Redis time-series regularizer: snaps jittery streams to a UTC grid, imputes or forecasts gaps
98
These images are currently intended for testing only and should not be used for production. We will remove this warning and add some documentation as soon as the images are ready for a broader evaluation.
Reads measurement streams from Redis with jittery timestamps and gaps, snaps samples onto a UTC grid (epoch + k * update_interval), fills missing steps by imputation or forecast, and writes a strictly regular output stream.
Each grid point is finalized once, in order, as measured, imputed, or forecast. Late measured values can re-emit corrections for points still in the in-memory window; the frontier never rewinds. Each configured channel runs in its own worker thread. Optional TimescaleDB bootstrap replays history so the grid does not start empty.
Configuration: YAML file, path via -c/--config or REGULARIZER_CONFIG; optional dotenv via --env or REGULARIZER_ENV. Values support !env-template "${VAR}". Sections: redis, timescale (required only if a channel uses history_provider), channels.<name>.
Input: Redis stream entries with JSON _time and _value (scalar or equal-length arrays), optional _metadata. Cursor starts at the stream tip.
Output: one XADD per poll batch with JSON arrays valid_time, value, quality.
Source, full config reference, and architecture diagrams: https://github.com/AIT-RDP/rdp-regularizer
License: MIT-0.
Content type
Image
Digest
sha256:3e0bc82bf…
Size
97.6 MB
Last updated
9 days ago
docker pull ait1/rdp-regularizer