diff --git a/scripts/import_markdown.py b/scripts/import_markdown.py index 336408f..2f9a1fe 100644 --- a/scripts/import_markdown.py +++ b/scripts/import_markdown.py @@ -1,33 +1,28 @@ #!/usr/bin/env python3 # -*- coding: utf-8 -*- """ -scripts/import_markdown.py (V2.3.1) +scripts/import_markdown.py (Mindnet V2 — Importer, v2.5.0) Zweck ----- - Liest Markdown-Notizen aus einem Vault ein - Erzeugt Note-Payload, Chunk-Payloads (+ optionale Embeddings) und Edges - Schreibt alles idempotent in Qdrant (Notes, Chunks, Edges) -- Integriert eine optionale Type-Registry (types.yaml), um z. B. chunk_profile - und retriever_weight pro Notiz-Typ zu steuern. +- Nutzt optional eine Type-Registry (`./config/types.yaml`) für: + - `chunk_profile` pro type + - `retriever_weight` pro type (Frontmatter > Registry > Default 1.0) -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. +Kernfeatures +------------ +- Deterministische Hashes via `make_note_payload(...)` (hash_mode/source/normalize) +- Idempotenter Upsert; Baseline-Hash-Merge je Modus +- Optionales `--sync-deletes` (Vault → Qdrant) +- Edges aus realen Referenzen + Nachbarschaft (next/prev/belongs_to) via `derive_edges.py`/`edges.py` +- `retriever_weight` wird **immer** in Notes & Chunks geschrieben +- Keine Duplikate wie `chunk_num`/`Chunk_Number` → wir halten uns an `index` (0‑basiert) und `ord` (= index+1) -Kompatibilität & Fixes ----------------------- -- Unterstützt sowohl app.core.derive_edges (bevorzugt) als auch app.core.edges als Fallback - → Aufruf erfolgt mit POSITIONSARGUMENTEN, damit alte Signaturen (note_level_refs vs. note_level_references) - 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. -- Baseline-Hash-Strategie: hash_mode (body|frontmatter|full), hash_source (parsed|raw), hash_normalize (canonical|none). - -Aufrufbeispiele ---------------- +Aufruf +------ # Import (Apply + Purge-Update) python3 -m scripts.import_markdown --vault ./vault --apply --purge-before-upsert --prefix "$COLLECTION_PREFIX" @@ -44,52 +39,53 @@ from typing import Dict, List, Optional, Tuple, Any, Set from dotenv import load_dotenv from qdrant_client.http import models as rest +import yaml # --- Projekt-Imports --- -from app.core.parser import ( - read_markdown, - normalize_frontmatter, - validate_required_frontmatter, -) +from app.core.parser import read_markdown, normalize_frontmatter, validate_required_frontmatter from app.core.note_payload import make_note_payload from app.core.chunker import assemble_chunks from app.core.chunk_payload import make_chunk_payloads try: - # bevorzugt der robuste Builder + # bevorzugt: derive_edges (regelbasiert + real) from app.core.derive_edges import build_edges_for_note except Exception: # pragma: no cover - # Fallback auf die einfache Variante + # Fallback: einfache Kanten in edges.py from app.core.edges import build_edges_for_note # type: ignore -from app.core.qdrant import ( - QdrantConfig, - get_client, - ensure_collections, - ensure_payload_indexes, -) -from app.core.qdrant_points import ( - points_for_chunks, - points_for_note, - points_for_edges, - upsert_batch, -) +from app.core.qdrant import QdrantConfig, get_client, ensure_collections, ensure_payload_indexes +from app.core.qdrant_points import points_for_chunks, points_for_note, points_for_edges, upsert_batch -# embeddings sind optional (z. B. beim Payload-Backfill) +# embeddings optional try: from app.core.embed import embed_texts # optional except Exception: # pragma: no cover - embed_texts = None + embed_texts = None # type: ignore -# ------------------------------------------------------------ -# Type-Registry (types.yaml) -# ------------------------------------------------------------ -import yaml + +# ============================================================ +# Type-Registry (config/types.yaml) +# ============================================================ def _env(name: str, default: Optional[str] = None) -> str: v = os.getenv(name) return