search.py

# 1. IMPORTS
import warnings

from dotenv import load_dotenv
load_dotenv()

from transformers import logging as hf_logging
hf_logging.set_verbosity_error()
hf_logging.disable_progress_bar()

warnings.filterwarnings("ignore")

from sentence_transformers import SentenceTransformer
from databricks.vector_search.client import VectorSearchClient

# 2. CONFIG
INDEX_NAME = "dv_projc.live_chat.messages_feelings_index"
ENDPOINT_NAME = "hate-or-love-sv"

# 3. MODELO
model = SentenceTransformer('paraphrase-multilingual-MiniLM-L12-v2')

# 4. CLIENTE
client = VectorSearchClient(disable_notice=True)
index = client.get_index(ENDPOINT_NAME, INDEX_NAME)

# 5. BUSCA
query = "quero matar todo mundo"
query_embedding = model.encode([query])[0].tolist()

results = index.similarity_search(
    query_vector=query_embedding,
    columns=["id", "text"],
    num_results=3,
    disable_notice=True
)

# 6. RESULTADOS
print("\nResultados:")

for r in results.get("result", {}).get("data_array", []):
    print(f"ID: {r[0]} | Texto: {r[1]}")