validators.py

import re
from typing import List, Tuple
from .schemas import Patch, Event, Update

INVEST_TOKENS = ['invest','funded','wired','purchase','bought','loaned']
DIST_TOKENS = ['distribution','profit distribution','tax distribution','dividend','received a wire','received','interest payment','interest received']
REPAY_TOKENS = ['repay','repayment','principal payment','paid back','buyback']
CONV_TOKENS = ['convert','converted','conversion']

OUR_PARTIES = [r'Radical Investments', r'Mark Cuban Companies', r'Mark Cuban', r'\bMCC\b']

def party_is_us(text: str) -> bool:
    return any(re.search(p, text or '', flags=re.I) for p in OUR_PARTIES) or bool(re.search(r'received (?:a )?wire', text or '', re.I))

def classify_confidence(e: Event) -> float:
    c = 0.0
    if e.date: c += 0.3
    if e.type in ['investment','distribution','repayment','conversion']: c += 0.2
    if e.amount is not None: c += 0.2
    if e.note and party_is_us(e.note): c += 0.3
    return round(min(c, 1.0), 2)

def validate_patch(p: Patch) -> Tuple[bool, float, List[str], List[str], Patch]:
    warnings, errors = [], []
    # Basic checks
    if not p.company_id:
        errors.append("company_id missing")
    # Normalize events: drop non-our-party cash events
    norm_events = []
    for e in p.events_to_append:
        if e.type in ['investment','distribution','repayment']:
            if not party_is_us(e.note or ''):
                warnings.append(f"Event on {e.date} filtered (not our cash party): {e.note}")
                continue
        conf = classify_confidence(e)
        norm_events.append(e)
    # Confidence heuristic: average over events/updates
    scores = [classify_confidence(e) for e in norm_events] or [0.0]
    confidence = round(sum(scores)/len(scores), 2)
    normalized = Patch(
        company_id=p.company_id,
        dossier_overrides=p.dossier_overrides,
        events_to_append=norm_events,
        updates_to_append=p.updates_to_append,
        documents_to_register=p.documents_to_register,
        contacts_to_set=p.contacts_to_set
    )
    ok = len(errors) == 0
    return ok, confidence, warnings, errors, normalized