| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| README.md | 2026-05-11 | 2.7 kB | |
| v2.9.39 -- Africa pack + match-service universalisation source code.tar.gz | 2026-05-11 | 457.5 MB | |
| v2.9.39 -- Africa pack + match-service universalisation source code.zip | 2026-05-11 | 459.3 MB | |
| Totals: 3 Items | 916.9 MB | 0 | |
Triple-rollup release — v2.9.37 + v2.9.38 + v2.9.39 land together on PyPI / GH.
v2.9.39 — Africa catalogue pack + ranker test fixes
Added
- 12 new
CWICR_V3_CATALOGUESrows: NG_LAGOS, KE_NAIROBI, GH_ACCRA, UG_KAMPALA, TZ_DARESSALAAM (Anglo → nrm); SN_DAKAR, CI_ABIDJAN, CM_DOUALA (Francophone → untec); AO_LUANDA (Lusophone → masterformat); MA_CASABLANCA, EG_CAIRO, TN_TUNIS (Maghreb-AR → masterformat). Registry 30 → 42.
Fixed
- 71 legacy ranker monkeypatch refs migrated from deleted
app.core.match_service.rankertoranker_qdrantacrosstests/integration/test_phase0_edge_cases.py,tests/perf/test_match_concurrency.py,tests/perf/test_phase0_perf.py. Import errors cleared; 122 previously-erroring tests now pass.
v2.9.38 — Match-service universalisation (Tier-1 / Tier-2 / Tier-3 hardcode killed)
Added
_KNOWN_CLASSIFICATION_STANDARDS+_REGION_PREFERRED_STANDARDnow data-driven via_derive_standards_from_catalogues(): 11 standards (DIN276, MasterFormat, NRM, UNTEC, VOCI, BC3, GB50500, SEKISAN, KBIM, GESN, BirimFiyat). FR→untec, IT→voci, ES→bc3, CN→gb50500, JP→sekisan, KR→kbim, RU/UA/BY/KZ→gesn, TR→birimfiyat.- Region groups →
data/match/region_groups.yaml(28 groups including ASEAN, MENA_AR, FRANCOPHONE_AFRICA, LUSOPHONE_AFRICA, ANGLO_AFRICA, ANDEAN, CAUCASUS, ANGLO_CARIBBEAN, LUSOPHONE_LATAM). - Region-language overlay →
data/match/region_language.yaml(54 entries). CwicrV3Catalogue.default_classification_standardfield.make seed-cwicr-v3+backend/scripts/seed_cwicr_v3.pyCLI.enumerate_qdrant_v3_collections()helper.- Per-encoder confidence profiles (
data/match/encoder_profiles.json): bge-m3 / bge-small / e5-small / sonnet-rerank. - Per-language lex thresholds (
data/match/lex_thresholds.json): 20 language profiles. config.boost_weights_for_standard()API stub.pyyaml>=6.0promoted to base dependency.
v2.9.37 — Africa pack + dashboard polish
Added
- 19 African currencies (NGN, KES, GHS, MAD, TND, DZD, ETB, UGX, TZS, RWF, XOF, XAF, AOA, MZN, BWP, ZMW, NAD, MGA).
- 24 new African region codes (Anglophone EN, Maghreb AR, Francophone FR, Lusophone PT, ET AM).
- Apple-style dashboard backdrop (aurora mesh + dot grid + cursor spotlight + vignette + fine-grain noise; honours
prefers-reduced-motion). - Dashboard stats badges (Projects / BOQs / Modules / Users) navigate to their pages.
/match-elementsCatalogueAdvisor: one-click install of project-language catalogues when none are loaded.- Demo register: email-domain-aggregate clause.
3675/3675 backend unit tests pass. Wheel published to PyPI: https://pypi.org/project/openconstructionerp/2.9.39/