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

@@ -1585,6 +1585,90 @@ CREATE INDEX IF NOT EXISTS idx_missing_precedents_citation_norm
 """
 
 
+# ── V41: canonical_halachot ──────────────────────────────────────────
+# Replaces the equivalent_halachot bidirectional-link model (V28) with a
+# first-class canonical entity. Instead of recording that halacha A ≡ halacha B,
+# we now have ONE canonical_halachot row that BOTH A and B point to.
+#
+# Each legal PRINCIPLE is defined ONCE here (canonical_statement = LLM-
+# synthesized abstraction, grounded in source statements per INV-AH). The
+# per-precedent halachot rows become INSTANCES that link to the canonical and
+# carry only their own quote, treatment, and context.
+#
+# Extraction pipeline change (Phase 3, separate PR): lookup-before-insert —
+# embed new extraction, cosine-search canonical_halachot (≥0.85); if match,
+# store a thin 'citation' instance; if not, create new canonical + 'original'
+# instance. This eliminates per-extraction duplication of the same principle.
+#
+# INV-DM7: authority (binding/persuasive) derived from
+#   first_established_in.precedent_level — never stored on canonical.
+# INV-G10: only 'published' canonicals reach drafting agents.
+# INV-AH: canonical_statement grounded in source statements, never invented;
+#   review_status='pending_synthesis' until chair verifies.
+# G2: equivalent_halachot (V28) deprecated post-backfill (no parallel path).
+SCHEMA_V41_SQL = """
+-- One row per unique legal principle across all precedents.
+CREATE TABLE IF NOT EXISTS canonical_halachot (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    canonical_statement TEXT NOT NULL,
+    rule_type TEXT NOT NULL DEFAULT 'interpretive',
+    practice_areas TEXT[] NOT NULL DEFAULT '{}',
+    subject_tags TEXT[] NOT NULL DEFAULT '{}',
+    embedding vector(1024),
+    review_status TEXT NOT NULL DEFAULT 'pending_synthesis'
+        CHECK (review_status IN
+            ('pending_synthesis','pending_review','approved','published','rejected')),
+    first_established_in UUID REFERENCES case_law(id) ON DELETE SET NULL,
+    instance_count INT NOT NULL DEFAULT 0,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+    updated_at TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+CREATE INDEX IF NOT EXISTS idx_canonical_halachot_status
+    ON canonical_halachot(review_status);
+CREATE INDEX IF NOT EXISTS idx_canonical_halachot_practice
+    ON canonical_halachot USING gin(practice_areas);
+CREATE INDEX IF NOT EXISTS idx_canonical_halachot_tags
+    ON canonical_halachot USING gin(subject_tags);
+CREATE INDEX IF NOT EXISTS idx_canonical_halachot_vec
+    ON canonical_halachot USING ivfflat (embedding vector_cosine_ops)
+    WITH (lists = 30);
+
+-- halachot: canonical linkage + role columns.
+-- canonical_id: NULL until backfill_canonical_halachot.py runs; 100% filled after.
+-- instance_type: role of this precedent's mention of the principle.
+--   'original'   = the precedent that FIRST established the principle (source)
+--   'citation'   = a later precedent that cites/applies the principle
+--   'application'= a later precedent that applies the principle to new facts
+-- treatment: how this precedent's mention relates to the canonical principle.
+--   Parallels halacha_citation_corroboration.treatment (X11) but for precedents
+--   (X11 tracks citations from internal decisions; this tracks per-precedent treatment).
+-- rule_statement + embedding become nullable: citation instances inherit these
+--   from canonical_halachot. 'original' instances keep their own stored values.
