brainy

soulcraft/brainy

Fork 0

Commit graph

Author	SHA1	Message	Date
David Snelling	0ddc05a5bb	feat: enforce data/metadata separation, numeric range queries, improved docs - Store data opaquely in add() and update() instead of spreading object properties into top-level metadata. data is for semantic search (HNSW), metadata is for structured where-filter queries (MetadataIndex). - Fix numeric range queries in MetadataIndex — use numeric-aware comparison instead of lexicographic string comparison for normalized values. - Add data field to RelateParams and Relation types for relationship content. - Add where.type → where.noun alias in metadata-only find() path. - Rewrite README: focused ~350 lines from 791, quick start first, feature showcase with mini-snippets, organized doc links, no version callouts. - Add DATA_MODEL.md and QUERY_OPERATORS.md reference docs. - Remove 10 outdated/redundant doc files consolidated into API reference. - Improve JSDoc on Entity, Relation, AddParams, FindParams, and core methods. - Fix tests asserting data properties appear in metadata (data model violation). - Deprecate verb.source/target in favor of from/to (public) and sourceId/targetId (storage).	2026-02-09 12:07:54 -08:00
David Snelling	0f3a88429d	feat: add SQ8 vector quantization, lazy loading, and two-phase rerank to HNSW - SQ8 scalar quantization (8-bit) for 4x vector storage reduction - Lazy vector loading: evict float32 vectors after graph construction, load on-demand from storage via UnifiedCache - Two-phase search: over-retrieve with SQ8 approximate distances, rerank top candidates with exact float32 distances - Configuration surface: hnsw.quantization and hnsw.vectorStorage in BrainyConfig - All features disabled by default (zero behavior change for existing users) - 27 new tests covering quantization accuracy, lazy loading, reranking - Remove GitHub Actions CI (build locally, cortex CI handles native builds)	2026-01-31 12:41:53 -08:00

Author

SHA1

Message

Date

David Snelling

0ddc05a5bb

feat: enforce data/metadata separation, numeric range queries, improved docs

- Store data opaquely in add() and update() instead of spreading object
  properties into top-level metadata. data is for semantic search (HNSW),
  metadata is for structured where-filter queries (MetadataIndex).
- Fix numeric range queries in MetadataIndex — use numeric-aware comparison
  instead of lexicographic string comparison for normalized values.
- Add data field to RelateParams and Relation types for relationship content.
- Add where.type → where.noun alias in metadata-only find() path.
- Rewrite README: focused ~350 lines from 791, quick start first, feature
  showcase with mini-snippets, organized doc links, no version callouts.
- Add DATA_MODEL.md and QUERY_OPERATORS.md reference docs.
- Remove 10 outdated/redundant doc files consolidated into API reference.
- Improve JSDoc on Entity, Relation, AddParams, FindParams, and core methods.
- Fix tests asserting data properties appear in metadata (data model violation).
- Deprecate verb.source/target in favor of from/to (public) and sourceId/targetId (storage).

2026-02-09 12:07:54 -08:00

David Snelling

0f3a88429d

feat: add SQ8 vector quantization, lazy loading, and two-phase rerank to HNSW

- SQ8 scalar quantization (8-bit) for 4x vector storage reduction
- Lazy vector loading: evict float32 vectors after graph construction,
  load on-demand from storage via UnifiedCache
- Two-phase search: over-retrieve with SQ8 approximate distances,
  rerank top candidates with exact float32 distances
- Configuration surface: hnsw.quantization and hnsw.vectorStorage in BrainyConfig
- All features disabled by default (zero behavior change for existing users)
- 27 new tests covering quantization accuracy, lazy loading, reranking
- Remove GitHub Actions CI (build locally, cortex CI handles native builds)

2026-01-31 12:41:53 -08:00

2 commits