Benchmark WaveMind against real LoCoMo memory systems

[CaspianG]

Summary

• add real Mem0 OSS 2.0.14 and Hindsight OSS 0.8.5 adapters to the shared LoCoMo evidence protocol
• publish exact-commit results for 5,882 memories and 1,977 evidence queries alongside WaveMind, Memory OS, Chroma, and Qdrant
• expose query latency and ingest scope separately, require native-version/profile/provenance metadata, and prevent local dataset paths from entering public artifacts
• refresh the benchmark registry, reports, leaderboard, dashboard, and README links

Measured result

Engine	Recall@5	Precision@1	MRR@5	Avg query
WaveMind	0.548	0.333	0.432	4.88 ms
WaveMind + Memory OS	0.548	0.332	0.431	5.99 ms
Mem0 OSS	0.500	0.263	0.369	270.24 ms
Chroma static	0.408	0.219	0.305	4.12 ms
Qdrant static	0.409	0.219	0.305	103.27 ms
Hindsight OSS	0.316	0.052	0.148	320.61 ms

This is retrieval evidence on one local machine, not final answer quality or hosted-service throughput. Internal systems start ingest from shared precomputed embeddings; Mem0 and Hindsight ingest includes their native embedding and persistence work.

Validation

• python -m pytest -q -> 887 passed, 5 skipped
• python -m ruff check wavemind tests benchmarks examples -> passed
• benchmark artifact audit -> pass (45 implemented, 3 runner-ready, 5 planned)
• python -m build -> passed
• python -m twine check dist\\* -> passed
• public evidence scan -> no secrets or local absolute paths