+ALTER TABLE halachot
+    ADD COLUMN IF NOT EXISTS canonical_id UUID
+        REFERENCES canonical_halachot(id) ON DELETE SET NULL,
+    ADD COLUMN IF NOT EXISTS instance_type TEXT NOT NULL DEFAULT 'original'
+        CHECK (instance_type IN ('original','citation','application')),
+    ADD COLUMN IF NOT EXISTS treatment TEXT NOT NULL DEFAULT 'mentioned';
+ALTER TABLE halachot ALTER COLUMN rule_statement DROP NOT NULL;
+ALTER TABLE halachot ALTER COLUMN embedding DROP NOT NULL;
+CREATE INDEX IF NOT EXISTS idx_halachot_canonical ON halachot(canonical_id);
+CREATE INDEX IF NOT EXISTS idx_halachot_instance_type ON halachot(instance_type);
+
+-- halacha_citation_corroboration (X11) gains canonical_id so the signal
+-- aggregates at the principle level rather than the per-instance level.
+-- Backfill: UPDATE halacha_citation_corroboration SET canonical_id =
+--   (SELECT canonical_id FROM halachot WHERE id = halacha_id).
+-- halacha_id is retained for audit trail.
+ALTER TABLE halacha_citation_corroboration
+    ADD COLUMN IF NOT EXISTS canonical_id UUID
+        REFERENCES canonical_halachot(id) ON DELETE CASCADE;
+CREATE INDEX IF NOT EXISTS idx_hcc_canonical
+    ON halacha_citation_corroboration(canonical_id)
+    WHERE canonical_id IS NOT NULL;
+"""
+
+
 # Stable, arbitrary key for the session-level advisory lock that serialises
 # schema DDL across processes. Every short-lived process (cron drains, services)
 # re-runs the idempotent migrations on startup; without this lock two processes
@@ -1602,7 +1686,7 @@ async def _run_schema_migrations(pool: asyncpg.Pool) -> None:
             await _apply_schema_ddl(conn)
         finally:
             await conn.execute("SELECT pg_advisory_unlock($1)", _MIGRATION_LOCK_KEY)
-    logger.info("Database schema initialized (v1-v40)")
+    logger.info("Database schema initialized (v1-v41)")
 
 
 async def _apply_schema_ddl(conn: asyncpg.Connection) -> None:
@@ -1647,6 +1731,7 @@ async def _apply_schema_ddl(conn: asyncpg.Connection) -> None:
         await conn.execute(SCHEMA_V38_SQL)
         await conn.execute(SCHEMA_V39_SQL)
         await conn.execute(SCHEMA_V40_SQL)
+        await conn.execute(SCHEMA_V41_SQL)
 
 
 async def init_schema() -> None:
@@ -5764,12 +5849,20 @@ async def store_corroboration(
     s_id = _UUID(source_id) if isinstance(source_id, str) else source_id
     cl_id = _UUID(citing_case_law_id) if (citing_case_law_id and isinstance(citing_case_law_id, str)) else citing_case_law_id
     d_id = _UUID(citing_decision_id) if (citing_decision_id and isinstance(citing_decision_id, str)) else citing_decision_id
+    # INSERT ... SELECT so we can pull canonical_id from halachot in one round-trip.
+    # canonical_id is NULL until backfill_canonical_halachot.py runs; COALESCE keeps
+    # existing canonical_id on conflict so a pre-backfill row is upgraded when the
+    # same corroboration is re-stored post-backfill.
     await pool.execute(
         "INSERT INTO halacha_citation_corroboration "
-        "(halacha_id, citing_case_law_id, citing_decision_id, source_citation_id, treatment, match_score, match_context) "
-        "VALUES ($1,$2,$3,$4,$5,$6,$7) "
+        "(halacha_id, canonical_id, citing_case_law_id, citing_decision_id, "
+        " source_citation_id, treatment, match_score, match_context) "
+        "SELECT $1, h.canonical_id, $2, $3, $4, $5, $6, $7 "
+        "FROM halachot h WHERE h.id = $1 "
         "ON CONFLICT (halacha_id, source_citation_id) DO UPDATE SET "
-        "treatment=EXCLUDED.treatment, match_score=EXCLUDED.match_score",
+        "treatment=EXCLUDED.treatment, match_score=EXCLUDED.match_score, "
+        "canonical_id=COALESCE(EXCLUDED.canonical_id, "
+        "                      halacha_citation_corroboration.canonical_id)",
         h_id, cl_id, d_id, s_id, treatment, score, context,
     )
 
@@ -5846,6 +5939,101 @@ async def list_equivalent_for_halacha(halacha_id: UUID) -> list[dict]:
     ]
 
 
+# ── Canonical halachot (V41) ─────────────────────────────────────────────────
+
+async def create_canonical_halacha(
+    statement: str,
+    rule_type: str = "interpretive",
+    practice_areas: list[str] | None = None,
+    subject_tags: list[str] | None = None,
+    embedding: list[float] | None = None,
+    first_established_in: "UUID | None" = None,
+    review_status: str = "pending_synthesis",
+) -> "UUID":
+    """Insert a new canonical principle and return its id."""
