Commit Graph

4 Commits

Author SHA1 Message Date
Benjamin Admin 4f29e5ff3c feat: LLM verification for regex FAILs + section-split hardening
Build + Deploy / build-admin-compliance (push) Successful in 1m49s
Build + Deploy / build-backend-compliance (push) Successful in 9s
Build + Deploy / build-ai-sdk (push) Successful in 8s
Build + Deploy / build-developer-portal (push) Successful in 8s
Build + Deploy / build-tts (push) Successful in 9s
Build + Deploy / build-document-crawler (push) Successful in 8s
Build + Deploy / build-dsms-gateway (push) Successful in 7s
Build + Deploy / build-dsms-node (push) Successful in 8s
CI / branch-name (push) Has been skipped
CI / guardrail-integrity (push) Has been skipped
CI / loc-budget (push) Failing after 15s
CI / secret-scan (push) Has been skipped
CI / go-lint (push) Has been skipped
CI / python-lint (push) Has been skipped
CI / nodejs-lint (push) Has been skipped
CI / nodejs-build (push) Successful in 2m55s
CI / dep-audit (push) Has been skipped
CI / sbom-scan (push) Has been skipped
CI / test-go (push) Failing after 45s
CI / test-python-backend (push) Successful in 42s
CI / test-python-document-crawler (push) Successful in 27s
CI / test-python-dsms-gateway (push) Successful in 26s
CI / validate-canonical-controls (push) Successful in 15s
Build + Deploy / trigger-orca (push) Successful in 2m13s
Path to 100% correctness: Regex finds 80%, LLM catches the rest.

1. LLM verification (llm_verify.py):
   - Every regex FAIL is re-checked by Qwen (qwen3:32b)
   - Binary YES/NO question with evidence extraction
   - Overturned checks marked with [LLM] prefix in matched_text
   - Graceful fallback if LLM unavailable

2. Section splitter hardening:
   - Short lines (<16 chars) only treated as headings if preceded
     by blank line — prevents table column headers ("Funktion",
     "Speicherdauer") from splitting cookie sections
   - Fixes IHK cookie section: 288 words → full section

3. DSFA documentation patterns expanded:
   - Recognizes "4.) Ergebnis:" numbered result sections
   - Matches risk assessment conclusions

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-05-07 15:34:07 +02:00
Benjamin Admin fa4fd87102 fix: 7 regex bugs from IHK Konstanz ground truth analysis
Build + Deploy / build-admin-compliance (push) Successful in 9s
Build + Deploy / build-backend-compliance (push) Successful in 8s
Build + Deploy / build-ai-sdk (push) Successful in 42s
Build + Deploy / build-developer-portal (push) Successful in 8s
Build + Deploy / build-tts (push) Successful in 7s
Build + Deploy / build-document-crawler (push) Successful in 7s
Build + Deploy / build-dsms-gateway (push) Successful in 8s
Build + Deploy / build-dsms-node (push) Successful in 8s
CI / branch-name (push) Has been skipped
CI / guardrail-integrity (push) Has been skipped
CI / loc-budget (push) Failing after 18s
CI / secret-scan (push) Has been skipped
CI / go-lint (push) Has been skipped
CI / python-lint (push) Has been skipped
CI / nodejs-lint (push) Has been skipped
CI / nodejs-build (push) Successful in 2m57s
CI / dep-audit (push) Has been skipped
CI / sbom-scan (push) Has been skipped
CI / test-go (push) Failing after 49s
CI / test-python-backend (push) Successful in 42s
CI / test-python-document-crawler (push) Successful in 28s
CI / test-python-dsms-gateway (push) Successful in 23s
CI / validate-canonical-controls (push) Successful in 15s
Build + Deploy / trigger-orca (push) Successful in 2m24s
Fixes based on manual verification of all 30 failed checks:
1. Cookie table: recognize "folgende cookies" + column headers as text
2. Cookie names: add JSESSIONID, cookieinfo, et_id, BT_* patterns
3. Essential justified: match "sitzung zuordnen", "betrieb der website"
4. Social bookmarks: recognize as 2-click alternative
5. DSFA plural: "kanaelen" now matches alongside "kanal"
6. Section splitter: skip-headings no longer lose subsequent text
   (Risikoabwaegung section was cut from DSFA, losing risk scores)
7. Cookie legal basis: accept Art. 6(1)(f) in cookie context

Reduces false positives from 7 to ~1-2 for IHK Konstanz test case.
Ground truth table: zeroclaw/docs/ground-truth-ihk-konstanz.md

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-05-07 14:51:09 +02:00
Benjamin Admin 293c58d0dd feat: Add actionable hints to all 138 compliance checks
Build + Deploy / build-admin-compliance (push) Successful in 1m40s
Build + Deploy / build-backend-compliance (push) Successful in 7s
Build + Deploy / build-ai-sdk (push) Successful in 35s
Build + Deploy / build-developer-portal (push) Successful in 8s
Build + Deploy / build-tts (push) Successful in 7s
Build + Deploy / build-document-crawler (push) Successful in 8s
Build + Deploy / build-dsms-gateway (push) Successful in 7s
Build + Deploy / build-dsms-node (push) Successful in 8s
CI / branch-name (push) Has been skipped
CI / guardrail-integrity (push) Has been skipped
CI / loc-budget (push) Failing after 16s
CI / secret-scan (push) Has been skipped
CI / go-lint (push) Has been skipped
CI / python-lint (push) Has been skipped
CI / nodejs-lint (push) Has been skipped
CI / nodejs-build (push) Successful in 2m50s
CI / dep-audit (push) Has been skipped
CI / sbom-scan (push) Has been skipped
CI / test-go (push) Failing after 40s
CI / test-python-backend (push) Successful in 37s
CI / test-python-document-crawler (push) Successful in 25s
CI / test-python-dsms-gateway (push) Successful in 23s
CI / validate-canonical-controls (push) Successful in 15s
Build + Deploy / trigger-orca (push) Successful in 2m28s
Each check now has a "hint" field explaining what is missing and
what the customer should do to fix it. Hints are shown in the
frontend below failed checks in red text.

Examples:
- "Bei Verarbeitung auf Basis von Art. 6(1)(f) muss dokumentiert
  werden, warum Ihr berechtigtes Interesse die Rechte der
  Betroffenen ueberwiegt."
- "Die ladungsfaehige Anschrift fehlt. Erforderlich: Strasse,
  Hausnummer, PLZ und Ort."

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-05-07 14:05:01 +02:00
Benjamin Admin b363c28539 feat: Add 76 Level-2 regex checks for document correctness verification
Split dsi_document_checker.py (466 LOC) into doc_checks/ package (9 files).
Two-pass L1→L2 logic: L1 checks "Is it mentioned?", L2 checks "Is it correct?"
(e.g. controller has full address, specific Art. 6 lit., concrete time periods).

138 total checks (62 L1 + 76 L2) across 7 doc types:
- DSE Art. 13: 31, Impressum §5 TMG: 16, Cookie §25 TDDDG: 15
- Widerruf §355: 15, AGB §305ff: 21, Social Media Art. 26: 20, DSFA Art. 35: 18

Frontend: hierarchical L1→L2 display with dual progress bars
(green=completeness, blue=correctness).

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-05-07 12:37:03 +02:00