refactor: replace manual base64 decoder with pytoniq_core Cell parser

- clean_decode now uses Cell.one_from_boc + load_snake_string
- no more regex/string hacks, native BOC parsing
- added full transaction_data debug log in process_transaction

app/utils/decoder.py

@@ -1,26 +1,38 @@
 import base64
 import logging
-import re
-import string
+
+from pytoniq_core import Cell
 
 logger = logging.getLogger(__name__)
 
 
+# OLD decoder (manual base64 + regex, kept for reference):
+#
+# import re, string
+# def clean_decode(payload: str) -> str:
+#     s = re.sub(r'[^A-Za-z0-9+/=]', '', payload.strip())
+#     s += '=' * (-len(s) % 4)
+#     text = base64.b64decode(s).decode('utf-8', errors='ignore')
+#     text = ''.join(c for c in text if c in string.printable or c.isspace())
+#     match = re.search(r'([0-9]*\s*Telegram .*?Ref#[A-Za-z0-9]+)', text, re.S)
+#     return match.group(1).strip() if match else text.strip()
+
+
 def clean_decode(payload: str) -> str:
     logger.debug("Original payload: %s", payload)
 
-    # Strip non-Base64 chars and decode
-    s = re.sub(r'[^A-Za-z0-9+/=]', '', payload.strip())
-    s += '=' * (-len(s) % 4)
-    text = base64.b64decode(s).decode('utf-8', errors='ignore')
+    # Pad and decode base64 → BOC bytes
+    s = payload.strip()
+    if not s:
+        return ""
+    s += "=" * (-len(s) % 4)
+    boc = base64.b64decode(s)
 
-    # Keep only printable characters
-    text = ''.join(c for c in text if c in string.printable or c.isspace())
-
-    # Extract "Telegram … Ref#XXXX" block
-    match = re.search(r'([0-9]*\s*Telegram .*?Ref#[A-Za-z0-9]+)', text, re.S)
-    result = match.group(1).strip() if match else text.strip()
+    # Parse BOC cell and read snake-encoded text (skipping 32-bit op prefix)
+    cell = Cell.one_from_boc(boc)
+    sl = cell.begin_parse()
+    sl.load_uint(32)  # op code — always 0 for text comment
+    result = sl.load_snake_string().strip()
 
     logger.debug("Decoded result: %s", result)
     return result
-