Spaces:
Paused
Paused
| """Pull the private runtime data into HOUND_DATA_DIR at Space startup. | |
| The Space CODE is public; the private ``graph.json`` / ``rooms.json`` / | |
| ``index.npz`` live in a PRIVATE Hugging Face dataset repo (design.md §6). At | |
| startup, when ``HOUND_DATA_REPO`` is set and ``HF_TOKEN`` is present, this | |
| downloads those three files into ``HOUND_DATA_DIR`` before uvicorn boots the | |
| app (api/main.py reads graph/rooms, ml/backend reads index.npz). | |
| No token or private file is ever baked into the image. If ``HOUND_DATA_REPO`` | |
| is unset (e.g. a local ``docker run`` without secrets) this no-ops gracefully so | |
| the container still boots to the 503 "data not ready" state rather than | |
| crashing. The same graceful no-op applies when ``HF_TOKEN`` is missing or the | |
| download fails: log and continue, never raise. | |
| """ | |
| from __future__ import annotations | |
| import os | |
| import sys | |
| from pathlib import Path | |
| # The three private runtime files the app loads from HOUND_DATA_DIR. | |
| DATA_FILES = ('graph.json', 'rooms.json', 'index.npz') | |
| OPTIONAL_DATA_FILES = ('recognition.json',) | |
| def log(message: str) -> None: | |
| print(f'[fetch_data] {message}', flush=True) | |
| def main() -> int: | |
| repo_id = os.environ.get('HOUND_DATA_REPO') | |
| data_dir = Path(os.environ.get('HOUND_DATA_DIR', 'data')) | |
| if not repo_id: | |
| log('HOUND_DATA_REPO not set, skipping private-data download ' | |
| '(app will report 503 until data is present).') | |
| return 0 | |
| token = os.environ.get('HF_TOKEN') | |
| if not token: | |
| log('HOUND_DATA_REPO is set but HF_TOKEN is missing, skipping download ' | |
| '(app will report 503 until data is present).') | |
| return 0 | |
| try: | |
| from huggingface_hub import hf_hub_download # lazy: runtime dep only. | |
| except Exception as exc: # pragma: no cover - only hit without the dep. | |
| log(f'huggingface_hub unavailable ({exc!r}), skipping download.') | |
| return 0 | |
| data_dir.mkdir(parents=True, exist_ok=True) | |
| for filename in DATA_FILES + OPTIONAL_DATA_FILES: | |
| try: | |
| path = hf_hub_download( | |
| repo_id=repo_id, | |
| filename=filename, | |
| repo_type='dataset', | |
| revision=os.environ.get('HOUND_DATA_REVISION') or None, | |
| token=token, | |
| local_dir=str(data_dir), | |
| ) | |
| log(f'downloaded {filename} -> {path}') | |
| except Exception as exc: | |
| # Never crash startup over one missing/failed file; the app's own | |
| # readiness check will surface 503 if required data is absent. | |
| log(f'could not download {filename} ({exc!r}), continuing.') | |
| return 0 | |
| if __name__ == '__main__': | |
| sys.exit(main()) | |