1d65d99d5fd45c29e2d2089d69211fb61fea8b29
2 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 Benjamin Admin
|
f2d445b891 |
fix(ucca): Cross-Reg 0070 — beide Regelwerk-Domaenen im Router-Top-K (Known Defects 0)
CI / detect-changes (pull_request) Successful in 13s
CI / branch-name (pull_request) Successful in 1s
CI / guardrail-integrity (pull_request) Successful in 9s
CI / secret-scan (pull_request) Successful in 10s
CI / dep-audit (pull_request) Failing after 56s
CI / sbom-scan (pull_request) Failing after 59s
CI / build-sha-integrity (pull_request) Successful in 5s
CI / validate-canonical-controls (pull_request) Successful in 3s
CI / test-python-document-crawler (pull_request) Successful in 15s
CI / test-python-dsms-gateway (pull_request) Successful in 13s
CI / loc-budget (pull_request) Successful in 23s
CI / go-lint (pull_request) Failing after 51s
CI / python-lint (pull_request) Failing after 18s
CI / nodejs-lint (pull_request) Failing after 1m8s
CI / nodejs-build (pull_request) Successful in 3m6s
CI / test-go (pull_request) Successful in 1m3s
CI / iace-gt-coverage (pull_request) Successful in 18s
CI / test-python-backend (pull_request) Successful in 28s
Der einzige offene Retrieval-Haertefall: eine Query mit >=2 genannten Regelwerken
("CRA und Maschinenverordnung") lieferte nur die keyword-dominante Domaene (CRA),
MaschVO fiel raus. Drei zusammenwirkende Ursachen, alle behoben:
1. CodeValues-Mismatch: MaschVO heisst je Collection anders (Slice MASCHVO ·
gesetze MVO · ce MACHINERY/MASCHINENVO), der Catalog hatte nur ["MASCHVO","MaschVO"]
→ Filter fand MaschVO nur in der Slice. Jetzt alle Varianten als CodeValues.
2. Per-Collection-Truncation: der Router gab perColl=3 → searchMultiRegulation holte
3+3=6, schnitt auf 3 → konnte eine Domaene je Collection verlieren. Multi-Reg-Queries
bekommen jetzt perColl = 3*len(regs).
3. Router-Score-Merge starvte die nicht-dominante Domaene. Neue balanceByRegulation()
gruppiert den gemergten Pool per Regelwerk (exakter regulation_code-Match) und nimmt
round-robin ueber die genannten Domaenen → jede Domaene mit Treffern ist im Top-K.
Generisch ueber jede genannte Menge; Single-Domain-Pfad unveraendert.
Validierung: Go-Unit (balanceByRegulation: dominante CRA verdraengt MaschVO NICHT mehr);
0070-e2e gegen dev (Retrieve() → [CRA MVO CRA MVO CRA MVO CRA MASCHINENVO] = beide
Domaenen, vorher nur CRA); CB-100-Stichprobe REGR 0 (Gain-Profil unveraendert).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
|
||
|
Benjamin Admin
|
1e5aaf7103 |
feat(advisor): Authority Router — Advisor collection-agnostisch, KB-2026.1-Gewinn im Produktpfad
CI / detect-changes (pull_request) Successful in 13s
CI / branch-name (pull_request) Successful in 2s
CI / guardrail-integrity (pull_request) Successful in 5s
CI / secret-scan (pull_request) Successful in 11s
CI / dep-audit (pull_request) Failing after 54s
CI / sbom-scan (pull_request) Failing after 1m1s
CI / build-sha-integrity (pull_request) Successful in 11s
CI / validate-canonical-controls (pull_request) Successful in 7s
CI / loc-budget (pull_request) Successful in 23s
CI / go-lint (pull_request) Successful in 53s
CI / python-lint (pull_request) Failing after 17s
CI / nodejs-lint (pull_request) Failing after 1m6s
CI / nodejs-build (pull_request) Successful in 2m59s
CI / test-go (pull_request) Successful in 1m0s
CI / iace-gt-coverage (pull_request) Successful in 17s
CI / test-python-backend (pull_request) Successful in 26s
CI / test-python-document-crawler (pull_request) Successful in 12s
CI / test-python-dsms-gateway (pull_request) Successful in 8s
Der Advisor fan-outete bisher selbst ueber eine feste Liste expliziter Collections
(advisor-rag.ts) und umging damit das #61-Scope-Routing (das nur den Default-Pfad
routet) → der gemessene +28-Retrieval-Gewinn (CB-100: 53→81, 0 Regr) kam nie beim
Antwort-LLM an. Dieser Router zieht den Fan-out in die Retriever-Schicht:
- SDK: LegalRAGClient.Retrieve() + POST /sdk/v1/rag/retrieve {query, top_k} —
fan-outet server-seitig ueber die Broad-Authority-Base + die KB-2026.1-Slice bei
inKBScope, merge+dedup, sortiert nach Authority-Score (rerankByAuthority je
Collection), top-K. Index-Warmup vor dem nebenlaeufigen Fan-out (Map-Race-frei).
Per-Env via RAG_ROUTER_COLLECTIONS.
- admin: advisor-rag.ts ruft EINMAL /retrieve statt 6-fach expliziter Collections.
Advisor ist collection-agnostisch (Vertrag Compiler→Collections→Retriever→Advisor);
COMPLIANCE_COLLECTIONS/searchCollection entfernt.
Validierung: Go-Unit (Router-Selektion, dedup); e2e gegen dev-Qdrant (echter
Retrieve(), CB-100-Stichprobe stride 5): OLD-hit 11/20 → NEW-hit 15/20, GAIN 4
(alle DS-Guidance), REGR 0 — reproduziert den +28/0-Regr durch den Produktionscode.
TS-Tests auf den Single-/retrieve-Call angepasst.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
|