# constants

Python API: `modelconverter.utils.constants`

Well-known directories and registries shared across modelconverter.

The same code runs twice: on the host, where it launches the per-backend Docker image, and inside that image, where the conversion
itself happens. The two see different filesystems -- the container has the cache bind-mounted at `/app/shared_with_container` and
writes its results to `/app/output`, while on the host the cache directory and a local `./output` are used. The constants below
settle that choice once, at import time, so the rest of the package can just refer to them.

## Functions

### get_cache_dir

```python
def get_cache_dir() -> Path:
```

Return the hidden, auto-managed cache directory used for staged inputs and remote downloads.

Respects `XDG_CACHE_HOME` and defaults to `~/.cache/modelconverter`.

### in_docker

```python
def in_docker() -> bool:
```

Whether this process runs inside a modelconverter container.

## Attributes

### CALIBRATION_DIR

### CONFIGS_DIR

### CONTAINER_SHARED_DIR

### CONVERSION_MARKER

### HOST_OUTPUT_DIR_ENV_VAR

### INFERENCE_MARKER

### LOADERS

### MISC_DIR

### MODELS_DIR

### OUTPUTS_DIR

### SHARED_DIR