מוסיף מסלול ייעודי לקליטת ההחלטה החתומה של היו"ר, ומפעיל אותו דרך שני
שלבים אוטומטיים מדורגים עם פאנלי-סוכנים (אוטו-אישור + אסקלציה ליו"ר).
Backend (web/):
- POST /api/cases/{case}/final/upload — קליטת final חיצוני: שמירה קנונית
(סופי-{case}.docx + עותק קורפוס-סגנון תחת case_number מלא כדי שבל"מ לא
יתנגש עם ערר באותו מספר), פתיחת draft_final_pairs (final_received). לא נוגע
ב-active_draft ולא מריץ retrofit (נבדל מ-exports/upload ו-mark-final → לא G2).
- POST .../final/run-learning + .../final/run-halacha — שלבים מדורגים שמעירים
worker מקומי (claude/DeepSeek/Gemini מקומיים בלבד) דרך הרחבת
wake_curator_for_final עם param task=learning|halacha.
פאנל-סגנון חדש (scripts/style_lesson_panel.py): שני שופטים (DeepSeek+Gemini)
על-גבי דיסטילציית-ה-Opus; הסכמה 2/2-keep → decision_lesson
(source=panel:deepseek+gemini); substance מדולג (INV-LRN5); הפיך + גיבוי CSV.
פאנל-הלכות: docstring/SCRIPTS.md עודכנו (--apply מחווט).
Frontend (web-ui/): כפתור "העלאת החלטה סופית של היו"ר" + שני כפתורים מדורגים
"הרץ למידת-קול"/"הרץ אימות-הלכות" ב-drafts-panel; כל התוויות בעברית
(badge מקור-לקח: "פאנל: דיפסיק+גמיני", "הרמס (סקירה)"...).
Spec: docs/spec/07-learning.md §0.6. Invariants: INV-LRN1/LRN4/LRN5, G10
(שער-יו"ר ידני להטמעה ל-SKILL.md/lessons.md — הפאנלים יוצרים הצעות בלבד);
G2 (מסלול-סופי הוא יכולת חסרה, לא מסלול-מקביל).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
הדף הציג את התורים באופן לא-אחיד (by_status גולמי), בלי הבחנה בין "ממתין"
(בקלוג: status=pending) ל"בתור" (התור הפעיל: requested_at IS NOT NULL), בלי
הצגת הפריט שרץ כרגע, ובלי שום שליטה בתהליכים.
מה נוסף:
1. כרטיסי-תור אחידים — בתור / ממתין(בקלוג) / בעיבוד / הושלם / נכשל + "רץ עכשיו"
(citation/case_number של הפריט בעיבוד) לכל drain (אחזור-פסיקה, מטא-דאטה,
הלכות, יומונים). שערי-אנוש (אישור-הלכות, פסיקה-חסרה) נשארים מוני-סטטוס.
2. פאנל ניהול-תהליכים בסגנון "שירותי Windows":
- דמון (court-fetch-service/xvfb/chat/reaper): הפעל-מחדש / עצור / הפעל.
- cron drain: "הרץ עכשיו" (pm2 restart) + מתג הפעל/כבה תזמון.
3. כל תגי-הסטטוס מתורגמים לעברית.
מנגנון:
- הפעל/כבה תזמון = דגל ב-DB (טבלה drain_controls). pm2 cron_restart מחיה תהליך
שעוצר ב-stop, לכן ה"כיבוי" האמין הוא דגל שכל drain בודק ב-startup (no-op מיידי
כשכבוי). הקונטיינר כותב/קורא ישירות מ-DB.
- הרץ-עכשיו + restart/stop/start = proxy ל-pm2 דרך endpoint חדש בגשר-המארח
(court_fetch_service /pm2/control), מאובטח Bearer + whitelist ל-legal-* בלבד.
- יומונים: drain_digests הועבר מ-crontab ל-pm2 (legal-digest-drain.config.cjs)
כדי שיופיע ויהיה שליט כמו כל drain. drain_halacha_queue.py הובא לבקרת-גרסאות.
Invariants: מקיים G2 (הרחבת /operations + הגשר הקיים, לא מסלול מקביל) ו-G1
(drain_controls = מקור-אמת יחיד לכיבוי, נורמליזציה במקור ולא תיקון-בקריאה).
אין בליעת שגיאות שקטה (הגשר מחזיר {ok,error}; המוטציות מציגות toast).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 211 open missing_precedents include 99 Supreme serial-format rulings
(בג"ץ/בר"מ/עע"מ NNNN/YY) with no נט-format triple — fetchable only from
supremedecisions.court.gov.il. Decoded its public JSON API (no browser, no
CAPTCHA, no smart-card); validated live on בג"ץ 3483/05 + בר"מ 10212/16.
- court_fetch_supreme.py: rewrite. POST Home/SearchVerdicts with a structured
`document` ({Year:"YYYY", CaseNum, OldMainNumFormat:true, SearchText:[…]}) +
X-Requested-With header → records; GET Home/Download?path=&fileName=&type=4 →
PDF. The earlier attempt failed only on the request shape (string vs object).
2-digit→4-digit year; try candidate docs best-first (פסק-דין→pages), skipping
the published-report 's'-prefix files the free endpoint WAF-blocks.
- orchestrator: on successful ingest, close matching open missing_precedents
(link to the new case_law). End-to-end validated (בר"מ 10212/16 → corpus).
- backfill_missing_precedents.py: enqueue fetchable open gaps (supreme + net)
into court_fetch_jobs; the drainer fetches+ingests+closes. dry-run default.
- X13 spec + SCRIPTS.md updated (Tier-0 decoded, no longer a limitation).
