app/core/derive_edges.py aktualisiert
All checks were successful
Deploy mindnet to llm-node / deploy (push) Successful in 2s

This commit is contained in:
Lars 2025-11-08 07:59:24 +01:00
parent 5f358d329b
commit 44e468fc21

View File

@ -2,40 +2,33 @@
# -*- coding: utf-8 -*- # -*- coding: utf-8 -*-
""" """
Modul: app/core/derive_edges.py Modul: app/core/derive_edges.py
Version: 1.4.0 Version: 1.9.0
Datum: 2025-10-01 Datum: 2025-11-07
Zweck Zweck
----- -----
Robuste Kantenbildung für mindnet (Notes/Chunks): Robuste Kantenbildung für mindnet (Notes/Chunks):
- belongs_to (chunk -> note) belongs_to (chunk note)
- next / prev (chunk-Kette) next / prev (Chunk-Sequenz)
- references (chunk-scope) aus Chunk.window/text references (chunk-scope) aus Chunk.window/text
- optional references/backlink (note-scope) optional note-scope references/backlink (Flag)
optional Default-Kanten pro Note-Type aus Type-Registry (falls vorhanden)
Wichtig: Wikilinks werden mit der Parser-Funktion `extract_wikilinks` extrahiert, Abwärtskompatibel zu v1.4.0 (keine Pflicht auf Registry).
damit Varianten wie [[id#anchor]] oder [[id|label]] korrekt auf 'id' reduziert werden.
Erwartete Chunk-Payload-Felder:
{
"note_id": "...",
"chunk_id": "...", # Alias "id" ist zulässig
"id": "...",
"chunk_index": int,
"seq": int,
"window": str,
"text": str,
"path": "rel/path.md",
...
}
""" """
from __future__ import annotations from __future__ import annotations
from typing import Dict, List, Optional, Iterable from typing import Dict, List, Optional, Iterable
# WICHTIG: benutze die Parser-Extraktion für saubere Wikilinks
from app.core.parser import extract_wikilinks from app.core.parser import extract_wikilinks
# Type-Registry (optional)
try:
from app.core.type_registry import get_edge_defaults # type: ignore
except Exception:
def get_edge_defaults(_note_type: str) -> List[str]:
return []
def _get(d: dict, *keys, default=None): def _get(d: dict, *keys, default=None):
for k in keys: for k in keys:
if k in d and d[k] is not None: if k in d and d[k] is not None:
@ -43,7 +36,6 @@ def _get(d: dict, *keys, default=None):
return default return default
def _chunk_text_for_refs(chunk: dict) -> str: def _chunk_text_for_refs(chunk: dict) -> str:
# bevorzugt 'window' → dann 'text' → 'content' → 'raw'
return ( return (
_get(chunk, "window") _get(chunk, "window")
or _get(chunk, "text") or _get(chunk, "text")
@ -62,13 +54,7 @@ def _dedupe(seq: Iterable[str]) -> List[str]:
return out return out
def _edge(kind: str, scope: str, source_id: str, target_id: str, note_id: str, extra: Optional[dict] = None) -> dict: def _edge(kind: str, scope: str, source_id: str, target_id: str, note_id: str, extra: Optional[dict] = None) -> dict:
pl = { pl = {"kind": kind, "scope": scope, "source_id": source_id, "target_id": target_id, "note_id": note_id}
"kind": kind,
"scope": scope, # "chunk" | "note"
"source_id": source_id,
"target_id": target_id,
"note_id": note_id, # Träger/Quelle der Kante (aktuelle Note)
}
if extra: if extra:
pl.update(extra) pl.update(extra)
return pl return pl
@ -79,14 +65,6 @@ def build_edges_for_note(
note_level_references: Optional[List[str]] = None, note_level_references: Optional[List[str]] = None,
include_note_scope_refs: bool = False, include_note_scope_refs: bool = False,
) -> List[dict]: ) -> List[dict]:
"""
Erzeugt Kanten für eine Note.
- belongs_to: für jeden Chunk (chunk -> note)
- next / prev: zwischen aufeinanderfolgenden Chunks
- references: pro Chunk aus window/text
- optional note-scope references/backlinks: dedupliziert über alle Chunk-Funde + note_level_references
"""
edges: List[dict] = [] edges: List[dict] = []
# belongs_to # belongs_to
@ -106,14 +84,14 @@ def build_edges_for_note(
edges.append(_edge("next", "chunk", a_id, b_id, note_id, {"chunk_id": a_id})) edges.append(_edge("next", "chunk", a_id, b_id, note_id, {"chunk_id": a_id}))
edges.append(_edge("prev", "chunk", b_id, a_id, note_id, {"chunk_id": b_id})) edges.append(_edge("prev", "chunk", b_id, a_id, note_id, {"chunk_id": b_id}))
# references (chunk-scope) Links aus window bevorzugen (Overlap-fest) # references (chunk-scope)
refs_all: List[str] = [] refs_all: List[str] = []
for ch in chunks: for ch in chunks:
cid = _get(ch, "chunk_id", "id") cid = _get(ch, "chunk_id", "id")
if not cid: if not cid:
continue continue
txt = _chunk_text_for_refs(ch) txt = _chunk_text_for_refs(ch)
refs = extract_wikilinks(txt) # <— Parser-Logik, kompatibel zu deinem System refs = extract_wikilinks(txt)
for r in refs: for r in refs:
edges.append(_edge("references", "chunk", cid, r, note_id, {"chunk_id": cid, "ref_text": r})) edges.append(_edge("references", "chunk", cid, r, note_id, {"chunk_id": cid, "ref_text": r}))
refs_all.extend(refs) refs_all.extend(refs)
@ -128,4 +106,19 @@ def build_edges_for_note(
edges.append(_edge("references", "note", note_id, r, note_id)) edges.append(_edge("references", "note", note_id, r, note_id))
edges.append(_edge("backlink", "note", r, note_id, note_id)) edges.append(_edge("backlink", "note", r, note_id, note_id))
# optional: Default-Kanten aus Registry (no-op, wenn leer)
# Beispiel: task → depends_on, concept → related_to etc.
# Wir erzeugen nur "formale" Kanten note→note_id selbst nicht; diese dienen Retri­ever-Gewichtung später.
# (Hier keine Targets das sind Typ-Hinweise und werden als self-hints abgelegt.)
try:
note_type = ""
if chunks:
note_type = str(_get(chunks[0], "type", default="") or "")
defaults = get_edge_defaults(note_type)
for k in defaults:
# Self-hint-Kante (kann vom Retriever gewichtet werden). Scope 'note'
edges.append(_edge(k, "note", note_id, note_id, note_id))
except Exception:
pass
return edges return edges