Storage and Local Data
Each Resultity node maintains a minimal local storage layer to support offline stability, reduce network dependency, and ensure execution traceability. This storage is scoped to operational needs and is not used for long-term history or reward tracking — those functions are handled server-side.
Purpose
Local storage allows the node to:
- Operate temporarily without an active connection;
- Resume execution after restart;
- Verify model integrity and configuration versions;
- Persist operational state across container reloads.
The storage is structured to separate short-lived data (e.g., job queue) from semi-persistent data (e.g., configuration snapshot).
Main Data Types
| Category | Description |
|---|---|
| Jobs | Incomplete jobs not yet acknowledged by the network; stored with TTL |
| Configuration Snapshot | Latest applied node_config, with model list, policies, and digest |
| Hashes and Versions | Recorded hashes of containers and models to detect drift |
| Logs and Status | Runtime metrics, errors, and diagnostic messages |
| Account Binding | Node's EVM key, node ID, account link metadata |
All data is stored in a local volume attached to the Agent container. Runtime containers are stateless.
Snapshot Rotation
Some data is periodically rotated based on:
- TTL defined in the configuration;
- Network-issued snapshot version changes;
- Manual node restart or container update.
For example, job traces from the previous configuration snapshot are deleted upon applying a new snapshot.
Logging Behavior
Logs are recorded locally and may include:
- Inference execution durations;
- Model loading results;
- Runtime error messages;
- Configuration application status.
These logs are periodically flushed to the network (or cloud) and removed from local storage to avoid unbounded growth.
Security and Scope
- Local data does not include credit balances, reward records, or token state;
- Sensitive keys (EVM private key) are stored securely and never exposed to containers;
- All local records are tied to a single network (Testnet or Mainnet), and are maintained separately.
Resultity nodes maintain just enough local state to remain robust and auditable — while deferring all critical coordination, validation, and billing logic to the network.