[SLICE 26][CORE-H] Add backend migration + scale benchmark harness for append, query, rebuild, projection, provenance, and ingestion #5

New issue

Open

opened 2026-09-08 01:56:59 +00:00 by nsaspy · 0 comments

nsaspy commented

2026-09-08 01:56:59 +00:00

Owner

Parent epic: #19
Depends on: #37 #38 #43

Goal

Measure the real performance envelope and prove backend migration without changing semantic meaning.

Implement:

bootstrap backend -> production backend export/import/migration;
deterministic semantic fingerprint comparison before/after migration;
synthetic corpora at increasing sizes (at least 10^3, 10^5, 10^6 records where practical in CI/local benchmark lanes);
append throughput/latency;
direct lookup and bounded query latency;
projection/update invalidation cost;
cold rebuild/checkpoint restore time;
provenance/explanation expansion cost;
ingestion/backfill throughput and restart cost;
memory/disk/index-size measurements;
benchmark profile/version/hardware metadata;
explicit unsupported/resource-limited thresholds.

Acceptance

Migration preserves canonical semantic/history fingerprints and normalized queries; benchmark report distinguishes measured values from estimates; no billion-record or latency claim is made unless actually measured; regression thresholds can be run reproducibly outside provider/model paths.

Parent epic: #19 Depends on: #37 #38 #43 ## Goal Measure the real performance envelope and prove backend migration without changing semantic meaning. Implement: - bootstrap backend -> production backend export/import/migration; - deterministic semantic fingerprint comparison before/after migration; - synthetic corpora at increasing sizes (at least 10^3, 10^5, 10^6 records where practical in CI/local benchmark lanes); - append throughput/latency; - direct lookup and bounded query latency; - projection/update invalidation cost; - cold rebuild/checkpoint restore time; - provenance/explanation expansion cost; - ingestion/backfill throughput and restart cost; - memory/disk/index-size measurements; - benchmark profile/version/hardware metadata; - explicit unsupported/resource-limited thresholds. ## Acceptance Migration preserves canonical semantic/history fingerprints and normalized queries; benchmark report distinguishes measured values from estimates; no billion-record or latency claim is made unless actually measured; regression thresholds can be run reproducibly outside provider/model paths.

nsaspy referenced this issue

2026-09-08 01:57:01 +00:00

[SLICE 04][CORE-D] Implement Prolog-RLM semantic compiler adapter and independent projection lifecycle #27

nsaspy referenced this issue

2026-09-08 01:57:02 +00:00

[EPIC][CORE-D] Semantic projection + corpus ingestion: compiler adapter, projection generations, structured adapters, resumable backfill, and artifact lineage #34

nsaspy referenced this issue

2026-09-08 01:57:03 +00:00

[EPIC][CORE] Library-first Symbolic Memory v1: durable semantic world model, learning history, federation, and zero-model recall #38

nsaspy referenced this issue

2026-09-08 01:57:03 +00:00

[semantic-memory] Expose general semantic ingestion, query, provenance, and bounded world-model projections through native/MCP tools #40