eSlider
a429b823e5
- bin/chats/: env-based paths, no /mnt/ /home/ hardcodes - bin/edelweiss-pilot: remove curasoft, use DOCS_BASE env var - bin/facts/crm: use KNOWLEDGE_MESH_SEED env var - compose.edelweiss.yml: remove curasoft volumes, use DOCS_BASE - docs/chat-import-plan.md: link to Gitea issue, no secrets - bin/seed-edelweiss-facts.py: removed (curasoft-only)
124 lines
4.8 KiB
Python
Executable File
124 lines
4.8 KiB
Python
Executable File
#!/usr/bin/env python3
|
|
"""facts/crm - prove person->company and company->project associations.
|
|
|
|
Two independent sources per fact:
|
|
|
|
S1 oo/OnlyOffice CRM (authoritative) : person.company_id -> company,
|
|
project.contacts -> company/person
|
|
S2 corpus SoT : eslider/cv/projects/knowledge-mesh-seed.yaml
|
|
(orgs: employer/client/... + projects)
|
|
|
|
Only associations supported by BOTH sources are written as root=facts.
|
|
Mismatches are reported (or, with --fix-crm, printed as oo CLI commands).
|
|
|
|
Usage:
|
|
bin/facts/crm write proven facts (needs var/kb.lbug)
|
|
bin/facts/crm --dry-run show proposed facts + mismatches only
|
|
bin/facts/crm --mismatches show associations found in only one side
|
|
"""
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
import os
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
ROOT = Path(__file__).resolve().parents[2]
|
|
sys.path.insert(0, str(ROOT / "bin" / "tools"))
|
|
|
|
from kblib import upsert_leaf, connect, leaf_id # noqa: E402
|
|
|
|
MESH_ENV = os.environ.get("KNOWLEDGE_MESH_SEED", "")
|
|
CORPUS_MESH = Path(MESH_ENV) if MESH_ENV else ROOT / "../knowledge-mesh-seed.yaml"
|
|
|
|
|
|
def corpus_orgs(raw: str) -> dict[str, dict]:
|
|
"""Delegate to tools.crmfacts.corpus_orgs (tested in tools/)."""
|
|
from crmfacts import corpus_orgs as _corpus_orgs
|
|
return _corpus_orgs(raw)
|
|
|
|
|
|
def main() -> int:
|
|
dry = "--dry-run" in sys.argv
|
|
mism = "--mismatches" in sys.argv
|
|
|
|
mesh = CORPUS_MESH.read_text()
|
|
orgs = corpus_orgs(mesh)
|
|
|
|
# CRM graph (produced by /tmp/opencode/crm/graph.py -> /tmp/opencode/crm/graph.json)
|
|
graph = json.load(open("/tmp/opencode/crm/graph.json"))
|
|
crm_person_company = graph["companies_with_persons"] # company -> [persons]
|
|
crm_project_companies = {} # pid -> title, companies
|
|
for pid, v in graph["projects_contacts"].items():
|
|
crm_project_companies[pid] = {"title": v["title"], "companies": v["companies"]}
|
|
|
|
facts: list[str] = []
|
|
mismatches: list[str] = []
|
|
|
|
# ---- person->company proven by CRM + corpus org ---- #
|
|
for org_name, org in orgs.items():
|
|
token = org.get("label", org_name)
|
|
# find CRM company whose name contains a significant token of the corpus org
|
|
key = next((k for k in crm_person_company
|
|
if token.split()[0].lower() in k.lower() or any(
|
|
t.lower() in k.lower() for t in org.get("label", "").split(" / "))),
|
|
None)
|
|
persons = crm_person_company.get(key, []) if key else []
|
|
if persons and org:
|
|
for p in persons:
|
|
facts.append(f"{p} is associated with {org.get('label')} "
|
|
f"(role: {org.get('kind', '?')}, {org.get('period', '')})")
|
|
elif org and key and not persons:
|
|
mismatches.append(f"corpus org '{org_name}' ({org.get('label')}) has no CRM persons")
|
|
elif org and not key:
|
|
mismatches.append(f"corpus org '{org_name}' ({org.get('label')}) not found in CRM")
|
|
|
|
# ---- corpus employer claims vs CRM ---- #
|
|
for org_name, org in orgs.items():
|
|
if not org or not org.get("kind"):
|
|
continue
|
|
if org["kind"] in ("employer", "own", "client", "agency", "apprenticeship"):
|
|
token = org.get("label", org_name).split()[0]
|
|
if not any(token.lower() in k.lower() for k in crm_person_company):
|
|
mismatches.append(f"corpus org '{org_name}' ({org['label']}) not found in CRM")
|
|
|
|
print(f"# CRM association facts proven (corpus x CRM): {len(facts)}")
|
|
for f in facts:
|
|
print(" -", f)
|
|
print(f"# mismatches / one-sided associations: {len(mismatches)}")
|
|
for f in mismatches:
|
|
print(" !", f)
|
|
|
|
if dry:
|
|
return 0
|
|
|
|
# ---- write proven facts into the brain (root=facts, 2 sources each) ---- #
|
|
import time
|
|
from model2vec import StaticModel
|
|
from kblib import MODEL # noqa: F401
|
|
model = StaticModel.from_pretrained(MODEL)
|
|
db, conn = connect(read_only=False)
|
|
try:
|
|
r = conn.execute("MATCH (l:Leaf) WHERE l.root='facts' RETURN count(*) AS n")
|
|
stats_before = r.get_all()[0][0]
|
|
except Exception:
|
|
stats_before = 0
|
|
rev = time.strftime("%Y%m%d-%H%M%S")
|
|
written = 0
|
|
for f in facts:
|
|
src = f"ooCRM x {CORPUS_MESH.name}"
|
|
lid = upsert_leaf(
|
|
conn,
|
|
text=f, root="facts", confidence="confirmed",
|
|
source=src, source_rev=rev,
|
|
how="crm-crosscheck", loc="bin/facts/crm", type_="association",
|
|
embedding=model.encode(f).tolist(),
|
|
)
|
|
written += 1
|
|
conn.close()
|
|
print(f"# wrote {written} facts into var/kb.lbug (facts was {stats_before})")
|
|
return 0
|
|
|
|
|
|
if __name__ == "__main__":
|
|
sys.exit(main()) |