Output A — Network State Snapshot

Hourly snapshot of the correlation network: which coins move together, how each cluster is evolving, and where capital is concentrated.

Cadence

Hourly, aligned to the bar boundary.

Schema

The pydantic model is memenet.schemas.network_state.NetworkStateSnapshot.

Example payload:

{
  "timestamp": "2026-05-02T14:00:00Z",
  "universe": ["BONK", "WIF", "POPCAT"],
  "clusters": [
    {
      "id": "cluster_0",
      "label_auto": null,
      "members": ["BONK", "WIF"],
      "internal_correlation_mean": 0.72,
      "internal_correlation_trend_24h": 0.08,
      "cluster_return_24h": 0.34,
      "cluster_volume_share_24h": 0.18,
      "state": "hot"
    }
  ],
  "cluster_count": 1,
  "concentration_index": 0.41
}

How to consume

Snapshots are written to partitioned Parquet under outputs/network_state/ and queryable via DuckDB. Downstream consumers should depend only on the pydantic model — the storage layout may evolve.