Very old un-digitized Supreme cases (e.g. בג"ץ 389/87 → 0 records) → manual.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The halacha extraction queue was stuck (same class as the metadata issue): 26
precedents requested extraction with no drainer, plus 1 orphaned in 'processing'
(status=processing, requested_at cleared → never re-picked by the queue).
- db.requeue_stale_processing_extractions(kind): re-stamp orphaned 'processing'
rows (requested_at IS NULL) so they re-drain; halacha extractor force=False
resumes from chunk checkpoints (no duplicates).
- process_pending_extractions calls it at the top — fully unattended, safe under
the global advisory lock. Mirrors the digests-drain self-heal.
- legal-halacha-drain.config.cjs: pm2 cron (every 2h, conservative — Claude is
slow/rate-limited and each run adds to the chair's pending_review queue).
drain_halacha_queue.py stays on claude_session (high reasoning quality for
holding/ratio; NOT moved to Gemini). SCRIPTS.md.
The chair-approval gate (INV-G10) is untouched — this only produces halachot;
Daphna still approves each in /approvals.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The /precedents metadata queue was stuck — 24 rows requested, nothing draining
them — and the agentic claude CLI hit error_max_turns on what is a single
structured text→JSON task (slow + flaky). Metadata extraction is bounded
extraction, the wrong fit for an agentic loop.
- gemini_session.py: query_json drop-in (gemini-2.5-flash, JSON mode, httpx —
no new SDK dep). Reads GEMINI_API_KEY (~/.env; SoT Infisical
nautilus:/external-apis/gemini). Host-side only — no LLM from the container.
- precedent_metadata_extractor: claude_session.query_json → gemini_session.
Validated live: rich, accurate fields (case_name/summary/appeal_subtype/tags).
- process_pending_extractions: kind-aware cooldown — metadata 2s (Gemini, fast),
halacha keeps 30s (Claude rate limits).
- drain_metadata_queue.py + legal-metadata-drain.config.cjs (pm2 cron */15) so
the queue never clogs again. SCRIPTS.md.
- X8 INV-FP5 updated: per-task engine choice (Gemini=bounded metadata,
claude_session=agentic halacha), both host-side, single canonical queue (G2).
Agentic/voice-sensitive work (writing, analysis, halacha) stays on claude_session
(Daphna's subscription). Gemini cost ≈ $0.10/1M tokens — negligible.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Periodic safety net for the multi-judge approval panel: samples panel-approved
halachot, re-runs the same 3-judge KEEP vote, and surfaces any that now lean
DROP — candidate false-keeps a human should glance at. Report-only by default;
--flag reopens flips to pending_review. Baseline 0/15 on the 2026-06-07 batch.
Closes the loop the literature prescribes (Trust-or-Escalate / selective
prediction): monitor the auto-decision error rate rather than trusting it
blindly. Reuses halacha_panel_approve's judges (single source of truth).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The chair cannot review every pending halacha. Three independent-lineage judges
(Opus via claude_session · DeepSeek · Gemini-2.5-flash — #1 on LegalBench) vote
on the COARSE axis we proved reliable across models (92%): "is this a genuine,
keepable rule?". Only an agreed verdict acts; every split escalates to the chair
(INV-G10). Buckets: clean→KEEP?; nli_unsupported→entailment re-adjudication;
extraction-defects→re-extraction.
halacha_panel_calibrate.py calibrates the voting policy on the gold-set's
is_holding (the coarse label) per Trust-or-Escalate (ICLR 2025): unanimous →
94.9% precision / 78% coverage; majority → 92.9% / 99%; ZERO false-drops in
both (the panel never rejects a good rule). Chosen policy (chair-approved):
clean→majority-2/3, nli→asymmetric (majority-reject, unanimous-approve),
defects→re-extraction. Reversible (--apply backs up review_status+flags first).
Sources: Panel-of-LLM-Evaluators (PoLL) · Trust-or-Escalate (ICLR 2025,
arXiv:2407.18370) · selective-prediction / learning-to-defer.
Invariants: upholds G10 (human gate — splits escalate, panel only collapses the
queue) and G9 (provenance — reviewer records the panel + policy). Read paths only
in calibrate; --apply writes review_status/quality_flags reversibly with backup.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- scripts/drain_court_fetch.py: drives orchestrator.drain_pending (host-only;
no-op when queue empty). Mirrors drain_halacha_queue.py.
- scripts/legal-court-fetch-drain.config.cjs: pm2 cron (hourly :17, one-shot),
COURT_FETCH_DRAIN_CRON override.
- fix: orchestrator default service URL 127.0.0.1 → 10.0.1.1 (the service binds
the docker0 gateway; the host can't reach it on loopback). Found live — the
first drain failed "connection refused" until corrected.
- SCRIPTS.md entries.
Validated end-to-end in PRODUCTION on a real digest: עת"מ 43830-12-24
(החברה להגנת הטבע) fetched from נט המשפט → case_law (79 chunks, source_url),
digest relinked (INV-DIG3 closed), halacha queued pending_review. job=done.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The gold-set's human role tags were made while seeing a claude AI recommendation,
so human↔AI agreement (~100%) is anchoring, not an independent accuracy signal.
This adds a third, genuinely independent judge — a DIFFERENT model (DeepSeek,
direct OpenAI-compatible API) classifies rule_role BLIND (never sees the human
tag nor the first AI's answer) — and reports an inter-rater agreement matrix.
