ruvector graph: --create/--query/--relate/--info all print success and do nothing (exit 0); GraphDatabase is never instantiated

[vidaunited]

Summary

Every subcommand of ruvector graph prints success-shaped output for work it never performs. --create says it is creating a node and creates nothing; --query echoes the Cypher and executes nothing; --relate is declared as an option but has no handler at all; --info advertises "Show graph info and stats" and prints a capability banner with no stats. All exit 0.

The command never instantiates GraphDatabase. It requires @ruvector/graph-node only to check that it is present, then falls through to console.log.

This matters because the failure is invisible: with the package correctly installed, a user gets Creating node: Person and an exit code of 0, and nothing has happened. Nothing distinguishes that from a working write.

Environment

ruvector 0.2.41 · @ruvector/graph-node 2.0.4 · node v22.23.0 · darwin arm64. ruvector doctor reports all checks passed, including ✓ @ruvector/graph-node installed.

Observed

$ ruvector graph --create Person --properties '{"name":"Alice"}'
  Creating node: Person
  Properties: { "name": "Alice" }
$ echo $?
0

Nothing is written — no file appears, and there is nowhere for it to go (see below).

$ ruvector graph --query "MATCH (n) RETURN n"
  Cypher Query: MATCH (n) RETURN n
  Note: Full Cypher execution requires running ruvector-server

$ ruvector graph --relate "a:KNOWS:b"
(no output beyond the banner — the option is parsed and ignored)

$ ruvector graph --info
  @ruvector/graph-node is available!
  Platform: darwin-arm64
  Available operations: --query / --create / --relate

--info's help string is "Show graph info and stats". It reports neither; it duplicates what doctor already prints.

Cause

In bin/cli.js, the graph action (around line 1770) handles each option with console.log only. The --query branch carries the comment:

// Actual implementation would execute the query

--relate has no if (options.relate) block at all.

Notably the code to do this correctly already exists in the same file — the demo --graph path (around line 2871) does const g = new GraphDatabase() and calls createNode / createEdge / query against it. The graph command could use the same approach.

A second, blocking gap

Even with handlers, the command has no way to address a stored graph. There is no --path / --db / --storage option, and:

const { GraphDatabase } = require('@ruvector/graph-node');
new GraphDatabase();   // constructor arity 0
// isPersistent() === false, getStoragePath() === null, stats() === {}

So any node created by the CLI would vanish at process exit, and --query would always run against an empty in-memory graph. A storage-path option looks necessary for the command to be useful at all, unless the intent is that everything goes through ruvector server.

Suggested resolution

Any of these would be an improvement over the current state, in descending order of preference:

1. Implement the subcommands against GraphDatabase, and add a --path option so the graph persists.
2. If Cypher genuinely requires ruvector server, make --query exit non-zero with that message instead of printing it as a note after echoing the query, and drop --create/--relate until they work.
3. At minimum, do not print Creating node: … when nothing is created, and fix --info's help text to match what it does.

Point 2 is the important one: a stub that exits 0 with encouraging output is worse than a missing feature, because it is indistinguishable from success in a script.

Unrelated minor bug spotted alongside: ruvector info prints Core Version: [object Object] — it interpolates core.version without calling it, since @ruvector/core exports it as a function. Called properly it returns 0.1.29.