mcp-server/src/legal_mcp/services/court_citation.py

@@ -117,6 +117,23 @@ def normalize_case_number(raw: str) -> str:
     return cleaned.replace("/", "-").strip("-")
 
 
+def case_number_from_citation(citation: str) -> str:
+    """Canonical ``case_number`` extracted from a full citation, or ``''``.
+
+    Returns the normalized number token only (e.g. ``85074-04-25``) — NEVER the
+    full citation string with party names / court / date. This is the
+    identifier-field rule from X1 (INV-ID2): a citation like
+    ``ערר (ת"א 85074-04-25) רפאל לוי ואח' נ' הוועדה … - חולון`` yields
+    ``85074-04-25``, not the whole display string.
+
+    Reuses ``classify`` (the one canonical citation parser) so callers that need
+    a case_number out of an arbitrary citation never roll their own regex (#137,
+    G2). Returns ``''`` when no number can be parsed — the caller MUST treat that
+    as "needs a manual case_number" and never fall back to the raw citation.
+    """
+    return classify(citation).case_number_norm
+
+
 def _split_filed(num_norm: str) -> tuple[str, str, str] | None:
     """Split a normalized NNNNN-MM-YY number into (file, month, year).
 

mcp-server/src/legal_mcp/services/db.py

@@ -1757,12 +1757,23 @@ def _canonical_case_number(s: str) -> str:
     Used at the write boundary for identifier-keyed corpora (internal
     committee decisions, active cases). NOT for external precedents, whose
     canonical identifier is the full citation.
+
+    Extracts the case-number TOKEN — the leading run of digits and internal
+    separators after the proceeding-type prefix — and drops any trailing display
+    text. Without this, a full citation mis-passed as the identifier (e.g.
+    ``ערר (ת"א 85074-04-25) רפאל לוי נ' …``) left party names glued to the number
+    (``85074-04-25) רפאל לוי …``), breaking equality lookups (#137, INV-ID2). A
+    clean number is returned unchanged.
     """
     s = (s or "").strip()
     m = re.search(r"\d", s)
-    if m:
+    if not m:
+        return ""
     s = s[m.start():]
-    return s.strip().replace("/", "-")
+    token = re.match(r"\d[\d/\-]*", s)
+    if token:
+        s = token.group(0)
+    return s.strip().rstrip("/-").replace("/", "-")
 
 
 def _content_hash(text: str) -> str:

mcp-server/tests/test_canonical_case_number.py

@@ -0,0 +1,34 @@
+"""Unit tests for db._canonical_case_number — #137 / INV-ID2 / X1 §1.
+
+The write-time canonicalizer must extract the case-number TOKEN only and drop
+any trailing display text (party names) that a mis-passed full citation glued
+onto the number. A clean number is returned unchanged; no month is invented.
+"""
+
+from __future__ import annotations
+
+from legal_mcp.services.db import _canonical_case_number as canon
+
+
+def test_clean_numbers_unchanged():
+    assert canon("8137-24") == "8137-24"
+    assert canon("85074-09-24") == "85074-09-24"
+    assert canon("8126-03-25") == "8126-03-25"
+    # Legacy two-part number — month is NOT invented (X1 §1).
+    assert canon("8126-25") == "8126-25"
+
+
+def test_prefix_stripped():
+    assert canon("ערר 8137/24") == "8137-24"
+    assert canon('בל"מ 85074-09-24') == "85074-09-24"
+
+
+def test_trailing_party_names_dropped():
+    # The #137 symptom: ingest left "85074-04-25) רפאל לוי …" in the identifier.
+    assert canon("85074-04-25) רפאל לוי ואח' נ' הוועדה המקומית - חולון") == "85074-04-25"
+    assert canon("8137-24 פלוני נ' אלמוני") == "8137-24"
+
+
+def test_empty_and_no_digit():
+    assert canon("") == ""
+    assert canon("ללא מספר") == ""

mcp-server/tests/test_court_citation.py

@@ -2,7 +2,11 @@
 
 from __future__ import annotations
 
-from legal_mcp.services.court_citation import classify, normalize_case_number
+from legal_mcp.services.court_citation import (
+    case_number_from_citation,
+    classify,
+    normalize_case_number,
+)
 
 
 def test_admin_filed_format_the_example():
@@ -80,6 +84,28 @@ def test_normalize_case_number():
     assert normalize_case_number("1110/20") == "1110-20"
 