Finding (100 tagged items): ai↔human 100% (anchored) vs deepseek↔human 50%
fine-grained — BUT 92% on the coarse axis (generalizable-rule vs application/
obiter). Conclusion: the fine sub-type (holding/interpretive/procedural) is an
inherently fuzzy boundary two capable models split differently; the coarse
"is this a real rule" axis is robust across models. Use the coarse axis as
ground truth; treat the sub-type as advisory, never as a gate.
Zero chair tagging, read-only on the gold-set. Key from ~/.hermes deepseek env.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
שלוש שכבות-הגנה נגד דליפת-זיכרון מדפדפנים יתומים, + טיפול בדליפה הגדולה
בפועל בשרת (task-master-mcp).
- camofox_client.py:
- asyncio.wait_for קשיח סביב כל ה-fetch (COURT_FETCH_HARD_TIMEOUT_S=180ש')
— hang → ביטול → async-with tear-down → reap.
- _reap_orphan_browsers(): הורג camoufox-bin יתומים (ppid=1) לפני ואחרי כל
fetch. סדרתיות (INV-CF4) → כל ppid=1 הוא שארית בטוחה.
- scripts/reap_orphan_procs.py: reaper כללי ל-task-master-mcp (~3GB יתומים)
+ camoufox-bin. רק ppid=1; /proc טהור. --dry-run / --loop N.
- scripts/legal-reaper.config.cjs: דמון pm2 (loop 180s, max_memory_restart 100M).
- X13 spec + SCRIPTS.md: תיעוד שכבות-ההגנה.
max_memory_restart בשירות (1.5G) כבר נותן רשת-ביטחון ברמת-התהליך.
Invariants: מקיים INV-CF4 (politeness/serial) — ללא שינוי חוזה.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The extractor classified rule_type by SOURCE bindingness (higher-court→binding,
committee→persuasive) instead of by rule KIND. The gold-set proved it: 'binding'
appeared on 19/19 external rulings & 0 committees; 'persuasive' on 13/13
committees & 0 external — only 58% agreement with the human role tags. The two
axes (authority vs rule role) were crammed into one enum.
This splits them per INV-DM7:
- authority (binding/persuasive) — DERIVED from case_law.precedent_level
(עליון/מנהלי→binding, ועדת_ערר_מחוזית→persuasive), never stored, never
LLM-guessed. New helper halacha_quality.derive_authority; surfaced read-only
in list_halachot / goldset_list / search results.
- rule_type — now the rule ROLE only: holding/interpretive/procedural/
application/obiter. Both extractor prompts unified to this vocabulary;
_coerce_halacha no longer defaults rule_type from the source; legacy
binding→holding / persuasive→interpretive fold for safety.
UI: authority shown as a separate read-only badge (gold=מחייב / muted=משכנע)
across the review queue, precedent detail, and gold-set; the gold-set role
selector drops binding/persuasive and adds מהותי (holding).
Migration: scripts/halacha_rule_role_backfill.py re-classifies the 276 pre-split
binding/persuasive rows into a genuine role via local claude_session (run after
deploy). Gold-set correct_type/ai_correct_type 'binding'→'holding' via SQL.
Sources (≥3, per research-decision policy): OASIS LegalRuleML v1.0
(appliesAuthority/Strength as metadata orthogonal to rule logic) · SemEval-2023
Task 6 LegalEval (rhetorical roles by function, authority kept separate) ·
Bluebook signals (weight-of-authority is a separate dimension).
Invariants: ESTABLISHES INV-DM7. Upholds G1 (normalize at source — extractor
classifies role, system derives authority) and G2 (single source of truth —
authority derived, not a parallel stored field). Tests: 211 pass + new
derive_authority/coerce coverage. web-ui build + tsc clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The chair wanted an independent recommendation beside each tag, to reconsider
his own judgments. Adds a NON-ground-truth AI second-opinion:
- schema: halacha_goldset.ai_is_holding / ai_correct_type / ai_rationale /
ai_generated_at (additive).
- db.goldset_set_ai_recommendation + goldset_list now returns the ai_* fields.
- scripts/goldset_ai_recommend.py — local claude_session judges is_holding +
type + a one-line rationale per item, INDEPENDENTLY (own legal rubric).
Independent of the rule-based validators #81.8 measures → no circularity.
Never auto-applied; QA aid only.
- web-ui: each card shows "🤖 המלצת AI: הלכה/לא · type" + rationale and an
agreement/disagreement chip vs the human tag (amber on disagree); a
"⚠ אי-הסכמות AI (N)" filter to review only the conflicts.
Methodology note kept explicit: the human stays the ground truth; the AI is a
prompt to reconsider, not to copy.
Verified: tsc --noEmit 0; generator stores recs and flags disagreements with
existing human tags.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Cross-precedent recurrence of a principle is real but is NOT citation
corroboration (X11) — the 5 candidate pairs have ZERO citations between their
precedents. Recording them in halacha_citation_corroboration would fabricate
citation data and inflate corroboration_count. This adds a proper, separate
halacha-level link for parallel authority.
Schema (V28): equivalent_halachot — symmetric (halacha_a < halacha_b, CHECK +
UNIQUE), non-citation, cross-precedent-only. ON DELETE CASCADE.
db.py:
- link_equivalent_halachot (idempotent; rejects same-id and SAME-precedent pairs
— parallel authority is cross-precedent by definition), unlink, and
list_equivalent_for_halacha.
- list_halachot gains include_equivalents → _annotate_equivalents attaches an
`equivalents` list (both directions) per row.
