lantern-ops

Use lantern-cli get vertex <key> to fetch a vertex and lantern-cli put vertex <key> <value> [ttl_seconds] to upsert one. Values are auto-typed (int, float, bool, datetime, string) and can be forced with a type= token; reads print JSON and writes print OK.

add edge is additive: repeated calls sum the weight on the same (tail, head) pair, suiting streaming counts and interaction signals. put edge replaces the weight wholesale and is idempotent, suiting measured properties like similarity or distance.

Run lantern-cli bfs <seed>, lantern-cli pagerank <seed>, or lantern-cli community <seed> to execute a read-only graph walk. Each returns the visited subgraph as JSON with vertices and edges, tunable via flags like --step, --fan-out, --top-n, and --weighting.

Yes. lantern-cli bulk streams NDJSON files into batched vertex or edge writes, while lantern-cli dump and restore capture and replay a consistent whole-graph snapshot in protobuf or NDJSON format. Already-sent batches are not rolled back on error.

The command is intentionally gated because it bulk-deletes every live vertex under a prefix. You must pass dry_run=true to preview the match count or confirm=yes to execute the deletion; otherwise it exits non-zero with suggested next steps.

Do not use it for editing Lantern's Go source code, regenerating protobuf or wire code, server deployment and infrastructure-as-code, or the MCP working-context tools. It covers only data operations against a running Lantern server via lantern-cli.