+    pool = await get_pool()
+    row = await pool.fetchrow(
+        "INSERT INTO canonical_halachot "
+        "(canonical_statement, rule_type, practice_areas, subject_tags, "
+        " embedding, first_established_in, review_status) "
+        "VALUES ($1,$2,$3,$4,$5,$6,$7) RETURNING id",
+        statement,
+        rule_type,
+        practice_areas or [],
+        subject_tags or [],
+        embedding,
+        first_established_in,
+        review_status,
+    )
+    return row["id"]
+
+
+async def nearest_canonical_halacha(
+    vec: list[float],
+    threshold: float = 0.85,
+    status_filter: tuple[str, ...] = ("approved", "published"),
+) -> "tuple[str, float] | None":
+    """Return (canonical_id, cosine_sim) of the nearest approved/published canonical
+    whose cosine similarity to `vec` meets `threshold`, or None if none qualifies.
+
+    Used by the extractor's lookup-before-insert (Phase 3) to detect whether
+    a newly extracted principle already exists in the registry.
+    """
+    pool = await get_pool()
+    row = await pool.fetchrow(
+        "SELECT id::text AS id, 1 - (embedding <=> $1) AS sim "
+        "FROM canonical_halachot "
+        "WHERE embedding IS NOT NULL AND review_status = ANY($2::text[]) "
+        "ORDER BY embedding <=> $1 LIMIT 1",
+        vec, list(status_filter),
+    )
+    if not row:
+        return None
+    sim = float(row["sim"])
+    return (row["id"], sim) if sim >= threshold else None
+
+
+async def refresh_canonical_instance_count(canonical_id: "UUID") -> None:
+    """Recount halachot rows pointing to this canonical and update instance_count."""
+    pool = await get_pool()
+    await pool.execute(
+        "UPDATE canonical_halachot SET "
+        "instance_count = (SELECT COUNT(*) FROM halachot WHERE canonical_id = $1), "
+        "updated_at = now() "
+        "WHERE id = $1",
+        canonical_id,
+    )
+
+
+async def get_canonical_halacha(canonical_id: "UUID") -> "dict | None":
+    """Fetch one canonical principle with its instance list."""
+    pool = await get_pool()
+    row = await pool.fetchrow(
+        "SELECT ch.id::text, ch.canonical_statement, ch.rule_type, "
+        "       ch.practice_areas, ch.subject_tags, ch.review_status, "
+        "       ch.instance_count, ch.created_at, ch.updated_at, "
+        "       cl.case_number AS first_established_case "
+        "FROM canonical_halachot ch "
+        "LEFT JOIN case_law cl ON cl.id = ch.first_established_in "
+        "WHERE ch.id = $1",
+        canonical_id,
+    )
+    if not row:
+        return None
+    instances = await pool.fetch(
+        "SELECT h.id::text, h.instance_type, h.treatment, h.supporting_quote, "
+        "       h.page_reference, h.review_status AS instance_status, "
+        "       cl.case_number, cl.case_name "
+        "FROM halachot h JOIN case_law cl ON cl.id = h.case_law_id "
+        "WHERE h.canonical_id = $1 ORDER BY h.instance_type, cl.case_number",
+        canonical_id,
+    )
+    return {
+        **dict(row),
+        "instances": [dict(i) for i in instances],
+    }
+
+
 async def _annotate_equivalents(pool, out: list[dict]) -> None:
     """Attach an `equivalents` list to each row (#84.2) — parallel-authority links.