API: include_equivalents on GET /api/halachot; GET/POST/DELETE
/api/halachot/{id}/equivalents for the chair to view/link/unlink manually.
scripts/halacha_batch_reconcile.py: --link records found cross-precedent pairs
as equivalent_halachot (non-destructive, idempotent).
web-ui: Halacha.equivalents type; the clean review queue fetches
include_equivalents; the review card shows a gold "עיקרון מקביל ב-N" badge + an
expandable list (case + rule + similarity) labeled "אסמכתה מקבילה — לא ציטוט".
Populated the 5 reviewed pairs (chair decision: keep all + link as parallel
authority). Verified: 5 rows; the 1023-20 hub annotates 3 of its halachot with
equivalents; tsc --noEmit exits 0.
Invariants: G1 (model recurrence at source in its own table, not by abusing the
citator); G2 (no parallel path — extends list_halachot); citator integrity
preserved (corroboration stays citation-only).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Halacha-extraction quality (#81) and dedup-on-insert (#82) — engine changes
(pure + tested) plus measurement/ops tooling.
halacha_quality.py
- #81.4 application gate: is_fact_dependent() (high-precision "applied to THIS
case" deixis per the strict rubric §3/§27) + FLAG_APPLICATION. compute_quality_flags
now takes rule_type and flags rule_type=='application' OR fact-dependent —
blocking auto-approve (an illustration is not a generalizable holding).
- #82.3 lexical tail signal: jaccard_shingles / normalized_levenshtein /
lexical_near_duplicate + FLAG_NEAR_DUPLICATE, for the 0.83–0.93 cosine band.
halacha_extractor.py — pass rule_type to the flag computation; re-type a
binding-labeled fact-application to 'application' (mirrors non_decision→obiter).
db.py (store_halachot_for_chunk) — dedup now fetches the nearest same-precedent
neighbor once: cosine ≥ DEDUP → skip (unchanged); cosine in [BAND, DEDUP) with
high lexical overlap → FLAG_NEAR_DUPLICATE (review, not skip — never drop a
possibly-distinct principle unreviewed).
config.py — HALACHA_DEDUP_BAND_COSINE (0.83).
Scripts:
- scripts/halacha_goldset.py (#81.7) — export stratified sample for human
tagging; score validators (P/R/F1) against the tags. Backbone for #81.8.
- scripts/halacha_batch_reconcile.py (#82.7) — conservative cross-precedent
dedup (cosine ≥0.95), dry-run report only.
- scripts/calibrate_halacha_dedup.py (#82.1) — calibrate the lexical thresholds
against the 2026-06-03 cleanup gold-set.
Deferred (documented): #82.4 merge-provenance and #82.5 DB ON CONFLICT/UNIQUE
on normalized quote are NOT included — the current skip+flag behavior is safe,
whereas a UNIQUE on normalized_quote would fail on existing dups and a blind
merge risks losing provenance; they need their own chair-reviewed migration.
#82.6 over-merge guard is moot until merge lands. #81.6 full rhetorical-role
classifier deferred (section pre-filter + application flag cover the practical
case); #81.8 blocked on the human-tagged gold-set (harness now provided).
Verified:
- pytest tests/test_halacha_quality.py — 52 passed (14 new).
- calibrate: configured (0.55,0.70) → precision 1.0 (zero false-merge), recall
0.30 — correct profile for an auto-approve-blocking signal.
- goldset export: 15-row sample CSV. batch reconcile: 819 halachot → 5
cross-precedent candidate pairs.
Invariants: G1 (normalize at source — flag at insert, not at read); §6 (no
silent swallow — suspect items flagged to review, never dropped); G2 (no
parallel path — same store_halachot_for_chunk / compute_quality_flags).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
#86.2 backfill + #86.3 benchmark, plus a #86.1 over-strip fix found en route.
extractor.py
- extract_nevo_ratio(): capture Nevo's מיני-רציו block (editorial holdings
summary) before it is stripped — a free professional gold-set (#86.3).
- _DECISION_START hardening (#86.2): the merged #86.1 regex over-stripped.
(a) פסק-דין headers are markdown-wrapped (**פסק דין**); the old anchor
required the keyword as the first line char with one separator, so it
missed the header and matched a citation 32K deep (עמ"נ 50567-07-21,
losing 45% of the body). Now tolerates leading markdown + 0-3 seps,
and the final-nun form (דין ן vs דינו נ).
(b) bare השופט/הנשיא matched CITATIONS ("השופט מ' חשין, פסקה 23"). The
authoring-judge line ends with a colon; we now require it.
ingest.py
- capture the ratio before stripping and store it on the row (best-effort,
non-fatal); also strip the text-upload path (was file-only).
db.py
- add case_law.nevo_ratio column (additive); allow it in update_case_law.
scripts/backfill_nevo_preamble.py (#86.2) — dry-run-by-default data migration:
finds historically-leaked rulings, captures ratio→nevo_ratio, rewrites
full_text (+content_hash), reindexes, and FLAGS (never deletes) halachot whose
quote lives in the removed preamble (review_status=pending_review +
nevo_preamble_leak flag). Safety guard: rows with keep%<--min-keep (60) are
excluded from --apply as suspected over-strip. --apply writes backup+manifest
to data/audit/ first. Chair-gated — NOT applied here.
scripts/nevo_ratio_benchmark.py (#86.3) — LLM-as-judge (local claude_session,
zero cost) measures recall/precision/granularity of our halachot vs the Nevo
ratio. Works pre- and post-backfill (reads nevo_ratio, falls back to full_text).
Verified:
- pytest tests/test_nevo_preamble.py — 12 passed (incl. citation/markdown
over-strip regressions).
- backfill dry-run: 19 leaked rulings, 27 contaminated halachot, all ≥75%
keep (the 32K over-strip is gone).
