Skip to main content

PostgreSQL

PostgreSQL is the core structured database of the Resultity platform, responsible for maintaining authoritative records across accounts, jobs, rewards, and network configuration.
It powers key components of the Network, Cloud, and Account services.

Purpose and Role​

PostgreSQL is used to store mission-critical metadata:

  • User accounts and associated wallets;
  • Node ownership and identity;
  • Inference job metadata (e.g., model, duration, result status);
  • Credit balances, deductions, and locks;
  • RCP reward history and epoch tracking;
  • Network configuration state, including model versioning;
  • Subcloud definitions and memberships.

All authoritative operations — from top-ups to node binding — are backed by PostgreSQL to ensure consistency, traceability, and auditability.

Structure and Queries​

The PostgreSQL layer is highly normalized and indexed for optimal performance:

  • Tables are structured by domain (e.g., accounts, jobs, credits, nodes, epochs);
  • Operations are mostly append-only, with historical snapshots for sensitive records;
  • Queries are optimized for common access patterns: job lookup, credit checks, node filtering, and reward computation.

The database supports cross-service access via role-based APIs and internal RPC bridges.

Snapshots and Rollups​

To support long-term integrity and performance:

  • Rollups are performed periodically, summarizing expired or completed jobs into compact aggregates;
  • Historical states are hashed and optionally exported;
  • These rollups serve as the basis for network snapshots, which can be archived or verified by external systems.

Snapshot features are designed for future support of auditable epochs, config diffs, and external cold storage.

Sync and Consistency​

PostgreSQL acts as the backbone for state synchronization:

  • Node state is validated against the latest records during config sync;
  • Cloud billing ensures credit deductions are committed before job dispatch;
  • Admin actions (e.g., model deprecation, Subcloud modifications) propagate from here to all services.

Consistency is enforced by central orchestrators via transaction-safe mechanisms.

Notes​

  • PostgreSQL is not used for storing raw job content — only metadata and accounting.
  • Structured logs and unstructured results are stored in MongoDB.
  • Caching and ephemeral state are delegated to Redis.