Add environment variable for configurable model storage path

[titusz]

Problem

The model storage path is hardcoded to use platformdirs.user_data_dir, which is problematic for:

• Container deployments (ephemeral filesystems)
• CI/CD environments (need to cache in specific locations)
• Multi-user systems (shared model cache)
• Custom deployment scenarios

Current Behavior

from platformdirs import PlatformDirs

dirs = PlatformDirs(appname=APP_NAME, appauthor=APP_AUTHOR, ensure_exists=True)
MODEL_PATH = Path(dirs.user_data_dir) / MODEL_FILENAME

The path cannot be overridden without modifying the source code.

Expected Behavior

Allow users to specify a custom model path via environment variable:

import os
from pathlib import Path
from platformdirs import PlatformDirs

# Allow override via environment variable
model_dir = os.getenv("ISCC_SCT_MODEL_DIR")
if model_dir:
    MODEL_PATH = Path(model_dir) / MODEL_FILENAME
else:
    dirs = PlatformDirs(appname=APP_NAME, appauthor=APP_AUTHOR, ensure_exists=True)
    MODEL_PATH = Path(dirs.user_data_dir) / MODEL_FILENAME

# Ensure directory exists
MODEL_PATH.parent.mkdir(parents=True, exist_ok=True)

Use Cases

Container Deployments

# Download model to /app/models during build
ENV ISCC_SCT_MODEL_DIR=/app/models
RUN python -c "import iscc_sct.utils; iscc_sct.utils.get_model()"

CI/CD Caching

- name: Cache iscc-sct model
  uses: actions/cache@v4
  with:
    path: .cache/iscc-sct
    key: iscc-sct-model-v1

- name: Run tests
  env:
    ISCC_SCT_MODEL_DIR: .cache/iscc-sct
  run: pytest

Shared Model Cache

# Multiple users on the same system share one model
export ISCC_SCT_MODEL_DIR=/opt/shared/iscc-sct
python my_app.py

Benefits

• Zero breaking changes (environment variable is optional)
• Follows common practice (HuggingFace uses HF_HOME, PyTorch uses TORCH_HOME, etc.)
• Enables efficient container deployments
• Simplifies CI/CD caching strategies
• Supports shared model caches

Implementation Notes

Also consider supporting ISCC_SCT_TOKENIZER_DIR for the tokenizer model, following the same pattern.