Persistence Operations¶

Methods for saving, loading, and streaming knowledge graph state.

save()¶

Serialize the entire knowledge graph to a JSON file.

r.save(path: str) -> None

The JSON file contains the complete graph state: all nodes, edges, senses, autonomy records, co-occurrence statistics, and configuration parameters. The file is self-contained and can be loaded in a different session.

load()¶

Class method. Deserialize a knowledge graph from a JSON file.

r = Rsvs.load(path: str) -> Rsvs

Reconstructs the full graph state including all Rust-side data structures. The loaded instance is identical to the one that was saved.

snapshot_v1()¶

Runtime snapshot for UI consumption.

json_string = r.snapshot_v1() -> str

Returns a JSON string containing the current graph state in the v1 snapshot format. This is used by the frontend for 3D visualization and is designed for low-latency consumption.

consume_events_v1()¶

Incremental event stream. Returns events with sequence numbers after the specified position.

events_json = r.consume_events_v1(
    after_seq: int | None = None,
    limit: int = 100,
) -> str

Parameters¶

Events are consumed (one-time read) from the Rust core's event buffer. Each event includes a monotonic sequence number, correlation ID, timestamp, event type, and payload.

latest_seq_v1()¶

Current monotonic event sequence number.

seq = r.latest_seq_v1() -> int

Use this to track the event stream position. Pass the result as after_seq in subsequent calls to consume_events_v1() to get only new events.

Event Schema¶

{
    "api_version": "v8.3",
    "schema_version": "v8.3",
    "seq": 42,
    "correlation_id": "corr_abc123",
    "event_type": "node_created",
    "payload": { "id": 25, "label": "batu" }
}

Event Types¶