 

scripts/SCRIPTS.md

@@ -64,7 +64,8 @@
 | `halacha_panel_audit.py` | python | **רשת-ביטחון לפאנל** (selective-prediction monitoring) — דוגם הלכות שאושרו ע"י הפאנל (`reviewer LIKE 'panel:%'`), מריץ עליהן **שוב** את הצבעת-ה-KEEP של 3 השופטים, ומציף כל מקרה שכעת נוטה DROP (false-keep פוטנציאלי). report-only כברירת-מחדל; `--flag` מחזיר את ה-flips ל-`pending_review` לסקירת-יו"ר. `--sample N`/`--seed`. בסיס 2026-06-07: 0/15. מיועד להרצה תקופתית (שבועי). מייבא שופטים מ-`halacha_panel_approve`. **חובה מקומי**. | תקופתי (שבועי) — ניטור |
 | `halacha_panel_calibrate.py` | python | **כיול + מדידת הפאנל** (Trust-or-Escalate, ICLR 2025). `--source live` (ברירת-מחדל): מריץ את שאלת-ה-KEEP על מדגם-הזהב ומודד מול `is_holding` precision+coverage+**split-rate** לכל מדיניות + false-keep/false-drop (מייבא שופטים מ-`halacha_panel_approve`, **חובה מקומי**). **#133/FU-5** — `--source captured`: **אפס-עלות** (בלי re-vote/LLM) — מצליב סבבים שמורים (FU-1) מול הכרעות-יו"ר (FU-2) דרך `db.panel_rounds_vs_chair` ומדווח split-rate+auto-precision **לכל סבב** (מגמת הלולאה: ככל שהרובריקה משתפרת precision נשמר ו-split יורד); משתף את `analyze_pairs` של FU-4 (מקור-יחיד). שתי המדידות מדווחות **anon-stability** (מבחן-אנונימיזציה #81.7) כמטריקת-בריאות נגד echo-chamber. `--batch`/`--limit`/`--concurrency`. | ידני — לפני חיווט `--apply` (live) / תקופתי — מעקב-לולאה (captured) |
 | `halacha_rubric_distill.py` | python | **#133/FU-4 — זיקוק-רובריקה PROPOSE-ONLY.** מצליב `halacha_panel_rounds` (FU-1, הצבעות+נימוקים) מול הכרעות-היו"ר (FU-2, seeds ב-`halacha_goldset` batch `chair-live`) דרך `db.panel_rounds_vs_chair` (read-only), מנתח דטרמיניסטית **כשלים שיטתיים** (false-keep/false-drop, פיצולים-שהוכרעו, שיעור-מחלוקת-עם-היו"ר לכל שופט), ומציע `KEEP_SYSTEM` v2 + exemplars מופשטים (claude_session מקומי, אפס עלות) כ**דוח-diff** ל-`data/learning/rubric-proposal-<ts>.md`. **לעולם לא auto-apply** — אימוץ v2 = עריכה אנושית של הקבוע דרך PR (INV-LRN1); exemplars מופשטים בלבד (INV-LRN5); הסיגנל היחיד = הכרעת-יו"ר, לא הצבעות-פאנל (anti-echo). מתחת ל-12 זוגות → "אין מספיק נתונים". `--no-llm` (סטטיסטיקה בלבד) / `--limit N`. **חובה מקומי**. | תקופתי — אחרי שהצטברו הכרעות-יו"ר על מחלוקות-פאנל |
-| `halacha_batch_reconcile.py` | python | **#82.7** — dedup חוצה-פסקים offline (שמרני, **dry-run בלבד**). dedup-on-insert משווה רק תוך-פסק; כאן סף מחמיר (cosine ≥0.95, `--cosine`) ולא-הרסני: מאתר זוגות הלכות near-duplicate בין פסקים שונים (pgvector `<=>` exact) עם איתות לקסיקלי (Jaccard/Levenshtein) ומדווח ל-CSV ב-`data/audit/` לסקירת היו"ר. לא מדלג/ממזג/מוחק. `--include-pending`. **`--link`** רושם את הזוגות שנמצאו כ-`equivalent_halachot` (parallel authority, #84.2 — קישור-מקביל ברמת-הלכה, **לא** ציטוט; idempotent, לא-הרסני). רץ עם venv של mcp-server. אומת: 800 הלכות → 5 זוגות (קושרו). | ידני — דוח-סקירה / `--link` לקישור |
+| `backfill_canonical_halachot.py` | python | **V41 — הקמת מודל ההלכות הקנוניות (חד-פעמי + idempotent).** (1) בונה רכיבים-קשורים (connected components) מ-`equivalent_halachot` (transitive closure — union-find). (2) לכל אשכול: בוחר נציג-קנוני (הכי הרבה corroboration → confidence → earliest), יוצר שורת `canonical_halachot`, ומעדכן `canonical_id` + `instance_type` לכל חברי האשכול. (3) לסינגלטונים (ללא קישורי-שוויון): 1:1 canonical. (4) מאכלס `halacha_citation_corroboration.canonical_id` מ-`halachot.canonical_id`. `--dry-run` (ברירת-מחדל, מחשב ומדווח בלבד) / `--apply` (כותב) / `--verbose`. לאחר הרצה: `canonical_statement` = ניסוח-נציג (pending_synthesis); עוקב: `backfill_canonical_synthesis.py` (Phase 4) יסנתז ניסוח-רחב דרך LLM. הרץ: `mcp-server/.venv/bin/python scripts/backfill_canonical_halachot.py --apply`. | **חד-פעמי** (לאחר deploy V41) / idempotent לפי צורך |
+| `halacha_batch_reconcile.py` | python | **#82.7** — dedup חוצה-פסקים offline (שמרני, **dry-run בלבד**). dedup-on-insert משווה רק תוך-פסק; כאן סף מחמיר (cosine ≥0.95, `--cosine`) ולא-הרסני: מאתר זוגות הלכות near-duplicate בין פסקים שונים (pgvector `<=>` exact) עם איתות לקסיקלי (Jaccard/Levenshtein) ומדווח ל-CSV ב-`data/audit/` לסקירת היו"ר. לא מדלג/ממזג/מוחק. `--include-pending`. **`--link`** רושם את הזוגות שנמצאו כ-`equivalent_halachot` (parallel authority, #84.2 — **deprecated post-V41** — השתמש ב-`backfill_canonical_halachot.py --apply` במקום). רץ עם venv של mcp-server. | **deprecated** — הוחלף ב-`backfill_canonical_halachot.py` (V41). נשמר לצורכי audit |
 | `calibrate_halacha_dedup.py` | python | **#82.1** — כיול ספי ה-dedup הלקסיקלי (#82.3) מול gold-set הניקוי. קורא `halacha-cleanup-manifest-*.csv` (זוגות duplicate↔survivor מתויגי-אדם), טוען טקסט-survivor מה-DB, ו-sweep של (jaccard_min × levenshtein_min) עם P/R/F1, מסמן את נקודת-העבודה המוגדרת. אימת ש-(0.55, 0.70) → **precision 1.0** (אפס false-merge), recall 0.30 — מתאים לאיתות-משני שחוסם auto-approve. `--manifest <path>`. רץ עם venv של mcp-server | חד-פעמי — כיול (בוצע 2026-06-06) |
 | `ab_halacha_opus48.py` | python | **A/B לא-הרסני לחילוץ הלכות (Claude)** — מריץ מחדש חילוץ הלכות על פסק-דין בודד דרך מודל/effort נבחרים (`AB_MODEL`/`AB_EFFORT`, ברירת-מחדל `claude-opus-4-8`/`xhigh`) ומשווה לסטטיסטיקות ההלכות הקיימות ב-DB **בלי למחוק/לכתוב כלום**. משכפל את `halacha_extractor.extract()` (אותם פרומפטים, בחירת-צ'אנקים, אימות-ציטוט) ומחליף רק את קריאת ה-LLM ב-`claude -p --model --effort`. מפיק `data/ab_halacha_<case>_<effort>.json`. הרצה: `DOTENV_PATH=/home/chaim/.env DATA_DIR=.../data .venv/bin/python scripts/ab_halacha_opus48.py <case_law_id>`. **ממצא 2026-05-31 (שטיין 1128-08-20):** Opus 4.8@xhigh חילץ 51 מול 124 בייצור (100% quote-verified מול 96%) אך ביטחון מכויל-נמוך יותר (חציון 0.75 מול 0.82) — ולכן **לא** מקטין את תור-האישור-הידני תחת sweep אוטו-אישור conf≥0.78 (26 מול 24). שיפור איכות, לא צמצום-תור. | ידני (החלטת מודל-חילוץ) |
