The update-integrity suite (bench/update_integrity.py, shipped repo-side with v0.2.3, results appendix in docs/competitive-landscape.md) has a designed-but-unimplemented comparison arm: run the same ten supersession scenarios against mem0 OSS for a side-by-side table.
Design (frozen in the WP2 plan, docs/superpowers/plans/2026-07-29-wp2-update-integrity.md Task 5):
--arm mem0 flag; default arm unchanged.- Adapter maps
Step.text → mem0.Memory.add and scenario queries → their search; assertion names stay identical so tables compare cell-for-cell.
import mem0 failure → exit 2 with install hint, never a silent skip.- Emitted header MUST pin
mem0.__version__ + the configured LLM backend.
Prerequisites: maintainer go-ahead + a configured mem0 LLM path (Ollama or API key). Publishing rule: nothing beyond reproduced, versioned behavior.
The update-integrity suite (
bench/update_integrity.py, shipped repo-side with v0.2.3, results appendix indocs/competitive-landscape.md) has a designed-but-unimplemented comparison arm: run the same ten supersession scenarios against mem0 OSS for a side-by-side table.Design (frozen in the WP2 plan,
docs/superpowers/plans/2026-07-29-wp2-update-integrity.mdTask 5):--arm mem0flag; default arm unchanged.Step.text→mem0.Memory.addand scenario queries → their search; assertion names stay identical so tables compare cell-for-cell.import mem0failure → exit 2 with install hint, never a silent skip.mem0.__version__+ the configured LLM backend.Prerequisites: maintainer go-ahead + a configured mem0 LLM path (Ollama or API key). Publishing rule: nothing beyond reproduced, versioned behavior.