YTsaurus Python Client

A lightweight Python helper library for day-to-day work with YTsaurus - https://ytsaurus.tech YQL CHYT and pandas DataFrames

PyPI · GitHub · Example notebook · Issues

The project wraps common analytics workflows into a small, readable API:

• run YQL queries and return results as pandas.DataFrame
• start long-running YQL queries without blocking the notebook
• write YQL results directly into YTsaurus tables
• read large query outputs through temporary YTsaurus tables with progress reporting
• execute CHYT queries through HTTP or the YTsaurus CLI
• upload pandas DataFrames into YTsaurus tables

This repository is designed as a clean portfolio-friendly version of the client: no company-specific hosts, pools, paths, tokens, or internal links are hardcoded

Installation

pip install ytsaurus_python_client

pip install -e .

For production packaging:

python -m build
pip install dist/ytsaurus_python_client-*.whl

Requirements

• Python 3.9+
• pandas
• requests
• numpy
• YTsaurus Python client with yt.wrapper
• Optional: YTsaurus CLI binary yt for CLI-based CHYT helpers

Configuration

The library is configured through environment variables or explicit constructor arguments.

Example:

export YT_PROXY="your-ytsaurus-proxy.example.com"
export YT_TOKEN_PATH="$HOME/.yt/token"
export YT_DEFAULT_TEMP_DIR="//home/your-login/tmp"
export CHYT_CLIQUE_ALIAS="ch_public"

Quick start

Run a YQL query

from ytsaurus_python_client import YTsaurusHook

hook = YTsaurusHook(
    yt_proxy="your-ytsaurus-proxy.example.com",
    yt_query_result_temp_dir="//home/your-login/tmp",
)

df = hook.yql("""
SELECT
    1 AS id,
    "hello" AS value;
""")

print(df)

Materialize a large YQL result into a temp table and read it in chunks

df = hook.yql_unlim(
    """
    SELECT *
    FROM `//home/your-login/large_table`;
    """,
    chunksize=500_000,
)

Upload a DataFrame to YTsaurus

import pandas as pd

from ytsaurus_python_client import YTsaurusHook

hook = YTsaurusHook(yt_proxy="your-ytsaurus-proxy.example.com")

df = pd.DataFrame({"id": [1, 2], "name": ["Alice", "Bob"]})
schema = hook.generate_yt_schema(df)

hook.upload_df_to_yt(
    df=df,
    yt_path="//home/your-login/users",
    schema=schema,
    overwrite=True,
)

Run a CHYT query over HTTP

from ytsaurus_python_client import chyt_df

df = chyt_df(
    """
    SELECT 1 AS ok
    """,
    host="your-chyt-host.example.com",
    clique_alias="ch_public",
)

Public API

from ytsaurus_python_client import (
    YTsaurusHook,
    DOYTHook,          # backward-compatible alias
    chyt_df,
    chyt_raw,
    chyt_to_yt,
)

Design notes

• Defaults are intentionally generic and safe for public repositories
• Secrets are never hardcoded. Use YT_TOKEN, YT_TOKEN_PATH, or explicit arguments
• Printed YTsaurus UI links are optional and controlled by YT_UI_BASE_URL
• YQL pragmas can be provided through query_pragma_config or environment variables such as YT_POOL
• DOYTHook is kept as a backward-compatible alias; new code should prefer YTsaurusHook

Repository hygiene

Before publishing, the project was cleaned from:

• macOS metadata files
• Python cache files
• internal company hosts and UI links
• internal pools and temp paths
• Russian comments and runtime messages
• local tokens or secret values

License

MIT © 2026 Alexey Voronko