Harden Smoke Signal noise learning and JSON decision writes

smoke_signal_tab.py

@@ -62,6 +62,7 @@ NOISE_GLOBAL_FILE = NOISE_DIR / "_global_noise.json"
 NOISE_MIN_CHARS = int(os.environ.get("SS_NOISE_MIN_CHARS", "3"))
 NOISE_MATCH_MIN_PATTERN_COVERAGE = float(os.environ.get("SS_NOISE_MATCH_MIN_PATTERN_COVERAGE", "0.65"))
 NOISE_MATCH_MIN_TEXT_COVERAGE = float(os.environ.get("SS_NOISE_MATCH_MIN_TEXT_COVERAGE", "0.08"))
+_NOISE_IO_LOCK = threading.RLock()
 
 
 def _normalize_noise_text(text: str) -> str:
@@ -96,19 +97,38 @@ def _load_noise_patterns_for_scope(path: Path) -> list[str]:
     return out
 
 
+def _write_json_atomic(path: Path, payload) -> None:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    fd, tmp_name = tempfile.mkstemp(prefix=path.name + ".", suffix=".tmp", dir=str(path.parent))
+    tmp_path = Path(tmp_name)
+    try:
+        with os.fdopen(fd, "w", encoding="utf-8") as fh:
+            json.dump(payload, fh, indent=2, ensure_ascii=False)
+            fh.flush()
+            os.fsync(fh.fileno())
+        os.replace(str(tmp_path), str(path))
+    finally:
+        if tmp_path.exists():
+            try:
+                tmp_path.unlink()
+            except Exception:
+                pass
+
+
 def _load_noise_patterns(book_id: str, include_global: bool = True) -> list[str]:
     combined: list[str] = []
     seen: set[str] = set()
     paths = [_noise_path(book_id)]
     if include_global:
         paths.append(NOISE_GLOBAL_FILE)
-    for path in paths:
-        for pattern in _load_noise_patterns_for_scope(path):
-            norm = _normalize_noise_text(pattern)
-            if norm in seen:
-                continue
-            seen.add(norm)
-            combined.append(pattern)
+    with _NOISE_IO_LOCK:
+        for path in paths:
+            for pattern in _load_noise_patterns_for_scope(path):
+                norm = _normalize_noise_text(pattern)
+                if norm in seen:
+                    continue
+                seen.add(norm)
+                combined.append(pattern)
     return combined
 
 
@@ -121,18 +141,47 @@ def _save_noise_pattern(book_id: str, pattern: str, save_global: bool = True) ->
     targets = [_noise_path(book_id)]
     if save_global:
         targets.append(NOISE_GLOBAL_FILE)
-    for target in targets:
-        existing = _load_noise_patterns_for_scope(target)
-        existing_norm = {_normalize_noise_text(p) for p in existing}
-        if norm in existing_norm:
-            continue
-        existing.append(pattern)
-        target.parent.mkdir(parents=True, exist_ok=True)
-        target.write_text(json.dumps(existing, indent=2, ensure_ascii=False))
-        changed = True
+    with _NOISE_IO_LOCK:
+        for target in targets:
+            existing = _load_noise_patterns_for_scope(target)
+            existing_norm = {_normalize_noise_text(p) for p in existing}
+            if norm in existing_norm:
+                continue
+            existing.append(pattern)
+            _write_json_atomic(target, existing)
+            changed = True
     return changed
 
 
+def _json_default(obj):
+    # numpy / pandas scalar safety for json dumps
+    try:
+        import numpy as np  # local import to avoid hard dependency at import-time
+        if isinstance(obj, np.generic):
+            return obj.item()
+    except Exception:
+        pass
+
+    if isinstance(obj, Path):
+        return str(obj)
+
+    # pandas Timestamp/NA etc.
+    try:
+        if isinstance(obj, pd.Timestamp):
+            return obj.isoformat()
+    except Exception:
+        pass
+
+    # Generic numeric coercion fallback
+    try:
+        if hasattr(obj, "item"):
+            return obj.item()
+    except Exception:
+        pass
+
+    return str(obj)
+
+
 def _text_matches_noise(text: str, patterns: list) -> bool:
     if not patterns or not text:
         return False
@@ -2766,7 +2815,7 @@ def save_review_decision(idx: int, final_text: str, action: str, noise_text_inpu
             "conf_class":    "verified-100" if conf_override else item.get("confidence_class",""),
             "punct_score":   punct_score,
             "punct_flags_count": punct_flags_count,
-        }) + "\n")
+        }, ensure_ascii=False, default=_json_default) + "\n")
 
     learned_pairs = 0
     learned_noise_added = 0