chronix-client (Python)

Async Python client for the Chronix time-series database.

Features

• Async-first — built on httpx for non-blocking I/O
• JSON + Line Protocol writes — both formats supported
• Structured queries + SQL — first-class support for both query APIs
• PromQL — instant & range queries via Prometheus-compatible endpoints
• Arrow Flight SQL — zero-copy queries via ADBC driver
• Schema introspection — list measurements, get column schemas
• Multi-tenant — namespace header support
• Typed — full type annotations with py.typed marker

Installation

pip install chronix-client

With pandas support:

pip install "chronix-client[pandas]"

Quick Start

import asyncio
from chronix_client import ChronixClient, Point, TimeRange

async def main():
    async with ChronixClient("http://localhost:8086") as client:
        # Write
        await client.write([
            Point("cpu", {"usage": 42.5}, tags={"host": "a"})
        ])

        # Query
        result = await client.query(
            "cpu",
            TimeRange(start=0, end=2**63 - 1),
            tags={"host": "a"},
        )
        for row in result:
            print(row)

        # SQL
        result = await client.sql("SELECT * FROM cpu ORDER BY _time DESC LIMIT 10")
        df = result.to_dataframe()  # requires pandas extra

asyncio.run(main())

Exact Decimals

f64 cannot represent 0.1. For a meter register a bill is computed from, write a decimal.Decimal and the digits stay digits — in the request body, in storage, and in the query result:

from decimal import Decimal

await client.write([
    Point("meter", {"z1nb_q": Decimal("1234.5678")}, tags={"device": "main"})
])

The client sends {"z1nb_q": {"decimal": "1234.5678"}} rather than a JSON number, because a JSON number is parsed as a double at the other end. Line protocol uses the d suffix: meter,device=main z1nb_q=1234.5678d.

A decimal column stores a fixed number of fractional digits, set when the column is created; a value needing more is refused rather than rounded. Declare it first when the first value might not carry the digits you mean to keep — see Data Model.

Line Protocol Writes

await client.write_line_protocol([
    "cpu,host=a usage=42.5 1700000000000000000",
    "mem,host=a total=16384i,used=8192i 1700000000000000000",
])

PromQL Queries

# Instant query
result = await client.prom_query('cpu_usage{host="server-1"}')

# Range query
result = await client.prom_query_range(
    'rate(cpu_usage[5m])',
    start="2024-01-01T00:00:00Z",
    end="2024-01-02T00:00:00Z",
    step="60s",
)

# Label discovery
labels = await client.prom_labels()
values = await client.prom_label_values("host")

Arrow Flight SQL (Zero-Copy)

For high-throughput analytical queries, use Arrow Flight SQL via ADBC:

import adbc_driver_flightsql.dbapi

uri = ChronixClient.flight_sql_uri("localhost", 8817)
conn = adbc_driver_flightsql.dbapi.connect(uri)
cursor = conn.cursor()
cursor.execute("SELECT * FROM cpu WHERE _time > now() - INTERVAL '1 hour'")
table = cursor.fetch_arrow_table()
df = table.to_pandas()

Authentication & Multi-Tenancy

client = ChronixClient(
    "http://localhost:8086",
    api_key="your-api-key",
    namespace="production",
)

Idempotent Writes

await client.write(
    [Point("cpu", {"usage": 42.5})],
    idempotency_key="batch-2024-01-15-001",
)
# Second call with same key → WriteError (HTTP 409)

API Reference

ChronixClient

Models

Exceptions

License

Apache-2.0 — same as Chronix.