hooks force-learn crashes with `TypeError: intel.tick is not a function` (v0.2.27 + main HEAD)

[KevinAIP]

Summary

npx ruvector@0.2.27 hooks force-learn exits with code 1 and TypeError: intel.tick is not a function at npm/packages/ruvector/bin/cli.js:4420. The command is unusable. Confirmed still present on main (HEAD as of 2026-06-02) and in npm latest (0.2.27) — the code at that line is unchanged.

Root cause (confirmed)

The force-learn action constructs the lightweight, engine-disabled Intelligence wrapper and then calls intel.tick():

// npm/packages/ruvector/bin/cli.js
hooksCmd.command('force-learn')
  .description('Force an immediate learning cycle')
  .action(() => {
    const intel = new Intelligence({ skipEngine: true });  // Fast mode
    intel.tick();   // <-- line 4420: TypeError
    console.log(JSON.stringify({ success: true, result: 'Learning cycle triggered', stats: intel.stats() }));
  });

This is doubly wrong:

1. tick() / forceLearn() exist only on the native IntelligenceEngine (dist/core/sona-wrapper.js, dist/core/intelligence-engine.js) — not on the cli.js Intelligence JSON wrapper, which has no tick().
2. { skipEngine: true } guarantees the engine is never initialized, so even getEngineIfReady() would return null here.

By contrast, the correct pattern is used elsewhere in the same file — e.g. sessionEnd() does const eng = this.getEngineIfReady(); if (eng) eng.forceLearn();.

The engine itself works

Constructing the engine directly and calling forceLearn() works fine:

const { IntelligenceEngine } = require('ruvector/dist/core/intelligence-engine.js');
const e = new IntelligenceEngine({ enableSona: true, enableOnnx: false });
e.forceLearn();
// => "Forced learning: 0 trajectories -> 0 patterns, status: skipped: insufficient trajectories"

So this is purely a CLI wiring bug, not an engine failure. (hooks stats and mcp start are unaffected.)

Environment

• ruvector: 0.2.27 (npm latest) and main HEAD
• node: v24.15.0
• OS: macOS (Darwin 25.4.0)
• Install: npx ruvector@0.2.27 …

Stack trace

.../node_modules/ruvector/bin/cli.js:4420
    intel.tick();
          ^
TypeError: intel.tick is not a function
    at Command.<anonymous> (.../ruvector/bin/cli.js:4420:11)
    at Command.listener [as _actionHandler] (.../commander/lib/command.js:494:17)
    ...
Node.js v24.15.0

Suggested fix

Make the force-learn action use the engine the way sessionEnd() does — initialize the engine, then call forceLearn():

.action(() => {
  const intel = new Intelligence();      // engine enabled
  const eng = intel.getEngine();         // full init (imports existing data)
  const result = eng ? eng.forceLearn() : 'engine unavailable';
  intel.save();
  console.log(JSON.stringify({ success: true, result, stats: intel.stats() }));
});

Impact

A nightly automation calls hooks force-learn as a best-effort consolidation step. The rc=1 was suppressed for ~a week as a "known issue" — which also masked an unrelated silent reset of intelligence.json during a v0.1.x → v0.2.x bump (~525 memories / 439 trajectories lost). The two together meant the learning layer was effectively non-functional for several days with no signal beyond an rc=1 that was being ignored. A crash here is especially costly because it hides other failures.