Export an rrlm_graph to a SQLite database

Writes nodes, edges, graph metadata, task traces, and the TF-IDF vocabulary to a SQLite file readable directly by better-sqlite3 in the TypeScript MCP server. All operations are idempotent: nodes/edges/metadata are replaced on each call; task-trace rows are inserted with INSERT OR IGNORE (deduplication on query, session_id, created_at).

Usage

export_to_sqlite(graph, db_path)

Arguments

graph

An rrlm_graph / igraph object.

db_path

Character(1). Path to the .sqlite file to create or update.

Value

db_path, invisibly.

Schema

nodes

node_id (PK), name, file, node_type, signature, body_text, roxygen_text, complexity, pagerank, task_weight, embedding (JSON text array of floats), pkg_name, pkg_version

edges

edge_id (PK AUTOINCREMENT), source_id, target_id, edge_type, weight, metadata (JSON)

task_traces

trace_id (PK AUTOINCREMENT), query, nodes_json, polarity, session_id, created_at. Unique on (query, session_id, created_at) to prevent duplicate imports.

tfidf_vocab

term (PK), idf, doc_count, term_count. Only populated when embed_method = "tfidf". Allows the TypeScript MCP server to encode queries in the same vector space without calling back into R.

graph_metadata

key (PK), value

Embedding format

The embedding column in nodes is stored as a JSON text array of floating-point numbers (e.g. [0.12, -0.34, ...]). In TypeScript, decode with JSON.parse(row.embedding || "[]").

See also

save_graph_cache()

Examples

if (FALSE) { # \dontrun{
g <- build_rrlm_graph("mypkg")
export_to_sqlite(g, "mypkg.sqlite")
} # }