5 Commits

Author	SHA1	Message	Date
Benjamin Admin	b87c27d104	fix(llm-verify): P13 — Default-Modell auf qwen3:30b-a3b (statt qwen3.5:35b-a3b) ... Bug: qwen3.5:35b-a3b liefert mit format='json' + Batch-Prompt leere Strings zurueck ('LLM batch: empty response from model'). Im echten Compliance-Check lief der LLM-Verifier deshalb wirkungslos — False-Positive-Findings wie 'Vorstand nicht erkannt' (BMW: Klammer- Liste) wurden nicht overturned. Fix: Default auf qwen3:30b-a3b umgestellt. Verifiziert mit BMW- Impressum-Text: representative_person wird mit Evidence 'Milan Nedeljkovic, Vorsitzender' overturned=True markiert. OLLAMA_VERIFY_MODEL Env-Var bleibt als Override-Moeglichkeit. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>	2026-05-19 09:11:01 +02:00
Benjamin Admin	8a44e67293	feat(compliance-check): unlock all 1874 MCs + close gap-table items ... User: 'wir haben 1800 MCs erstellt um sie zu 10% zu nutzen — das ist Schwachsinn'. Fixed all 6 gaps from the audit. #1 max_controls=0 (was 20): - agent_compliance_check_routes _check_single: passes max_controls=0 to check_document_with_controls -> ALL MCs evaluated per doc_type. - 8 doc_types now use 1874 MCs instead of 160 (10x coverage). - Regex matching is cheap (<1s per doc); LLM-enrich cap of 10 stays. #2 LLM-verify fixed: - llm_verify.py was getting 0/N parsed. Causes: qwen3 thinking-mode wrapped output in <think>...</think>, /api/generate doesn't enforce JSON, prompt didn't handle code-fence wrappers. - Now uses /api/chat with format='json' (forces valid JSON). - _parse_batch_response strips <think> tags, accepts {results:[...]} AND bare [...], adds richer regex-fallback parse, logs raw head on total parse failure for diagnosis. #3 Loeschkonzept checklist (new): - doc_checks/loeschkonzept_checks.py — 9 L1 + 7 L2 checks per DIN 66398 + Art. 5(1)(e)/17/32 DSGVO: scope+responsibility, data categories, retention periods, legal basis refs (HGB/AO/BGB), deletion trigger, deletion process+technical+systems, deletion proof, exceptions + Art. 18 lock, review cycle, DSGVO references. - runner.py registered for loeschkonzept/loeschung/loeschfristen. #4 regulation backfill script: - backend-compliance/scripts/backfill_mc_regulation.py — regex-detects DSGVO/TDDDG/TMG/BGB/HGB/AO/MStV/UWG/VSBG/PAngV/GwG/BDSG/EU-VO references in MC title+question+pass_criteria, UPDATEs regulation + article fields. - Idempotent (only NULL rows), --dry-run flag, batched 200/UPDATE. - Run inside container: docker exec bp-compliance-backend python3 \ /app/scripts/backfill_mc_regulation.py #5 MC alias-fallback: - rag_document_checker._MC_ALIAS_FALLBACK maps doc_types without own MCs to a related set: nutzungsbedingungen->agb, social_media->dse, sub_processor/scc/tom_annex->avv, loeschfristen->loeschkonzept, eu_institution/dsb->dse. - _load_controls retries with the alias when the primary query returns 0 rows. - 14 additional doc_types now get MC coverage transparently. #6 cross-domain auto-discovery: - _autodiscover_missing builds a crawl plan: primary submitted base + up to 2 related domains sharing the owner SLD (e.g. BMW Group: bmw.de + bmwgroup.com + bmwgroup.jobs). - Detection: regex over submitted texts for https?://...<owner>... hostnames distinct from the primary base. - Each crawled base contributes documents + cmp_payloads to the discovery pool. Net effect for BMW: 1874 MCs evaluated (90 from cookie alone, was 20), Loeschkonzept Pflichtangaben benoten-bar, LLM overturns false regex FAILs, Joint-Controller policies on bmwgroup.jobs (Social Media) jetzt entdeckbar. Same wins will apply to CRA-Compliance check.	2026-05-17 13:07:50 +02:00
Benjamin Admin	ce77cde309	fix(compliance-check): batch LLM verification + increase poll timeout ... - LLM verify now sends ALL failed checks in one batched call instead of one Ollama call per check (80+ calls → 1 per document) - Increase frontend poll timeout from 6 min to 15 min Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	2026-05-12 11:49:30 +02:00
Benjamin Admin	f51671737a	fix: Correct Ollama model name + strict blank-line heading detection ... 1. LLM model: qwen3:32b → qwen3.5:35b-a3b (actual model on Mac Mini) 2. Section splitter: headings MUST be preceded by a blank line. This prevents cookie table entries ("Funktionale Cookies", "Session Cookies") from splitting the cookie section. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	2026-05-07 15:53:53 +02:00
Benjamin Admin	4f29e5ff3c	feat: LLM verification for regex FAILs + section-split hardening ... 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