.

help_processor.py

@@ -1352,6 +1352,61 @@ def fallback_classify(title: str, text: str) -> tuple[str, str]:
     return t[:200], ", ".join(seen)
 
 
+KEYWORDS_MAX_LEN = 300  # rip_help_sections.keywords VARCHAR(300)
+
+
+def parse_keyword_list(raw: str) -> list[str]:
+    """Разделя ключови думи. Запетая/точка и запетая = фрази; иначе — интервал."""
+    text = (raw or "").strip()
+    if not text:
+        return []
+    if "," in text or ";" in text:
+        parts = re.split(r"[,;]+", text)
+    else:
+        parts = text.split()
+    out: list[str] = []
+    seen: set[str] = set()
+    for p in parts:
+        k = " ".join(p.split())
+        if not k:
+            continue
+        key = k.casefold()
+        if key in seen:
+            continue
+        seen.add(key)
+        out.append(k)
+    return out
+
+
+def merge_section_keywords(
+    manual: str,
+    generated: str,
+    max_len: int = KEYWORDS_MAX_LEN,
+) -> str:
+    """Ръчните думи първи, после генерираните; без дубликати (без значение на регистъра)."""
+    first = parse_keyword_list(manual)
+    seen = {k.casefold() for k in first}
+    rest: list[str] = []
+    for k in parse_keyword_list(generated):
+        key = k.casefold()
+        if key in seen:
+            continue
+        seen.add(key)
+        rest.append(k)
+    merged = first + rest
+    if not merged:
+        return ""
+    out: list[str] = []
+    used = 0
+    for i, k in enumerate(merged):
+        extra = len(k) + (2 if i else 0)  # ", "
+        if used + extra > max_len:
+            break
+        out.append(k)
+        used += extra
+    return ", ".join(out)
+
+
 def classify_section(client: anthropic.Anthropic, title: str, text: str) -> tuple[str, str]:
     """Връща (наименование, 'кл1, кл2, кл3') чрез Claude."""
     snippet = text[:MAX_AI_CHARS]
@@ -1420,6 +1475,7 @@ def process_file(
     force: bool = False,
     remote_root: Optional[str] = None,
     source_key: Optional[str] = None,
+    seed_keywords: str = "",
 ) -> dict:
     """Обработва един файл. Връща статистика (saved, codes, ok)."""
     rel = source_key or path.name
@@ -1512,6 +1568,7 @@ def process_file(
         if not (keywords or "").strip():
             title, keywords = fallback_classify(title or sec.title or f"Секция {sec_index}", text)
             ai_errors += 1
+        keywords = merge_section_keywords(seed_keywords, keywords)
 
         images_json = json.dumps(image_rel_paths, ensure_ascii=False)
         ps = ProcessedSection(
@@ -1558,6 +1615,7 @@ def process_directory(
     force: bool = False,
     purge_missing: bool = False,
     remote_root: Optional[str] = None,
+    seed_keywords: str = "",
 ):
     if not _PREFIX_RE.match(prefix):
         raise ValueError(
@@ -1598,6 +1656,7 @@ def process_directory(
                 path, idx, db, client, output_dir,
                 prefix=prefix, force=force, remote_root=remote_root,
                 source_key=identity,
+                seed_keywords=seed_keywords,
             )
             file_results.append(info)
             total_sections += int(info.get("saved") or 0)
@@ -1693,6 +1752,11 @@ def main():
              "Файловете се пишат локално; в БД се записва remote път. "
              "Или env HELP_REMOTE_ROOT."
     )
+    parser.add_argument(
+        "--keywords",
+        default="",
+        help="Ръчни ключови думи (запетая). Стоят в началото на всяка секция от това сканиране."
+    )
     args = parser.parse_args()
 
     if not args.api_key:
@@ -1709,6 +1773,7 @@ def main():
         force=args.force,
         purge_missing=args.purge_missing,
         remote_root=args.remote_root,
+        seed_keywords=args.keywords,
     )