 
+def test_case_number_from_citation_strips_party_names():
+    """#137 — a full ועדת-ערר citation yields ONLY the number, never the
+    display string with party names (INV-ID2). This is the exact precedent
+    1bf0bae0 that planted ``85074-04-25) רפאל לוי …`` into case_number."""
+    cit = 'ערר (ת"א 85074-04-25) רפאל לוי ואח\' נ\' הוועדה המקומית - חולון'
+    assert case_number_from_citation(cit) == "85074-04-25"
+
+
+def test_case_number_from_citation_various_forms():
+    assert case_number_from_citation('ערר (ת"א 1198-12-25) זאטוס') == "1198-12-25"
+    assert case_number_from_citation("85074-04-25") == "85074-04-25"
+    assert case_number_from_citation('בל"מ 85074-09-24') == "85074-09-24"
+    assert case_number_from_citation("ערר 8137/24") == "8137-24"
+
+
+def test_case_number_from_citation_empty_when_unparseable():
+    """No number → '' so the caller demands a manual number, never the raw
+    citation (the #137 fallback that caused the bug)."""
+    assert case_number_from_citation("") == ""
+    assert case_number_from_citation("פסק דין בלי מספר") == ""
+
+
 def test_supreme_with_net_format_triple():
     """A Supreme prefix carrying a נט-format number exposes the triple so the
     orchestrator can route it to Tier-1 (נט המשפט serves Supreme too)."""

scripts/SCRIPTS.md

@@ -21,6 +21,7 @@
 | `test_retrieval_by_name.py` | python | בדיקת אחזור-לפי-שם (#52/RC-A) — מאמת ש`search_precedent_library`/`search_internal_decisions` מדרגים את ההחלטה עצמה (אגסי) מעל מי שמצטט אותה, + רגרסיות לשאילתות מהותיות. הרצה: `DOTENV_PATH=/home/chaim/.env DATA_DIR=.../data mcp-server/.venv/bin/python scripts/test_retrieval_by_name.py` (exit 0 = עבר). | ידני אחרי שינוי שכבת חיפוש |
 | `fu2b_reconcile_internal_case_numbers.py` | python | **FU-2b (GAP-07/08) — תיאום `case_number` של `internal_committee`** מציטוט-מלא למספר-בסיס קנוני (X1: trim·prefix-strip·`/`→`-`, חודש נשמר). דטרמיניסטי (token יחיד; 0/>1 → flag). `--dry-run` (ברירת-מחדל) מפיק טבלת-תיאום ל-`data/audit/fu2b-reconciliation-*.{csv,md}` עם flags (DUP_CHECK / PROC_MISMATCH / MISMATCH). `--apply --approved <csv>` מגבה ואז מעדכן רק שורות שאושרו ע"י היו"ר. scope: internal בלבד (external → #68). FK-safe. | חד-פעמי, **chair-gated** (apply רק אחרי אישור דפנה) |
 | `fu2c_reconcile_external_case_numbers.py` | python | **FU-2c (GAP-08, #68) — תיאום `case_number` של פסיקה חיצונית** (`source_kind <> internal_committee`) מציטוט-מלא לצורה קנונית **מציין-הליך + docket** (החלטת-יו"ר 2026-05-31, Option A: `/` נשמר, *לא* `-`; תואם db.py:369 ו-INV-ID2). דטרמיניסטי (designator+docket; 0/>1 docket → flag). `--dry-run` (ברירת-מחדל) מפיק `data/audit/fu2c-reconciliation-*.{csv,md}` עם flags (MISMATCH / NO_CITATION / CIT_NO_DOCKET / DESIG_MISMATCH / DUP_CHECK). `--apply --approved <csv>` מגבה ואז מעדכן שורות לא-חוסמות (כולל ADVISORY/NO_CITATION). `--overrides <csv>` (id,proposed_canonical,reason) פותח שורות-חוסמות בהכרעת-יו"ר מפורשת (למשל פס"ד מאוחד — ראה `data/audit/fu2c-overrides.csv` לרשומת לויתן/קלמנוביץ). לוגיקת-החילוץ + פיצול flags אומתו offline על 24 רשומות. scope: external בלבד (internal = FU-2b). FK-safe. | חד-פעמי, **chair-gated** (apply רק אחרי אישור דפנה) |