- benchmark on בג"ץ 1764/05: recall=0.875 precision=1.0 granularity=1.75x.
Invariants: G1 (normalize at source — strip/capture at ingest, not at read);
no silent swallow (contaminated halachot flagged + reported, not dropped);
data-migration is dry-run-default with backup+manifest, chair-gated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
הספ (docs/spec/, G1–G11) חובר לסוכני Paperclip דרך INV-AG1 אבל לא למסלול
שבו רוב הקוד נכתב בפועל — הסשן האינטראקטיבי של Claude Code. סוגר את הפער
לפני מחזור-2 (FU-9..15), שהוא כולו כתיבת-קוד.
שלוש שכבות אכיפה:
1. תיעוד — CLAUDE.md §"פרוטוקול כתיבת-קוד" + docs/spec בטבלת-הייחוס
2. hook — scripts/spec-guard.sh (PreToolUse על Edit/Write/MultiEdit, רשום
ב-.claude/settings.json) מזכיר פעם-בסשן בכל נגיעה בקובץ-קוד; non-blocking
3. PR — .gitea/PULL_REQUEST_TEMPLATE.md עם סעיף-חובה "Invariants"
המקבילה האינטראקטיבית ל-INV-AG1 שכבר אוכף על הסוכנים (HEARTBEAT §"קריאת-ספ").
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds scripts/rechunk_legacy_precedents.py: selects every case_law with a tiny
chunk (content<50 — the pre-fix chunker fingerprint) and runs
ingest.reindex_case_law (re-chunk+re-embed from stored full_text only, no
re-OCR/LLM, idempotent). Batch-idempotent (re-queries the affected set).
Run result (2026-06-03): 73 precedents reindexed, 0 failed. Tiny chunks
483 -> 4 (99.2%); total precedent_chunks 5019 -> 3115 (fragments merged).
Search verified healthy (substantial coherent passages, no errors).
The 4 residual tiny chunks are isolated section headings ('דיון',
'טענות המשיבים', ...) emitted by the CURRENT (fixed) chunker — not legacy
fragments — and are already filtered at query time (>=50, #55). Minor
chunker edge case, candidate #55 follow-up.
The DB chunk migration is already applied to prod; this commit is the script
+ SCRIPTS.md entry only (no app code change, no deploy needed).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Covers GAP-11 (INV-RET4/G8) and GAP-14 (INV-QA1/G10). Retrieval quality was
never measured (only telemetry observation) and the halacha review backlog was
invisible (the 10/19 gap was found by accident).
Unit B — backlog visibility (pure code, container):
- metrics.halacha_backlog(conn) → {pending_review, approved, rejected, published,
total, oldest_pending_at}; surfaced in metrics.get_dashboard() (get_metrics MCP
tool) and /api/system/diagnostics. Live count revealed 178 pending / 1552 total,
oldest from 2026-05-03 — previously invisible.
Unit A — retrieval eval harness (host-side scripts):
- scripts/eval_gold_bootstrap.py — seeds data/eval/gold-set.jsonl. Two sources:
citations (cited==relevant via search_relevance_feedback — empty until decisions
cite precedents) and known_item (query=case_name → relevant=self; a real
citation-free signal, the methodology #52 checked by hand). Idempotent; preserves
source='chair' rows.
- scripts/eval_retrieval.py — runs the production retrieval path (search_library /
search_internal) over the gold-set; computes precision@k, recall@k, MRR, nDCG@k
(k=5,10); aggregates overall + per-corpus + per-practice_area; writes a report and
a delta vs committed baseline.json (which records the retrieval_config it reflects).
--self-test unit-checks the metric math offline.
Gold-set strategy = hybrid (chair decision): bootstrap + chair review. The citation
source is empty today (0 cited precedents in decisions), so the seed is known-item
(77 queries: 54 internal_decisions + 23 precedent_library). The gold-set is
PROVISIONAL until Dafna reviews it (the domain chair-gate).
Baseline (production config: multimodal+rerank on): R@10=0.987, MRR=0.837,
nDCG@10=0.872. Finding: MULTIMODAL_ENABLED=true slightly lowers known-item recall
(image-page results displace exact name matches) — relevant to #15. precedent_library
weaker than internal (R@10 0.957 vs 1.0) — one external precedent unfindable by name.
"CI gate" realized as discipline (re-runnable harness + committed baseline + run
before/after any retrieval-layer change) — retrieval needs prod DB + Voyage, no CI
runner has that access.
Spec: docs/superpowers/specs/2026-05-31-fu5-eval-harness-design.md
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Root cause of "agent can't find the Agasi decision in the corpus" (CMPA-55):
the decision was fully ingested, but the retrieval layer failed on the
realistic agent query — searching by case name.
- RC-A (#52): lexical tsvector covered only chunk content + halacha text,
so a bare-name query ("אגסי") matched decisions that *cite* the case, not
the case itself. Add meta_tsv on case_law(case_name, case_number) (SCHEMA
V20) and OR it into the lexical halacha/chunk SQL with a match boost, so a
name/number hit surfaces the case's own rows. Agasi: rank 4 → rank 1.
- RC-B (#53): precedent_library_list hard-defaulted source_kind=external_upload
and never exposed the param, hiding uploaded ערר/בל"מ (internal_committee)
decisions. Thread source_kind through service → tool → MCP tool (supports
'internal_committee' / 'all_committees').
- #54: agent instructions (researcher/analyst/writer) — search-by-name
protocol: add content/case-number, search both corpora, use all_committees
before declaring "not in corpus".
