Input and storage
Read-only telemetry enters the local environment and is stored on-site. A pilot needs a defined signal schema, quality checks, sampling expectations and retention policy.
RAIN / Software
Follow the signal.Understand the system.
An architecture guide to RAIN’s local telemetry, inference, evaluation and audit-history components, with current implementation boundaries.
Architecture documentation · simulation stage
INTELLIGENCE BELONGS WHERE THE DATA LIVES.
01 / THE IDEA
One local loop, distinct responsibilities.
The current project describes telemetry capture, local storage, inference, retraining, promotion, rollback and audit components running against a battery-storage simulator. This is a guide to those responsibilities. It is not a released API reference or a production deployment manual.
Inference and learning
The incumbent runs inference while retraining produces candidate models. Held-out evaluation gates promotion so retraining does not automatically replace the active model.
History and review
A signed, linked history records inferences and model changes, including rejected candidates. The local view presents advisories and model context for an operator.
02 / IN PRACTICE
Put the thinking into practice.
Define the scope, preserve the boundary, and make the outcome something a person can inspect.
- 01
Trace the input boundary
Identify where source measurements enter, what can be read and where raw data is retained. Equipment commands are outside this architecture.
- 02
Trace the model lifecycle
Follow the incumbent, training data, candidate evaluation, promotion decision and rollback path. Treat each transition as something to verify.
- 03
Trace the evidence boundary
Determine what an advisory can show and what its recorded history can support. Production uplink transport and hardware-backed signing remain unfinished work.
03 / A FEW DETAILS
Worth understanding.
Does the runtime require a cloud connection?
The local inference and learning loop is designed to operate without runtime network egress. The simulator uplink is a local no-op, and production transport is not yet implemented.
Where is the production deployment guide?
A production deployment manual is not published. Hardware configuration, equipment interfaces and operating procedures still need to be scoped and verified through a first field pilot.
BUILD WITH RAIN
One site. One question.
A useful next step.
RAIN is preparing for its first field pilot. Start by defining the environment, the data boundary, and what success should look like.