+| `fix_137_committee_case_number.py` | python | **#137 — תיקון-נתון חד-פעמי**: רשומת `internal_committee` בודדת (1bf0bae0) שבה ציטוט-מלא זיהם את שדה-המזהה (case_number=`85074/0425`, case_name=ציטוט שלם) — הפרת INV-ID2 ממסלול `missing_precedent_upload` (לפני תיקון-הקוד ב-#137). מתקן `case_number`→`85074-04-25`, `case_name`→צדדים, ו-token ב-`citation_formatted`. אומת היחיד עם `_canonical_case_number(num)≠num` ב-internal_committee (138 ה"מזוהמים" האחרים = מקור-חיצוני/cited_only מקודמים-קידומת, X1 §5 — מחוץ-לתחום). `document_id=NULL`, 0 ציטוטים-נכנסים → ללא נתיב/קובץ לשנות. guard-התנגשות על `(case_number,proceeding_type)`. אידמפוטנטי, dry-run כברירת-מחדל / `--apply`. הרצה: `HOME=/home/chaim PYTHONPATH=mcp-server/src mcp-server/.venv/bin/python scripts/fix_137_committee_case_number.py --apply`. | חד-פעמי (בוצע 2026-06-15) |
 | `eval_gold_bootstrap.py` | python | **FU-5 (GAP-11) — bootstrap ל-gold-set** של הערכת-אחזור ל-`data/eval/gold-set.jsonl`. שני מקורות: `--source citations` (cited==relevant מ-`search_relevance_feedback`; ריק עד שייצברו ציטוטים) ו-`--source known_item` (query=שם-תיק → relevant=עצמו; אות אמיתי היום). Idempotent — שומר שורות `source=chair`, מחדש `bootstrap_*`. דורש POSTGRES. | לפני eval; חוזר כשנצבר ground-truth |
 | `eval_retrieval.py` | python | **FU-5 (GAP-11, INV-RET4/G8) — harness הערכת-אחזור** — מריץ את מסלול-האחזור בייצור (`search_library`/`search_internal`) על ה-gold-set, מחשב precision@k/recall@k/MRR/nDCG@k (k=5,10), מצרף overall+per-corpus+per-PA ל-`data/eval/eval-report-<ts>.{json,md}` + delta מול `data/eval/baseline.json` (מתעד retrieval_config). `--self-test` בודק את המטריקות offline; `--update-baseline` מאמץ snapshot. **שער-CI במשמעת:** הרץ לפני/אחרי כל שינוי בשכבת-האחזור באותו קונפיג. דורש POSTGRES+VOYAGE_API_KEY. | לפני/אחרי שינוי RRF/k/embedder/rerank |
 | `legal-court-fetch-service.config.cjs` | pm2/js | **שירות-מארח Tier-1 לאחזור פסקי-דין מנט המשפט (X13)** — 2 apps: (א) `legal-court-fetch-xvfb` (Xvfb :99, צג-וירטואלי ל-Camoufox); (ב) `legal-court-fetch-service` (`python -m legal_mcp.court_fetch_service.server`, bound `10.0.1.1:8771`, Bearer `COURT_FETCH_SHARED_SECRET` מ-`~/.legal-court-fetch-service.env`, `DISPLAY=:99`). מריץ Camoufox דרך חבילת-הפייתון (in-process) כי הקונטיינר לא יכול דפדפן. תלות: `pip install -e "mcp-server[court-fetch]" && python -m camoufox fetch`. אחזור = ניווט→צופה→`GetImages`(X-Requested-With)→PDF, ללא CAPTCHA; כשל→`ok:false`→orchestrator מסלים ל-fallback אנושי. **אומת על עת"מ 46111-12-22 (34 עמ').** מראָה לדפוס `legal-chat-service.config.cjs`. ספ: `docs/spec/X13-court-fetch.md`. התקנה: `pm2 start scripts/legal-court-fetch-service.config.cjs && pm2 save`. בריאות: `curl http://10.0.1.1:8771/health`. | pm2 (host-side) |

