senti/backend/api/read_layer.py

from backend.database.postgres.db import SessionLocal
from backend.database.postgres.models import User, Transaction, FinancialProfile, UserConsent, Invoice
from layers.layer_1_data.dag_adapter import dag_adapter
from layers.layer_1_data.vector_store import SentiVectorStore, SentiEmbeddingFunction
from core.market_profiles import get_market_profile
from core.legal_intelligence import legal_intelligence
import hashlib

vector_store = SentiVectorStore()
_embedding_fn = SentiEmbeddingFunction()

async def get_full_context(user_address: str, country_code: str, db):
    user_hash = hashlib.sha256(user_address.encode()).hexdigest()[:32]
    
    # 1. Market Profile
    market = get_market_profile(country_code)
    
    # 2. Ledger Transactions
    transactions = dag_adapter.fetch_live_ledger(user_address, days_window=30, max_transactions=500)
    
    # 3. User Info
    user = db.query(User).filter(User.phone_hash == user_hash).first()
    
    # 4. Financial Profile
    profile = db.query(FinancialProfile).filter(FinancialProfile.user_id == user_address).first()
    
    # 5. Language Detection (simple heuristic for now)
    language = "en"
    if user and user.language_preference:
        language = user.language_preference
    
    return {
        "user_address": user_address,
        "user_hash": user_hash,
        "country_code": country_code,
        "market": market,
        "transactions": transactions,
        "profile": profile,
        "language": language
    }

async def get_transactions(user_hash: str, db, days: int = 30):
    return dag_adapter.fetch_live_ledger(user_hash, days_window=days, max_transactions=500)

async def get_profile(user_hash: str, transactions, db):
    profile = db.query(FinancialProfile).filter(FinancialProfile.user_id == user_hash).first()
    return profile

def search_knowledge(query: str, domain: str, language: str, n_results: int = 3):
    collection_name = "legal_knowledge" if domain in ["TAXATION", "LENDING", "COMPLIANCE"] else "financial_knowledge"
    try:
        coll = vector_store.client.get_collection(
            name=collection_name,
            embedding_function=_embedding_fn
        )
        res = coll.query(
            query_texts=[query], 
            n_results=n_results,
            include=['documents', 'distances', 'metadatas']
        )
        if res and res.get('documents') and len(res['documents'][0]) > 0:
            docs = res['documents'][0]
            distances = res['distances'][0]
            metas = res['metadatas'][0] if res.get('metadatas') else [{}] * len(docs)
            return [{"text": doc, "score": 1.0 - dist, "source": meta.get('source', 'unknown')} for doc, dist, meta in zip(docs, distances, metas) if (1.0 - dist) > 0.35]
    except Exception:
        pass
    return []

def get_legal_rules(jurisdiction: str, domain: str):
    return legal_intelligence.get_rules(jurisdiction, domain)