feat(fortune): add download_model.py utility with resume and hash validation

tools/fortune_ai/download_model.py

@@ -0,0 +1,95 @@
+# -*- coding: utf-8 -*-
+import hashlib
+import json
+import os
+import sys
+import time
+import urllib.request
+
+ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), '../..'))
+MANIFEST_PATH = os.path.join(ROOT, 'ai/model.json')
+
+with open(MANIFEST_PATH, 'r', encoding='utf-8') as f:
+    manifest = json.load(f)
+
+dest_path = os.path.join(ROOT, 'ai', manifest['file'])
+part_path = dest_path + '.part'
+os.makedirs(os.path.dirname(dest_path), exist_ok=True)
+
+expected_bytes = manifest['bytes']
+expected_sha256 = manifest['sha256'].lower()
+
+if os.path.exists(dest_path):
+    actual_bytes = os.path.getsize(dest_path)
+    if actual_bytes == expected_bytes:
+        print(f"[INFO] Destination file already exists ({actual_bytes} bytes). Verifying SHA256...")
+        h = hashlib.sha256()
+        with open(dest_path, 'rb') as f:
+            while chunk := f.read(1024 * 1024):
+                h.update(chunk)
+        if h.hexdigest().lower() == expected_sha256:
+            print("[SUCCESS] Model already downloaded and verified!")
+            sys.exit(0)
+        else:
+            print("[WARN] Existing file has checksum mismatch, re-downloading...")
+            os.remove(dest_path)
+
+url = f"https://huggingface.co/{manifest['source']}/resolve/{manifest['revision']}/{manifest['source_file']}?download=true"
+print(f"Downloading model from:\n{url}")
+print(f"Destination: {dest_path}")
+print(f"Expected size: {expected_bytes} bytes (~{expected_bytes / (1024**3):.2f} GB)")
+
+headers = {'User-Agent': 'Mozilla/5.0'}
+existing_bytes = 0
+if os.path.exists(part_path):
+    existing_bytes = os.path.getsize(part_path)
+    if existing_bytes < expected_bytes:
+        print(f"Resuming download from byte {existing_bytes}...")
+        headers['Range'] = f"bytes={existing_bytes}-"
+    else:
+        existing_bytes = 0
+        os.remove(part_path)
+
+req = urllib.request.Request(url, headers=headers)
+try:
+    response = urllib.request.urlopen(req, timeout=30)
+except Exception as e:
+    print(f"[ERROR] Failed to open URL: {e}")
+    sys.exit(1)
+
+mode = 'ab' if existing_bytes > 0 else 'wb'
+downloaded = existing_bytes
+start_time = time.time()
+last_report_time = start_time
+last_report_bytes = downloaded
+
+with open(part_path, mode) as out:
+    while True:
+        chunk = response.read(1024 * 512) # 512 KB
+        if not chunk:
+            break
+        out.write(chunk)
+        downloaded += len(chunk)
+        now = time.time()
+        if now - last_report_time >= 5.0 or downloaded == expected_bytes:
+            speed_mb = (downloaded - last_report_bytes) / (now - last_report_time) / (1024 * 1024)
+            pct = (downloaded / expected_bytes) * 100.0
+            print(f"Progress: {downloaded / (1024**3):.2f} / {expected_bytes / (1024**3):.2f} GB ({pct:.1f}%) | {speed_mb:.2f} MB/s")
+            last_report_time = now
+            last_report_bytes = downloaded
+
+print("\nDownload finished! Verifying SHA256 checksum...")
+h = hashlib.sha256()
+with open(part_path, 'rb') as f:
+    while chunk := f.read(1024 * 1024):
+        h.update(chunk)
+
+actual_sha256 = h.hexdigest().lower()
+if actual_sha256 != expected_sha256:
+    print(f"[ERROR] SHA256 mismatch!\nExpected: {expected_sha256}\nGot:      {actual_sha256}")
+    sys.exit(1)
+
+if os.path.exists(dest_path):
+    os.remove(dest_path)
+os.rename(part_path, dest_path)
+print("[SUCCESS] Model successfully downloaded and verified!")