scripts/fix_137_committee_case_number.py

@@ -0,0 +1,100 @@
+#!/usr/bin/env python3
+"""One-off data fix for TaskMaster #137 — committee precedent whose identifier
+field was polluted by a full citation.
+
+Background: ``missing_precedent_upload`` (committee branch) fell back to the raw
+``citation`` when the form left ``case_number`` blank, so a citation like
+``ערר (ת"א 85074-04-25) רפאל לוי ואח' נ' הוועדה … - חולון`` landed as the
+identifier / display name (INV-ID2 violation). The code fix (PR for #137)
+prevents recurrence; this script corrects the one already-stored row.
+
+Scope: a SINGLE internal_committee row (1bf0bae0). Verified the only
+internal_committee row where ``_canonical_case_number(case_number) != case_number``
+(the legacy 138 "polluted" matches are external/cited_only PREFIXED numbers —
+the X1 §5 external-identifier item, deliberately out of scope here). The row has
+``document_id = NULL`` (no file/storage key to rename) and 0 incoming citations,
+so only three columns change: ``case_number``, ``case_name``, ``citation_formatted``.
+
+Idempotent: if the row already carries the canonical number it is a no-op.
+Dry-run by default; pass ``--apply`` to write.
+
+Run (local, reads ~/.env for POSTGRES_URL):
+    HOME=/home/chaim PYTHONPATH=mcp-server/src python scripts/fix_137_committee_case_number.py [--apply]
+"""
+
+from __future__ import annotations
+
+import asyncio
+import sys
+
+import asyncpg
+
+from legal_mcp import config
+
+# The verified target row + its corrected values (see module docstring).
+CASE_LAW_ID = "1bf0bae0-1cb7-4110-ba1b-b956e42b0355"
+BAD_CASE_NUMBER = "85074/0425"
+NEW_CASE_NUMBER = "85074-04-25"
+NEW_CASE_NAME = "רפאל לוי ואח' נ' הוועדה המקומית לתכנון ובניה - חולון"
+
+
+async def main(apply: bool) -> int:
+    conn = await asyncpg.connect(config.POSTGRES_URL)
+    try:
+        row = await conn.fetchrow(
+            "SELECT case_number, case_name, citation_formatted, proceeding_type "
+            "FROM case_law WHERE id = $1",
+            CASE_LAW_ID,
+        )
+        if row is None:
+            print(f"row {CASE_LAW_ID} not found — nothing to do")
+            return 0
+
+        if row["case_number"] == NEW_CASE_NUMBER:
+            print(f"already canonical (case_number={NEW_CASE_NUMBER!r}) — no-op")
+            return 0
+
+        if row["case_number"] != BAD_CASE_NUMBER:
+            print(
+                f"UNEXPECTED current case_number={row['case_number']!r} "
+                f"(expected {BAD_CASE_NUMBER!r}) — refusing to guess; inspect manually"
+            )
+            return 1
+
+        # Collision guard: the (case_number, proceeding_type) partial-unique key.
+        clash = await conn.fetchval(
+            "SELECT id FROM case_law WHERE source_kind='internal_committee' "
+            "AND case_number = $1 AND proceeding_type = $2 AND id <> $3",
+            NEW_CASE_NUMBER, row["proceeding_type"], CASE_LAW_ID,
+        )
+        if clash:
+            print(f"COLLISION: {NEW_CASE_NUMBER!r}/{row['proceeding_type']!r} "
+                  f"already exists as {clash} — aborting")
+            return 1
+
+        new_citation = (row["citation_formatted"] or "").replace(
+            BAD_CASE_NUMBER, NEW_CASE_NUMBER)
+
+        print("WILL UPDATE:")
+        print(f"  case_number: {row['case_number']!r} -> {NEW_CASE_NUMBER!r}")
+        print(f"  case_name:   {row['case_name']!r}\n           -> {NEW_CASE_NAME!r}")
+        print(f"  citation_formatted: {row['citation_formatted']!r}\n"
+              f"                   -> {new_citation!r}")
+
+        if not apply:
+            print("\n(dry-run — pass --apply to write)")
+            return 0
+
+        await conn.execute(
+            "UPDATE case_law SET case_number = $2, case_name = $3, "
+            "citation_formatted = $4 WHERE id = $1",
+            CASE_LAW_ID, NEW_CASE_NUMBER, NEW_CASE_NAME, new_citation,
+        )
+        print("\n✓ updated")
+        return 0
+    finally:
+        await conn.close()
+
+
+if __name__ == "__main__":
+    sys.exit(asyncio.run(main("--apply" in sys.argv)))

web/app.py

@@ -7081,6 +7081,7 @@ async def bulletin_upload(file: UploadFile = File(...)):
 
 
 from legal_mcp.services import internal_decisions as int_decisions_service  # noqa: E402
+from legal_mcp.services import court_citation  # noqa: E402
 
 
 @app.post("/api/internal-decisions/upload")
@@ -7917,8 +7918,20 @@ async def missing_precedent_upload(
                 or int_decisions_service._district_from_court(citation)
                 or PLACEHOLDER_PENDING_EXTRACTION
             )
-            # case_number for the committee decision (not the cited-in case)
+            # case_number for the committee decision (not the cited-in case).
-            committee_case_number = case_number.strip() or citation
+            # When the form leaves it blank, derive the canonical number from
+            # the citation — NEVER fall back to the raw citation string, which
+            # planted the full display text (party names) into the identifier
+            # field (#137, INV-ID2). If no number can be parsed, demand one.
+            committee_case_number = (
+                case_number.strip()
+                or court_citation.case_number_from_citation(citation)
+            )
+            if not committee_case_number:
+                raise HTTPException(
+                    400,
+                    "לא ניתן לחלץ מספר-תיק מהציטוט — נא להזין מספר-תיק ידנית.",
+                )
             result = await int_decisions_service.ingest_internal_decision(
                 case_number=committee_case_number,
                 case_name=(case_name.strip() or mp.get("case_name") or "").strip(),