- #55: chunker produced tiny fragment chunks ("דיון", "החלטה") from header
keywords matched mid-sentence. Anchor SECTION_PATTERNS to line start +
merge sub-min sections; exclude <50-char fragments at query time (484
existing fragments hidden; full re-chunk tracked as #57).
Tests: scripts/test_retrieval_by_name.py (name ranks case above citer +
substantive regressions); chunker unit checks (0 tiny chunks). New findings
filed as tasks #56 (halacha source_kind leak) and #57 (re-chunk migration).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Six-phase upgrade of /training from a read-only dashboard into a full
Style Studio for managing Daphna's style corpus.
- Upload Sheet on /training: file → proofread preview → commit (no more
CLI-only `upload-training` skill).
- Rich corpus metadata: GET /api/training/corpus returns summary, outcome,
key_principles, page_count, parties (regex), legal_citation, lessons_count.
PATCH endpoint for chair edits. CorpusDetailDrawer with 4 tabs (details
/content/lessons/patterns) replaces the bare table row.
- LLM metadata enrichment: style_metadata_extractor + MCP tools
(style_corpus_enrich, style_corpus_pending_enrichment) fill summary
/outcome/key_principles via claude_session (free, host-side).
- Per-decision lessons: new decision_lessons table + 4 REST endpoints +
LessonsTab in drawer; hermes-curator now auto-posts findings as
decision_lessons(source=curator).
- Curator Portrait tab: prompt rendered with link to Gitea, recent
curator findings, style_analyzer training prompts, propose-change
form that writes proposals to data/curator-proposals/ for manual
chair review (no auto-mutation of the agent file).
- Style chat tab: SSE-streamed conversations with the style agent.
New host-side pm2 service (legal-chat-service, port 8770) wraps
claude CLI with stream-json + --resume continuation; FastAPI proxies
via host.docker.internal. Zero API cost — uses chaim's claude.ai
subscription. chat_conversations + chat_messages persist history.
Architecture: keeps the existing rule that claude_session only runs
on the host (not the container). The new legal-chat-service is the
canonical bridge between the container and the local CLI for the chat
feature; everything else (upload, metadata, lessons) stays within the
container's existing capabilities.
Audit script (scripts/audit_training_corpus.py) included for verifying
which corpus rows still need enrichment.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
After the proceeding_type field landed, users started flipping cases
to בל"מ via the edit dialog. But the case-header badge + cases-table
filter were still gated on isBlamSubtype(appeal_subtype), so the badge
didn't appear when only the proceeding_type changed. Now the badge
shows when either proceeding_type === 'בל"מ' OR appeal_subtype is an
extension_request_* variant — the legacy path stays so existing rows
that never got a proceeding_type still render correctly.
Also regen types.ts from prod (proceeding_type now in OpenAPI schema)
and register the one-shot process_pending_blam.py script.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Same case_number can exist as both a regular appeal (ערר) and an
extension-of-time request (בל"מ), and we were inferring the difference
from appeal_subtype prefixes — fragile, and case-number lookups
weren't disambiguated. Now stored as a first-class field on both
case_law (corpus) and cases (live cases), with partial unique indexes
on (case_number, proceeding_type).
- SCHEMA_V15: column + CHECK constraints + backfill from
appeal_subtype LIKE 'extension_request_%' + partial unique indexes
replace the old global UNIQUE(case_number).
- derive_proceeding_type() centralizes the inference rule
(extension_request_* → בל"מ; subject regex fallback; default ערר).
- Metadata extractor prompt asks Claude to populate the new field
explicitly; apply_to_record writes it for internal_committee rows.
- internal_decision_upload, case_create, case_update accept an
optional proceeding_type; FastAPI request models expose it.
- Wizard + edit dialog get a sided Select; case header renders the
resolved label (ערר / בל"מ).
- Uploaded the 2 staged בל"מ decisions on betterment levy:
8126/24 (סופר נוח, 13 chunks), 8047/23 (הרנון, 48 chunks).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Four parallel sub-agents closed the remaining critical gaps from the
26/05 Stage A/B sprint. Each block independently tested; aggregated here.
## #30/#31 finalizers (sub-agent A)
* Auto-derive practice_area in case_create from case_number prefix
(1xxx→rishuy_uvniya, 8xxx→betterment_levy, 9xxx→compensation_197);
default for CaseCreateRequest is now "" (the DB constraint catches
any stray "appeals_committee").
* practice_area.py: derive_subtype now handles axis-B domain values
(rishuy_uvniya/betterment_levy/compensation_197) without parsing the
case number; new helper derive_domain_practice_area().
* Halacha re-extraction verified unnecessary — all 6 reclassified
records already had is_binding=false and approved halachot.
* Regression tests: 6 cases in tests/test_corpus_constraints.py
covering practice_area enum, internal-committee chair/district,
external-upload arar prefix, MCP guard.
* UI: district input → Select dropdown (7 districts) in
precedent-edit-sheet.tsx, preserving legacy free-text values.
## #37 בל"מ subtypes (sub-agent B)
* 3 new appeal_subtypes: extension_request_{building_permit,
betterment_levy,compensation}. APPEALS_COMMITTEE_SUBTYPES extended,
SUBTYPES_BY_AREA mappings added.
* New helpers: is_blam_subject(), is_blam_subtype(),
derive_subtype_with_blam(case_number, subject, practice_area).
case_create now uses it to auto-detect "בקשה להארכת מועד" subjects.
* 3 methodology templates under docs/methodology/extension-request-*.md.
* paperclip_client.py mapping updated for the 3 new subtypes
(extension_request_building_permit→CMP, the other two→CMPA).
* Frontend: bilingual "בל"מ" badge + filter dropdown on cases list +
detail header; appeal-type-bars collapseBlam() merges בל"מ into its
parent domain for aggregate bars.
