Benchmark performance

[zmoshansky]

Could you consider running benchmarks about utility compared to GBrain and other memory systems that augment Hermes.

Especially considering features like multi-user support.

[ClaudioDrews]

This is on the roadmap. With v0.2.0 we shipped an automated installer and a smoke test suite — we now have the infrastructure to run reproducible benchmarks across different hardware and configurations without manual setup overhead.

Comparative benchmarks against GBrain and other memory systems are planned for the coming days as we start exercising the test harness. Multi-user support is a longer-term item but absolutely worth measuring early.

Will update this issue when results land.

[ClaudioDrews]

Honest answer: no benchmarks yet against GBrain or other memory systems. The project has a test suite (_test_sanitize.py, 24 tests + collapse/attestation tests in the Elyan Edition PR #18), but no comparative performance benchmarks.

This is a good idea — a proper benchmark would need:

• Standardized recall precision/recall across a shared query set
• Latency comparisons at comparable hardware profiles
• Multi-user isolation tests

Tagging as help-wanted if anyone in the community wants to contribute a benchmark harness. I'll also note this in the roadmap.