Dateien nach "scripts" hochladen

scripts/import_markdown.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
 # -*- coding: utf-8 -*-
 """
-scripts/import_markdown.py
+scripts/import_markdown.py  (V2.3.1)
 
 Zweck
 -----
@@ -11,6 +11,12 @@ Zweck
 - Integriert eine optionale Type-Registry (types.yaml), um z. B. chunk_profile
   und retriever_weight pro Notiz-Typ zu steuern.
 
+Fix in V2.3.1
+-------------
+- `retriever_weight` wird nun **immer deterministisch** gesetzt:
+  Frontmatter-Override > types.yaml > Default=1.0 (falls nichts konfiguriert).
+  Das Feld wird **zwingend** in mindnet_notes **und** mindnet_chunks geschrieben.
+
 Kompatibilität & Fixes
 ----------------------
 - Unterstützt sowohl app.core.derive_edges (bevorzugt) als auch app.core.edges als Fallback
@@ -18,7 +24,6 @@ Kompatibilität & Fixes
     nicht zu TypeError führen.
 - `scroll_filter` wird für alle Scrolls verwendet (Qdrant >= 1.7.x).
 - `--purge-before-upsert` entfernt alte Chunks/Edges einer Note, wenn sich die Note geändert hat.
-- `retriever_weight` aus types.yaml bzw. Frontmatter wird in Note- und Chunk-Payload gespiegelt.
 - Baseline-Hash-Strategie: hash_mode (body|frontmatter|full), hash_source (parsed|raw), hash_normalize (canonical|none).
 
 Aufrufbeispiele
@@ -113,7 +118,7 @@ def effective_chunk_profile(note_type: str, reg: dict) -> Optional[str]:
         return prof
     return None
 
-def effective_retriever_weight(note_type: str, reg: dict) -> Optional[float]:
+def effective_retriever_weight_from_registry(note_type: str, reg: dict) -> Optional[float]:
     cfg = get_type_config(note_type, reg)
     w = cfg.get("retriever_weight")
     try:
@@ -121,6 +126,25 @@ def effective_retriever_weight(note_type: str, reg: dict) -> Optional[float]:
     except Exception:
         return None
 
+def compute_effective_retriever_weight(fm: Dict[str, Any], note_type: str, reg: dict) -> float:
+    """Ermittelt den finalen retriever_weight:
+       1) Frontmatter-Override
+       2) types.yaml (für den type)
+       3) Default 1.0
+    """
+    # 1) Frontmatter-Override
+    if fm.get("retriever_weight") is not None:
+        try:
+            return float(fm.get("retriever_weight"))
+        except Exception:
+            pass
+    # 2) Registry
+    r = effective_retriever_weight_from_registry(note_type, reg)
+    if r is not None:
+        return float(r)
+    # 3) Default
+    return 1.0
+
 # ------------------------------------------------------------
 # Sonstige Helper
 # ------------------------------------------------------------
@@ -345,12 +369,9 @@ def main() -> None:
         if prof:
             fm["chunk_profile"] = prof
 
-        weight = effective_retriever_weight(note_type, reg)
-        if weight is not None:
-            try:
-                fm["retriever_weight"] = float(weight)
-            except Exception:
-                pass
+        # NEU: finalen retriever_weight deterministisch bestimmen
+        rw = compute_effective_retriever_weight(fm, note_type, reg)
+        fm["retriever_weight"] = rw  # Frontmatter spiegeln, damit nachfolgende Builder konsistent sind
 
         # --- Payload aufbauen (inkl. Hashes) ---
         try:
@@ -369,12 +390,11 @@ def main() -> None:
         if not note_pl.get("fulltext"):
             note_pl["fulltext"] = getattr(parsed, "body", "") or ""
 
-        # retriever_weight sicher in Note-Payload spiegeln
-        if "retriever_weight" not in note_pl and fm.get("retriever_weight") is not None:
-            try:
-                note_pl["retriever_weight"] = float(fm.get("retriever_weight"))
-            except Exception:
-                pass
+        # NEU: retriever_weight **immer** in Note-Payload setzen
+        try:
+            note_pl["retriever_weight"] = float(rw)
+        except Exception:
+            note_pl["retriever_weight"] = 1.0
 
         note_id = note_pl.get("note_id") or fm.get("id")
         if not note_id:
@@ -410,15 +430,13 @@ def main() -> None:
             print(json.dumps({"path": path, "note_id": note_id, "error": f"chunk build failed: {type(e).__name__}: {e}"}))
             continue
 
-        # retriever_weight auf Chunk-Payload spiegeln
-        if fm.get("retriever_weight") is not None:
-            try:
-                rw = float(fm.get("retriever_weight"))
-                for pl in chunk_pls:
-                    if "retriever_weight" not in pl:
-                        pl["retriever_weight"] = rw
-            except Exception:
-                pass
+        # NEU: retriever_weight **immer** auf Chunk-Payload spiegeln
+        try:
+            rwf = float(rw)
+        except Exception:
+            rwf = 1.0
+        for pl in chunk_pls:
+            pl["retriever_weight"] = rwf
 
         # Embeddings (fallback: Nullvektoren)
         vecs: List[List[float]] = [[0.0] * int(cfg.dim) for _ in chunk_pls]
@@ -450,6 +468,8 @@ def main() -> None:
         summary = {
             "note_id": note_id,
             "title": fm.get("title"),
+            "type": fm.get("type"),
+            "rw": rw,
             "chunks": len(chunk_pls),
             "edges": len(edges),
             "edges_failed": edges_failed,