* Wizard auto-detects בל"מ from subject during case creation.
* 3 Berlinger cases (1017/1018/1019-03-26) migrated to
appeal_subtype=extension_request_building_permit via psql.
## #35 missing_precedents feature (sub-agent C)
* Schema V13: missing_precedents table (citation, case_id, party,
legal_topic, status, linked_case_law_id, claim_quote, ...) +
FK constraints + 3 indexes. Applied via psql + idempotent migration.
* 6 db.py service functions, 3 MCP tools, 6 FastAPI endpoints
(POST/GET/PATCH/DELETE/upload — upload routes by citation prefix
to ingest_internal_decision or ingest_precedent).
* Next.js page /missing-precedents with 5 status tabs + filters +
sidebar badge counter + detail drawer with metadata edit + smart
upload form that switches fields per committee/court.
* Bootstrap: 7 rows imported from the JSON file
(3 citations × cases, all status=closed with linked_case_law_id).
* legal-researcher.md: new §2ב.5 with missing_precedent_create
usage + dedup semantics + tool grant.
## #36 legal_arguments aggregation (sub-agent D)
* Schema V14: legal_arguments + legal_argument_propositions M:M.
Applied via psql.
* New service argument_aggregator.py with two functions —
aggregate_claims_to_arguments() (Claude CLI / claude_session) and
get_legal_arguments(). Graceful llm_unavailable handling when CLI
is missing (containers).
* 2 MCP tools + 2 API endpoints (POST .../aggregate-arguments as
BackgroundTask, GET .../legal-arguments).
* Frontend: shadcn Accordion + new legal-arguments-panel.tsx with
hierarchical (party → priority badge → arguments) display, "טיעונים"
tab on the case page, "חשב/חשב מחדש" buttons.
* scripts/backfill_legal_arguments.py + SCRIPTS.md entry — dry-run
found 8 candidate cases including 1017/1018/1019.
## Open follow-ups (intentionally deferred)
* npm run api:types in web-ui (CLAUDE.md flow) — recommended before
the next UI commit; not required for backend deployment.
* Run backfill_legal_arguments.py --apply once the container picks up
the new aggregator service.
* webhook on missing-precedents upload-close to Paperclip (optional).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
A/B test (2026-05-05) showed DeepSeek V4-Pro is 2-3x faster and ~20x cheaper
than Sonnet for style/lexicon pattern analysis, with comparable quality.
Adds adapters/deepseek-paperclip-adapter/ package, documents adapter requirements
(env injection, run-id headers), updates CLAUDE.md with adapter integration notes,
and records lessons from ערר 1200-25 (block order for 1xxx, "להלן מתוך" pattern,
expanded factual background, bridge planning analysis, flat heading structure).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
paperclip-dev is for maintaining the Paperclip codebase itself — not
relevant to legal work. Removed from all 14 agents (was on CMPA mirror).
paperclip-converting-plans-to-tasks helps decompose a plan into assigned
issues. Useful for the planning-heavy agents (CEO, analyst). Now scoped
to those two — removed from the other 5 in CMPA where it had crept in.
Net effect: zero drift on paperclipai/* skills across all 7 master+mirror
pairs. Verified via the new Agents tab dashboard.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Brings the legal-ai ↔ Paperclip integration in line with the official
Paperclip skill. Net effect: HEARTBEAT.md -47% (370→195 lines), all 14
agents on uniform runtime_config + budget + instructionsBundleMode, and
two cross-company helpers replacing manual SQL.
Highlights:
- HEARTBEAT.md refactor: project-specific only, delegates to the official
paperclipai/paperclip skill (loaded per agent). Adds heartbeat-context
fast-path (§1.7) and PAPERCLIP_WAKE_PAYLOAD_JSON shortcut (§1.5).
- Issue Thread Interactions API: legal-ceo.md now uses
ask_user_questions / request_confirmation / suggest_tasks instead of
free-text comments — gives chair structured UI with idempotency keys.
- pc.sh + paperclip_api.pc_request: every API call goes through helpers
that inject Authorization + X-Paperclip-Run-Id (audit trail).
- sync_agents_across_companies.py: master(CMP)→mirror(CMPA) sync via
Paperclip API, idempotent, with --verify and --apply modes.
- skills/new-company-setup: 11-step blueprint distilling all 11 gaps
into a single onboarding runbook for the next company.
- .taskmaster: 12 tasks covering each gap (one already closed: #29).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The legacy chunker did not track which PDF page each chunk came from.
Stored chunks had page_number=NULL, which blocked the multimodal
hybrid retriever's text+image boost — it joins (chunk, image) on
(document_id, page_number) and the join could never fire.
This change:
- extractor.extract_text now returns (text, page_count, page_offsets);
page_offsets[i] is the start char offset of page (i+1) in the joined
text. None for non-PDFs.
- chunker.chunk_document accepts an optional page_offsets and tags
each chunk with the page that contains its first character (uses
the existing chunker logic; pages assigned post-hoc by content
search to keep the diff minimal).
- processor.process_document and precedent_library.ingest_precedent
forward page_offsets through the chunker. New uploads now carry
accurate page_number on every chunk.
- Other extract_text callers (tools/documents, tools/workflow,
web/app.py) updated to unpack the third element (ignored).
- scripts/backfill_chunk_pages.py: per-case retrofit. Re-extracts each
PDF (re-OCRs via Google Vision if needed, ~$0.0015/page), computes
page_offsets, and updates page_number on every chunk by content
search. Idempotent; --force re-runs on already-tagged docs.
