Data-write
River Forecaster
Most forecasting setups demand a separate pipeline. The River Forecaster plugin for InfluxDB 3 eliminates that entirely: a single WAL trigger learns incrementally from every incoming observation, produces multi-step-ahead forecasts automatically, and writes predictions back to your database as queryable time series data. No batch retraining, no external serving layer, no separate configuration per series.
Configuration
Arguments can be passed inline (space-separated lists) or via TOML config file (native TOML lists). Parameter names are the same in both modes.
| Argument | Required | Default | Description |
|---|---|---|---|
include_tables |
Yes | — | Space-separated tables to forecast (e.g., system_cpu system_memory). In TOML config use a list. |
include_fields |
No | (all) | Space-separated fields to forecast (e.g., idle used). Omit to use all numeric fields. In TOML config use a list. |
string_fields |
No | (none) | Space-separated string column names that should be treated as fields, not tags. All other string columns become tags. In TOML config use a list. |
max_series |
No | 50 | Max unique series to track (LRU eviction) |
min_observations |
No | 30 | Minimum observations before forecasting |
default_horizon |
No | 12 | Fallback horizon if the model’s own write_interval estimate is unavailable |
forecast_target_seconds |
No | 3600 | Target forecast window for auto-horizon |
snarimax_p |
No | 0 (auto) | SNARIMAX autoregressive order. 0 = auto-tune from write frequency (~1h lookback, clamped 4-30) |
snarimax_d |
No | 1 | SNARIMAX differencing order |
snarimax_q |
No | 1 | SNARIMAX moving average order |
calibration_count |
No | 30 | Observations buffered before finalizing SNARIMAX p (only used when snarimax_p=0). Larger = more reliable interval estimate, longer warm-up. |
log_forecasts |
No | false | Log forecast details |
checkpoint_interval_seconds |
No | 1800 | Seconds between model checkpoints to the database |
max_checkpoint_age_hours |
No | 24 | Ignore checkpoints older than this when restoring |
min_checkpoint_observations |
No | 10 | Skip checkpointing models with fewer observations than this (avoids persisting cold-start models) |
config_file_path |
No | — | Path to TOML config file |
Ready to get started?
Download InfluxDB 3 and have running in minutes.
