Time Series Reference

ArcadeDB’s native time series engine is documented in the Time Series Concepts section, which covers:

  • Schema — CREATE TIMESERIES TYPE syntax with all options (timestamp precision, tags, fields, shards, retention, compaction, block size)

  • Ingestion — Four ingestion methods: InfluxDB Line Protocol, Prometheus remote_write, SQL INSERT, Java Embedded API

  • SQL Functions — ts.timeBucket(), ts.rate(), ts.percentile(), ts.interpolate(), ts.delta(), ts.movingAvg(), ts.correlate(), ts.lag(), ts.lead(), ts.rowNumber(), ts.rank()

  • PromQL — Native PromQL parser and evaluator with HTTP endpoints for instant queries, range queries, and label discovery

  • Continuous Aggregates — Watermark-based incremental refresh with CREATE CONTINUOUS AGGREGATE

  • Retention & Downsampling — Automatic data lifecycle management with multi-tier downsampling

  • Integrity Checking — CHECK DATABASE coverage of the mutable bucket, tag dictionary and sealed store, the DEEP decoding tier, and what FIX may repair

  • Grafana Integration — DataFrame-compatible endpoints for the Infinity datasource plugin

  • HTTP API — 13 REST endpoints for ingestion, querying, PromQL, and Grafana

Quick Reference

Create a type

CREATE TIMESERIES TYPE SensorReading TIMESTAMP ts TAGS (sensor_id STRING) FIELDS (temperature DOUBLE)

Ingest (Line Protocol)

POST /api/v1/ts/{db}/write with SensorReading,sensor_id=s1 temperature=22.5 1708430400000000000

Ingest (SQL)

INSERT INTO SensorReading SET ts = '2026-02-20T10:00:00Z', sensor_id = 's1', temperature = 22.5

Query (SQL)

SELECT ts.timeBucket('1h', ts) AS hour, avg(temperature) FROM SensorReading GROUP BY hour

Query (PromQL)

GET /ts/{db}/prom/api/v1/query?query=avg(SensorReading{sensor_id="s1"})

Retention

CREATE TIMESERIES TYPE …​ RETENTION 90 DAYS

Downsampling

ALTER TIMESERIES TYPE …​ ADD DOWNSAMPLING POLICY AFTER 7 DAYS GRANULARITY 1 MINUTES

Continuous aggregate

CREATE CONTINUOUS AGGREGATE hourly_temps AS SELECT ts.timeBucket('1h', ts) AS hour, avg(temperature) FROM SensorReading GROUP BY hour

Integrity check

CHECK DATABASE TYPE SensorReading (add DEEP to decode every sealed block, FIX to repair the derived counters)

See Time Series Concepts for full documentation, examples, and architecture details.