Forward-only would leave the 419 image embeddings backfilled on
cases 8174-24 + 8137-24 unable to boost their corresponding text
chunks. The retrofit script closes that gap (cost ~$0.60).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Stage C: per-page image embeddings via voyage-multimodal-3 + hybrid
text+image search. Off by default; enable with MULTIMODAL_ENABLED=true.
- Schema V9: document_image_embeddings + precedent_image_embeddings
(vector(1024), page_number, image_thumbnail_path)
- extractor.render_pages_for_multimodal renders PDF pages at
MULTIMODAL_DPI (144) for embedding + JPEG thumbnails at
MULTIMODAL_THUMB_DPI (96) for UI preview, in one pass
- embeddings.embed_images calls voyage-multimodal-3 in 50-page batches
- services/hybrid_search.py orchestrator: rerank applied to text side
first (rerank-2 is text-only); image side cosine; weighted merge
with text_weight 0.65 (env-tunable); image-only pages surface as
match_type='image' so dense scanned content still appears
- processor.process_document and precedent_library.ingest_precedent
gated by flag — non-fatal on multimodal failure
- scripts/multimodal_backfill.py — idempotent per-case CLI to embed
existing documents without re-extracting text
Validated locally on a 5-page response brief: render 0.31s, embed 8.32s,
hybrid merge surfaces image rows correctly. Production rollout starts
with flag=false (no behavior change), then per-case A/B.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The claude_session bridge had two structural defects that made any
non-trivial document extraction unreliable:
1. subprocess.run() blocks the asyncio event loop in the MCP server
for the full duration of every LLM call (60-180s typical).
2. The 120-second timeout was below the cold-cache cost of any
document over ~12K Hebrew characters. Three back-to-back timeouts
on case 8174-24 dropped 43 appellant claims on the floor.
Phase 1 of the remediation plan — keeps claude_session as the engine
(no Anthropic API switch) and restructures around it:
claude_session.py
• query / query_json are now async — asyncio.create_subprocess_exec
instead of subprocess.run, so MCP server can serve other coroutines
while a call is in flight.
• DEFAULT_TIMEOUT 120 → 1800 (30 min). High enough that no realistic
document hits it; bounded so a runaway never zombifies forever.
• LONG_TIMEOUT 300 → 3600 for opus block writing on full case context.
• TimeoutError now actually kills the subprocess (asyncio.wait_for
cancellation alone leaves the child running).
claims_extractor.py
• _split_by_sections: chunks at numbered sections / Hebrew letter
headings / "פרק" markers / markdown ##, falls back to paragraph
breaks, then to hard splits. Targets 12K chars per chunk — small
enough that each chunk reliably finishes inside the timeout.
• _extract_chunk: per-chunk retry (1 attempt by default) with
structured logging on failure. Failed chunks no longer crash the
overall extraction; they're skipped with a partial-result warning.
• extract_claims_with_ai now runs chunks in parallel via
asyncio.gather bounded by a semaphore (CHUNK_CONCURRENCY=3).
For a 25K-char appeal: was sequential 150-300s, now ~70-90s.
Updated all 9 callers (claims, appraiser facts, block writer, qa
validator, brainstorm, learning loop, style analyzer × 3) to await
the now-async API.
The one-shot scripts/extract_claims_8174.py used to recover 43
appellant claims on case 8174-24 has been moved to .archive/ — phase 1
makes it obsolete. SCRIPTS.md updated.
Phase 2 (background-task wrapper around LLM-bound MCP tools, persistent
llm_tasks table, SSE progress) is the structural follow-up — separate PR.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Fixes critical bug in 1033-25: user-uploaded עריכה-*.docx files were
orphaned on disk while exports kept rebuilding from stale DB blocks.
New architecture:
- User-uploaded DOCX becomes the source of truth (cases.active_draft_path)
- System edits via XML surgery with real Word <w:ins>/<w:del> revisions
- User can Accept/Reject each change from within Word
Components:
- docx_reviser.py: XML surgery for Track Changes (15 tests)
- docx_retrofit.py: retroactive bookmark injection with Hebrew marker
detection + heading heuristic (9 tests)
- docx_exporter.py: emits bookmarks around each of the 12 blocks
- 3 new MCP tools: apply_user_edit, list_bookmarks, revise_draft
- 4 new/updated endpoints: upload (auto-registers active draft),
/exports/revise, /exports/bookmarks, /exports/{filename}/retrofit,
/active-draft
- DB migration: cases.active_draft_path column
- UI: correct banner using real v-numbers, "מקור האמת" badge,
detailed upload toast with bookmarks_added/missing_blocks
- agents: legal-exporter (3 export modes), legal-ceo (stage G for
revision handling), legal-writer (revision mode)
Multi-tenancy:
- Works for both CMP (1xxx cases) and CMPA (8xxx/9xxx cases)
- New revise-draft skill added to both companies
- deploy-track-changes.sh syncs skills CMP ↔ CMPA
- retrofit_case.py: one-off retrofit of existing files
Tests: 34 passing (15 reviser + 9 retrofit + 4 exporter bookmarks + 6 e2e)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Active scripts (5): auto-sync-cases.sh, backup-db.sh, restore-db.sh,
notify.py, bidi_table.py
Archived (17): one-time migration/seeding scripts whose functionality
is now in MCP server or web API. Moved to scripts/.archive/
Deleted (5): zero-value scripts (duplicates, hardcoded single-case,
debug scripts)
Added scripts/SCRIPTS.md — registry of all scripts with purpose,
status, and what superseded them. CLAUDE.md updated with rule:
any script change requires SCRIPTS.md update.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
