mahenzon

import requests

API_KEY = "EMPTY"
# BASE_URL = "http://localhost:12345/engines/llama.cpp/v1"
BASE_URL = "http://localhost:11223/v1"
URL = f"{BASE_URL}/chat/completions"
# MODEL = "ai/gemma3"
# MODEL = "ai/gemma3n"
MODEL = "gemma3:1b"

mahenzon

3.13

mahenzon

Examples with Redis Vector and ChromaDB

Embeddings via ru-en-RoSBERTa https://huggingface.co/ai-forever/ru-en-RoSBERTa

Text file: https://github.com/ai-forever/gigachain/blob/master/cookbook/sample_data/%D0%BC%D0%B0%D1%81%D1%82%D0%B5%D1%80_%D0%B8_%D0%BC%D0%B0%D1%80%D0%B3%D0%B0%D1%80%D0%B8%D1%82%D0%B0.txt

mahenzon

from functools import wraps
from timeit import default_timer
from typing import reveal_type, Callable, Concatenate

type ExecutionTime = float


def with_execution_time[**P, T](
    func: Callable[P, T],
) -> Callable[P, tuple[T, ExecutionTime]]:

mahenzon

import logging

# DEFAULT_LOG_FORMAT = "[%(asctime)s.%(msecs)03d] %(funcName)20s %(module)s:%(lineno)d %(levelname)-8s - %(message)s"
DEFAULT_LOG_FORMAT = (
    "%(funcName)10s %(module)s:%(lineno)d %(levelname)-8s - %(message)s"
)


def configure_logging(level: int = logging.INFO) -> None:
    logging.basicConfig(

mahenzon

ChatGPT 5 solution after several attempts and corrections.

You're right—just setting _server_version_info isn’t enough; the MSSQL dialect caches feature flags at init. We need to refresh those flags so the compiler knows OFFSET/FETCH is supported.

Use this function (note the call to _setup_version_attributes) and make sure your statement has ORDER BY and .offset(...):

from typing import Optional, Tuple, Union

mahenzon

import os

TOKEN = os.getenv(
    "BOT_TOKEN",
    "123:abc",
)

mahenzon

mahenzon

mahenzon