deploy: add model/download_model.py to TemplateA; Dockerfile runs downloader

Dockerfile

@@ -2,6 +2,10 @@ FROM python:3.9-slim
 
 ARG HF_TOKEN
 ENV HF_TOKEN=${HF_TOKEN}
+ENV MODEL_REPO="TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF"
+ENV MODEL_FILE="TinyLlama-1.1B-Chat-v1.0.Q4_K_M.gguf"
+ENV MODEL_DIR="/tmp/models"
+ENV MODEL_PATH="/tmp/models/${MODEL_FILE}"
 
 # cache-bust on demand
 ARG BUILD_ID=deploy-001

model/download_model.py

@@ -0,0 +1,21 @@
+#!/usr/bin/env python3
+import os, pathlib, sys
+from huggingface_hub import hf_hub_download
+
+MODEL_REPO = os.getenv("MODEL_REPO", "TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF")
+MODEL_FILE = os.getenv("MODEL_FILE", "TinyLlama-1.1B-Chat-v1.0.Q4_K_M.gguf")
+MODEL_DIR  = pathlib.Path(os.getenv("MODEL_DIR", "/tmp/models"))
+MODEL_DIR.mkdir(parents=True, exist_ok=True)
+
+print(f"[download] repo={MODEL_REPO} file={MODEL_FILE} -> {MODEL_DIR}", file=sys.stderr)
+path = hf_hub_download(
+    repo_id=MODEL_REPO,
+    filename=MODEL_FILE,
+    local_dir=str(MODEL_DIR),
+    local_dir_use_symlinks=False,
+    token=os.getenv("HF_TOKEN") or None,
+)
+print(f"[download] saved: {path}", file=sys.stderr)
+
+# Optional: write MODEL_PATH file for debugging
+print(path, file=open(MODEL_DIR / "MODEL_PATH.txt", "w"))