Skip to content

feat(bgz17): RaBitQ encoding, PCDVQ weighted distance, sigma-band palette D1: rabitq_compat.rs — RaBitQEncoding struct with: - OrthogonalMatrix (Hadamard rotation) - encode(): normalize → rotate → sign-quantize → binary + corrections - distance_rabitq(): unbiased L2² estimate with correction factors - distance_palette(): O(1) matrix lookup - distance_corrected(): generative decompression with LFD correction D6: Base17::l1_weighted() — PCDVQ direction-weighted L1 distance. dim 0 (sign) weighted 20×, dims 1-6 (exponent) 3×, dims 7-16 (mantissa) 1×. DistanceMatrix::build_pcdvq_weighted() uses weighted distances. D7: Palette::from_sigma_bands() — training-free codebook from empirical distribution. Equal-percentile bands = GQ-style guaranteed uniform utilization. No empty clusters by construction. 114 tests pass (was 108), 6 new tests for the added features. https://claude.ai/code/session_01CdqyUTUfjKZuk8YGJzv6LB#33

Merged
AdaWorldAPI merged 1 commit into
mainfrom
claude/continue-session-0mAVa
Mar 22, 2026

Conversation

@AdaWorldAPI
Copy link
Copy Markdown
Owner

No description provided.

…ette

D1: rabitq_compat.rs — RaBitQEncoding struct with:
  - OrthogonalMatrix (Hadamard rotation)
  - encode(): normalize → rotate → sign-quantize → binary + corrections
  - distance_rabitq(): unbiased L2² estimate with correction factors
  - distance_palette(): O(1) matrix lookup
  - distance_corrected(): generative decompression with LFD correction

D6: Base17::l1_weighted() — PCDVQ direction-weighted L1 distance.
  dim 0 (sign) weighted 20×, dims 1-6 (exponent) 3×, dims 7-16 (mantissa) 1×.
  DistanceMatrix::build_pcdvq_weighted() uses weighted distances.

D7: Palette::from_sigma_bands() — training-free codebook from empirical
  distribution. Equal-percentile bands = GQ-style guaranteed uniform
  utilization. No empty clusters by construction.

114 tests pass (was 108), 6 new tests for the added features.

https://claude.ai/code/session_01CdqyUTUfjKZuk8YGJzv6LB
@AdaWorldAPI AdaWorldAPI merged commit 6f5f4c0 into main Mar 22, 2026
AdaWorldAPI pushed a commit that referenced this pull request May 16, 2026
…route mirror from WoA PR #33

User pointed at WoA PR #33 (merged 2026-05-16) which already harvested
the full WoA routing surface — 514 routes across `app.py` (91 routes
in the monolith) + 53 blueprint modules under `woa/blueprints/*.py`.

The prior commit in this branch (87f5076) harvested only the
2,904-line app.py via a local AST walker and captured 96 routes —
the monolith remnant only, missing the entire modular blueprint
architecture (423 routes). That harvest is now superseded by this
canonical mirror.

Mirror fetched via `mcp__github__get_file_contents` against
AdaWorldAPI/WoA `.claude/reference/routing_table.{json,csv}`, with
the MCP response wrapper (the "[Resource from github at repo://..."
prefix and the surrounding `[{"type":"text", "text":...}]` envelope)
stripped.

Schema per WoA PR #33:
  endpoint / function / path / methods / source_file / line /
  blueprint / cluster

Coverage: 514 of 515 runtime routes; the delta is Flask's auto
`/static/` rule (no decorator → invisible to AST).

README updated to:
- name WoA PR #33 as the canonical source
- document the mirror schema + refresh procedure
- record the supersession history (96 partial → 514 canonical) so
  future sessions don't reinvent the partial harvest
- list the cross-repo use cases (membrane wiring planning, route-
  disappearance audit across refactors, etc.)

https://claude.ai/code/session_01UwJuKqP828qyX1VkLgGJFS
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants