refactor: deduplicate filename cleanup and generic token helpers

core/matcher.py

@@ -284,7 +284,7 @@ def base_model_score(candidate: str, preferred: Optional[str]) -> float:
 
 # ==================== CENTRALIZED MATCHING & CONFIDENCE HELPERS ====================
 
-from .type_utils import MODEL_EXTENSIONS
+from .type_utils import MODEL_EXTENSIONS, PRECISION_FORMAT_SUFFIXES
 MODEL_FILE_EXTENSIONS = MODEL_EXTENSIONS
 
 
@@ -392,6 +392,19 @@ def calculate_candidate_rank(
     return matches, confidence + score
 
 
+def clean_filename_for_search(filename: str) -> str:
+    """
+    Clean up filename for better search results.
+    Remove common suffixes that might prevent matches.
+    """
+    if not filename:
+        return ""
+    basename = os.path.basename(filename).strip()
+    base = os.path.splitext(basename)[0]
+    pattern = r"[-_]?(" + "|".join(PRECISION_FORMAT_SUFFIXES) + r").*$"
+    return re.sub(pattern, "", base, flags=re.IGNORECASE).strip(" -_")
+
+
 def build_filename_search_queries(filename: str) -> List[str]:
     """
     Build a list of candidate search queries from a filename by stripping
@@ -405,7 +418,7 @@ def build_filename_search_queries(filename: str) -> List[str]:
     stem_lower = stem.lower()
     
     # Pattern to match common precision/format suffixes
-    pattern = r"[-_]?(fp16|fp32|fp8|fp4|bf16|e4m3fn|mixed|scaled|pruned|emaonly|q4|q8)"
+    pattern = r"[-_]?(" + "|".join(PRECISION_FORMAT_SUFFIXES) + r")"
     
     # 1. Strip suffix followed by any other characters (like clean_filename_for_search)
     cleaned = re.sub(pattern + r".*$", "", stem, flags=re.IGNORECASE).strip(" -_")

core/sources/huggingface.py

@@ -19,7 +19,7 @@ log = create_module_logger(__name__)
 
 from ..path_utils import write_json_atomic, METADATA_DIR, read_json_safe
 from ..type_utils import check_credential_http, parse_size_header as _parse_size_header, fetch_remote_file_size, fetch_remote_file_size_cached, clear_remote_size_cache, extract_file_size
-from ..matcher import build_filename_search_queries
+from ..matcher import build_filename_search_queries, clean_filename_for_search
 
 HF_API_URL = "https://huggingface.co/api"
 HF_AUTHOR_FALLBACKS = ["Comfy-Org"]
@@ -358,23 +358,6 @@ def _fetch_remote_file_size_bytes(
     return fetch_remote_file_size_cached(probe_url, headers=headers, timeout=timeout)
 
 
-
-def clean_filename_for_search(filename: str) -> str:
-    """
-    Clean up filename for better search results.
-    Remove common suffixes that might prevent matches.
-    """
-    base = os.path.splitext(filename)[0]
-    # Remove common precision/format suffixes
-    base = re.sub(
-        r"[-_]?(fp16|fp32|fp8|fp4|bf16|e4m3fn|scaled|pruned|emaonly|mixed|q4|q8).*$",
-        "",
-        base,
-        flags=re.IGNORECASE,
-    )
-    return base
-
-
 def _build_huggingface_result(
     repo_id: str,
     file_path: str,

core/sources/lora_manager_archive.py

@@ -18,7 +18,7 @@ from ..matcher import (
     base_model_score as _base_model_score,
     calculate_candidate_rank,
 )
-from ..type_utils import select_primary_model_file
+from ..type_utils import select_primary_model_file, get_generic_filename_tokens
 from ..progress import report_progress, get_progress_reporter
 from ..log_system import create_module_logger
 log = create_module_logger(__name__)
@@ -29,22 +29,7 @@ DEFAULT_DB_RELATIVE_PATH = os.path.join("civitai", "civitai.sqlite")
 
 _search_cache: Dict[str, Any] = {}
 _db_path_cache: Optional[str] = None
-GENERIC_FILENAME_TOKENS = {
-    "safetensors",
-    "ckpt",
-    "pt",
-    "pth",
-    "bin",
-    "onnx",
-    "gguf",
-    "fp16",
-    "fp32",
-    "bf16",
-    "e4m3fn",
-    "scaled",
-    "pruned",
-    "emaonly",
-}
+GENERIC_FILENAME_TOKENS = get_generic_filename_tokens()
 
 
 def clear_search_cache():

core/type_utils.py

@@ -60,6 +60,29 @@ MODEL_EXTENSIONS = {
 }
 
 
+PRECISION_FORMAT_SUFFIXES = [
+    "fp16",
+    "fp32",
+    "fp8",
+    "fp4",
+    "bf16",
+    "e4m3fn",
+    "mixed",
+    "scaled",
+    "pruned",
+    "emaonly",
+    "q4",
+    "q8",
+]
+
+
+def get_generic_filename_tokens() -> set:
+    """Return a set of generic filename tokens (extensions and suffixes) in lowercase."""
+    dotless_extensions = {ext.lstrip(".") for ext in MODEL_EXTENSIONS}
+    return dotless_extensions | set(PRECISION_FORMAT_SUFFIXES)
+
+
+
 def first_non_empty(*values: Any, default: Any = "") -> Any:
     """
     Return the first value that is not None, not an empty/whitespace-only string,