> ## Documentation Index
> Fetch the complete documentation index at: https://fluxcrud.mahimai.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Concurrency

> Scaling throughput with Batcher and ParallelExecutor

## The Batcher

When inserting thousands of records, individual `await repo.create()` calls are slow due to network latency.

The `Batcher` queues items and flushes them in chunks, reducing network round-trips drastically.

```python theme={null}
from fluxcrud.async_patterns import Batcher

# Usage via context manager on Repository
async with repo.batch_writer(batch_size=100) as writer:
    for i in range(5000):
        # Adds to buffer, auto-flushes at 100
        await writer.add({"name": f"Item {i}"})
```

## ParallelExecutor

Processing many independent tasks (like calling external APIs or multiple distinct DB selects) sequentially is inefficient.

`ParallelExecutor` runs them concurrently but with a **limit** (Semaphore) to avoid crashing your DB or exhausting connections.

```python theme={null}
from fluxcrud.async_patterns import ParallelExecutor

tasks = [lambda: api.fetch_data(i) for i in ids]

# Run at most 50 at once
results = await ParallelExecutor.gather_limited(limit=50, tasks=tasks)
```