries | Automatic exponential backoff and error handling. |
| Wait for multiple results | Synchronization | Powerful wait() and gather() primitives. |
Why Concurry?¶
1. Unified Future Interface¶
Stop writing different code for asyncio.Future, concurrent.futures.Future, and Ray ObjectRef. Concurry wraps them all in a single, consistent Unified Future interface.
2. The Actor Pattern made Simple¶
Define a class, inherit from Worker, and you have a stateful actor that can run anywhere.
from concurry import Worker
class DataProcessor(Worker):
def process(self, data):
return data * 2
# Run efficiently on threads...
worker = DataProcessor.options(mode="thread").init()
# ...or move to a separate process...
worker = DataProcessor.options(mode="process").init()
# ...or scale to a cluster!
worker = DataProcessor.options(mode="ray").init()
3. Production-Ready Features¶
Concurry comes batteries-included with the tools you need for robust systems: * 🚦 Limits: Distributed rate limiting and resource semaphores. * 🔁 Retries: Smart retry logic with jitter and validation. * 📊 Progress: Beautiful, zero-config progress bars.
Where to Start?¶
- 🚀 Getting Started: Build a high-performance LLM pipeline in 5 minutes.
- 📚 User Guide: Deep dive into Workers, Pools, and more.
- 🖼️ Gallery: Real-world examples and recipes.