# Configuration Management

The configuration management module provides classes and utilities for managing RAPiD feature extraction settings.

## RAPiDConfig

The main configuration class that holds all parameters for RAPiD computation.

### Parameters

- `k_mid`: Number of neighbors for mid-range processing (default: 8)
- `device`: Processing device ("cpu", "cuda", or "auto")
- `batch_size`: Batch size for processing (default: 32)
- `precision`: Floating point precision ("float16", "float32")

### Example

```python
from rapid_seg.config import RAPiDConfig

# Create custom configuration
config = RAPiDConfig(
    k_mid=12,
    device="cuda",
    batch_size=64,
    precision="float32"
)

print(f"K neighbors: {config.k_mid}")
print(f"Device: {config.device}")
```

## Predefined Configurations

Use the `create_config` function to get predefined configurations:

```python
from rapid_seg.config import create_config

# Balanced configuration (default)
balanced_config = create_config("balanced")

# Fast processing configuration
fast_config = create_config("fast")

# High accuracy configuration
accurate_config = create_config("accurate")
```

### Configuration Presets

| Preset | k_mid | Batch Size | Description |
|--------|-------|------------|-------------|
| `fast` | 4 | 64 | Optimized for speed |
| `balanced` | 8 | 32 | Balanced performance |
| `accurate` | 16 | 16 | High accuracy mode |

## Dynamic Configuration

The configuration system supports dynamic parameter adjustment:

```python
# Get k value for standard RAPiD
k = config.get_k_for_standard_rapid()

# Get k values for range-aware processing
k_close, k_mid, k_far = config.get_k_for_range_aware_rapid()

# Adjust batch size based on available memory
config.optimize_batch_size(available_memory_gb=8.0)
```

## Device Management

Automatic device selection based on system capabilities:

```python
# Auto-detect best available device
config = RAPiDConfig(device="auto")

# Force specific device
config = RAPiDConfig(device="cuda:0")

# Check device availability
if config.is_cuda_available():
    print("CUDA is available")
```