import sqlite3, json, os, torch
from sentence_transformers import SentenceTransformer, util

BASE_DIR = os.path.abspath(os.path.dirname(__file__))
DB_PATH  = os.path.join(BASE_DIR, "platform.db")

conn = sqlite3.connect(DB_PATH)
conn.row_factory = sqlite3.Row

print("=== SAMPLE ENTRIES (first 15, org_id=2) ===")
rows = conn.execute(
    "SELECT id, question, processus, intent, response FROM entries WHERE org_id=2 ORDER BY id LIMIT 15"
).fetchall()
for r in rows:
    print(f"[{r['id']}] Q: {r['question']}")
    print(f"     processus={r['processus']} | intent={r['intent']}")
    print(f"     R: {(r['response'] or '')[:100]}")
    print()

print("=== LIVE SEARCH TEST ===")
model = SentenceTransformer("paraphrase-multilingual-MiniLM-L12-v2", device="cpu")

test_queries = [
    "comment bloquer un paiement",
    "ouvrir un compte bancaire",
    "faire une reclamation",
    "demande de credit",
    "carte bancaire perdue",
]

all_rows = conn.execute(
    "SELECT id, question, processus, intent, response FROM entries WHERE org_id=2 ORDER BY id"
).fetchall()

questions = [r["question"] for r in all_rows]
emb_base  = model.encode(questions, convert_to_tensor=True, normalize_embeddings=True, batch_size=64, show_progress_bar=False)

for q in test_queries:
    emb_q  = model.encode(q, convert_to_tensor=True, normalize_embeddings=True)
    scores = util.pytorch_cos_sim(emb_q, emb_base)[0]
    top3   = torch.topk(scores, 3).indices.tolist()
    print(f"QUERY: '{q}'")
    for idx in top3:
        r = all_rows[idx]
        print(f"  [{float(scores[idx])*100:.1f}%] {r['question']} | intent={r['intent']}")
        print(f"         R: {(r['response'] or '')[:80]}")
    print()