v if v is not None else (default or "") +def _deep_get(root: Any, path: str) -> Any: + cur = root + for key in path.split("."): + if not isinstance(cur, dict) or key not in cur: + return None + cur = cur[key] + return cur + +def _as_float(x: Any) -> Optional[float]: + try: + return float(x) + except Exception: + return None + def load_type_registry() -> dict: - # ENV kann Pfad überschreiben; Standard: ./config/types.yaml + # ENV überschreibt Pfad path = _env("MINDNET_TYPES_FILE", "./config/types.yaml") if not os.path.isfile(path): return {} @@ -118,48 +114,51 @@ def effective_chunk_profile(note_type: str, reg: dict) -> Optional[str]: return prof return None -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: - return float(w) if w is not None else None - 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 +def effective_retriever_weight_from_registry(note_type: str, reg: dict) -> Tuple[Optional[float], Optional[str]]: """ - # 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 + Liefert (Wert, Pfad) aus Registry, wenn vorhanden. + Unterstützte Pfade (in dieser Reihenfolge): + - types..retriever_weight + - types..retriever.weight + - types..retrieval.weight + - defaults.retriever_weight + - defaults.retriever.weight + - global.retriever_weight + - global.retriever.weight + """ + candidates = [ + f"types.{note_type}.retriever_weight", + f"types.{note_type}.retriever.weight", + f"types.{note_type}.retrieval.weight", + "defaults.retriever_weight", + "defaults.retriever.weight", + "global.retriever_weight", + "global.retriever.weight", + ] + for path in candidates: + val = _deep_get(reg, path) + v = _as_float(val) + if v is not None: + return v, path + return None, None -# ------------------------------------------------------------ -# Sonstige Helper -# ------------------------------------------------------------ -def iter_md(root: str) -> List[str]: - out: List[str] = [] - for dirpath, _, filenames in os.walk(root): - for fn in filenames: - if not fn.lower().endswith(".md"): - continue - p = os.path.join(dirpath, fn) - pn = p.replace("\\", "/") - if any(ex in pn for ex in ["/.obsidian/", "/_backup_frontmatter/", "/_imported/"]): - continue - out.append(p) - return sorted(out) +def compute_effective_retriever_weight(fm: Dict[str, Any], note_type: str, reg: dict) -> Tuple[float, str]: + """ + Priorität: Frontmatter > Registry > Default (1.0). Gibt (Wert, Quelle) zurück. + """ + if fm.get("retriever_weight") is not None: + v = _as_float(fm.get("retriever_weight")) + if v is not None: + return v, "frontmatter.retriever_weight" + r, rpath = effective_retriever_weight_from_registry(note_type, reg) + if r is not None: + return float(r), f"types.yaml:{rpath}" + return 1.0, "default:1.0" + + +# ============================================================ +# Qdrant Helpers +# ============================================================ def collections(prefix: str) -> Tuple[str, str, str]: return f"{prefix}_notes", f"{prefix}_chunks", f"{prefix}_edges" @@ -169,7 +168,7 @@ def fetch_existing_note_payload(client, prefix: str, note_id: str) -> Optional[D f = rest.Filter(must=[rest.FieldCondition(key="note_id", match=rest.MatchValue(value=note_id))]) points, _ = client.scroll( collection_name=notes_col, - scroll_filter=f, # wichtig: scroll_filter (nicht: filter) + scroll_filter=f, with_payload=True, with_vectors=False, limit=1, @@ -244,9 +243,11 @@ def artifacts_missing(client, prefix: str, note_id: str) -> Tuple[bool, bool]: edges_missing = not _has_any_point(client, edges_col, note_id) return chunks_missing, edges_missing -# ------------------------------------------------------------ + +# ============================================================ # Main -# ------------------------------------------------------------ +# ============================================================ + def _resolve_mode(m: Optional[str]) -> str: m = (m or "body").strip().lower() return m if m in {"body", "frontmatter", "full"} else "body" @@ -270,7 +271,7 @@ def main() -> None: ap.add_argument("--hash-normalize", help="canonical|none (Default canonical)") ap.add_argument("--compare-text", action="store_true", help="Parsed fulltext zusätzlich direkt vergleichen") ap.add_argument("--baseline-modes", action="store_true", help="Fehlende Hash-Varianten still nachtragen (Notes)") - ap.add_argument("--sync-deletes", action="store_true", help="Qdrant->Vault Lösch-Sync (Dry-Run; mit --apply ausführen)") + ap.add_argument("--sync-deletes", action="store_true", help="Qdrant->Vault Lösch-Sync (mit --apply ausführen)") ap.add_argument("--prefix", help="Collection-Prefix (überschreibt ENV COLLECTION_PREFIX)") args = ap.parse_args() @@ -289,7 +290,7 @@ def main() -> None: ensure_collections(client, cfg.prefix, cfg.dim) ensure_payload_indexes(client, cfg.prefix) - # Type-Registry laden (optional) + # Type-Registry laden reg = load_type_registry() root = os.path.abspath(args.vault) @@ -299,7 +300,17 @@ def main() -> None: only = os.path.abspath(args.only_path) files = [only] else: - files = iter_md(root) + files: List[str] = [] + for dirpath, _, filenames in os.walk(root): + for fn in filenames: + if not fn.lower().endswith(".md"): + continue + p = os.path.join(dirpath, fn) + pn = p.replace("\\", "/") + if any(ex in pn for ex in ["/.obsidian/", "/_backup_frontmatter/", "/_imported/"]): + continue + files.append(p) + files.sort() if not files: print("Keine Markdown-Dateien gefunden.", file=sys.stderr) sys.exit(2) @@ -358,7 +369,7 @@ def main() -> None: processed += 1 - # --- Type-Registry anwenden (chunk_profile / retriever_weight) --- + # --- Typ & Profile aus Registry übernehmen --- try: note_type = resolve_note_type(fm.get("type"), reg) except Exception: @@ -369,9 +380,9 @@ def main() -> None: if prof: fm["chunk_profile"] = prof - # 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 + # --- retriever_weight auflösen (FM > Registry > 1.0) --- + rw, rw_source = compute_effective_retriever_weight(fm, note_type, reg) + fm["retriever_weight"] = rw # Spiegeln ins FM für Builder # --- Payload aufbauen (inkl. Hashes) --- try: @@ -390,7 +401,7 @@ def main() -> None: if not note_pl.get("fulltext"): note_pl["fulltext"] = getattr(parsed, "body", "") or "" - # NEU: retriever_weight **immer** in Note-Payload setzen + # retriever_weight **immer** in Note-Payload schreiben try: note_pl["retriever_weight"] = float(rw) except Exception: @@ -430,11 +441,18 @@ def main() -> None: print(json.dumps({"path": path, "note_id": note_id, "error": f"chunk build failed: {type(e).__name__}: {e}"})) continue - # NEU: retriever_weight **immer** auf Chunk-Payload spiegeln - try: - rwf = float(rw) - except Exception: - rwf = 1.0 + # **Felder-Policy**: `index` (0-based) & `ord` (1-based) sicherstellen; keine doppelten Alias-Felder + for i, pl in enumerate(chunk_pls): + if "index" not in pl: + pl["index"] = i + pl["ord"] = int(pl.get("index", i)) + 1 + # Entferne ggf. Alt-Aliase, um Duplikate zu vermeiden + for alias in ("chunk_num", "Chunk_Number"): + if alias in pl: + pl.pop(alias, None) + + # retriever_weight **immer** auf Chunk-Payload spiegeln + rwf = float(rw) if isinstance(rw, (int, float)) else 1.0 for pl in chunk_pls: pl["retriever_weight"] = rwf @@ -456,8 +474,8 @@ def main() -> None: should_build_edges = (changed and (not do_baseline_only)) or edges_missing if should_build_edges: try: - # ACHTUNG: POSITIONSARGUMENTE benutzen → kompatibel zu edges.py UND derive_edges.py note_refs = note_pl.get("references") or [] + # Positionsargumente → kompatibel mit derive_edges.py UND edges.py edges = build_edges_for_note(note_id, chunk_pls, note_refs, include_note_scope_refs=note_scope_refs) except Exception as e: edges_failed = True @@ -470,6 +488,7 @@ def main() -> None: "title": fm.get("title"), "type": fm.get("type"), "rw": rw, + "rw_source": rw_source, "chunks": len(chunk_pls), "edges": len(edges), "edges_failed": edges_failed,