 | `ab_halacha_codex.py` | python | **A/B לא-הרסני לחילוץ הלכות (Codex/gpt-5.5)** — עמית ל-`ab_halacha_opus48` אך מחליף את `claude -p` ב-`codex exec --model gpt-5.5` (אימות ChatGPT, ללא OPENAI_API_KEY). אותם פרומפטים ואותו הסקת quote-verification. הפלט האחרון של הסוכן (`-o FILE`) נפענח כ-JSON. `AB_MODEL` (default `gpt-5.5`), `AB_REASONING` low/medium/high/xhigh (default `medium`), `AB_CONCURRENCY` (default 1), `CODEX_BIN`. מפיק `data/ab_halacha_codex_<case>_<model>_<reasoning>.json`. הרצה: `DOTENV_PATH=/home/chaim/.env DATA_DIR=.../data mcp-server/.venv/bin/python scripts/ab_halacha_codex.py <case_law_id>`. **ממצא 2026-06-17 (8181-21 האוניברסיטה העברית):** gpt-5.5@medium חילץ 27 מול 28 של Opus (quote-verified 100%/100%), ביטחון חציון 0.86 מול 0.78 — אך **0 פריטים מתחת ל-0.7** (לעומת 9/28 של Opus = 32%), דבר המצביע על over-confidence. holding↑ (12 מול 7), procedural↓ (4 מול 7). **מסקנה: ריאלי כ-fallback חירום; לא מוכן לייצור ללא כיול-ביטחון.** | ידני (בנצ'מרק מודל codex) |

scripts/backfill_canonical_halachot.py

@@ -0,0 +1,236 @@
+#!/usr/bin/env python3
+"""Backfill canonical_halachot table from existing halachot + equivalent_halachot.
+
+WHAT THIS DOES
+--------------
+1. Finds connected components in equivalent_halachot (transitive closure).
+2. For each cluster of ≥2 equivalent halachot: picks a canonical representative
+   (highest approved corroboration, then highest confidence, then earliest created),
+   creates ONE canonical_halachot row, and sets canonical_id on all cluster members.
+3. For singleton halachot (not in any cluster): creates a 1:1 canonical.
+4. Updates halacha_citation_corroboration.canonical_id from halachot.canonical_id.
+5. Refreshes canonical_halachot.instance_count.
+6. Marks cluster non-representative instances as instance_type='citation'.
+
+The backfill sets canonical_statement = representative's rule_statement. A
+subsequent LLM synthesis pass (backfill_canonical_synthesis.py, Phase 4) will
+replace this with a broader synthesized statement and set review_status='pending_review'.
+Until then, review_status stays 'pending_synthesis'.
+
+IDEMPOTENCY
+-----------
+Halachot with canonical_id already set are skipped. Re-running only fills gaps.
+
+USAGE
+-----
+cd ~/legal-ai/mcp-server
+.venv/bin/python ../scripts/backfill_canonical_halachot.py            # dry-run
+.venv/bin/python ../scripts/backfill_canonical_halachot.py --apply    # execute
+.venv/bin/python ../scripts/backfill_canonical_halachot.py --apply --verbose
+"""
+from __future__ import annotations
+
+import argparse
+import asyncio
+import os
+import sys
+from collections import defaultdict
+from datetime import datetime, timezone
+from uuid import UUID
+
+sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", "mcp-server", "src"))
+
+from legal_mcp.services import db  # noqa: E402
+
+
+# ── connected-components helpers ──────────────────────────────────────
+
+def _build_components(pairs: list[tuple[UUID, UUID]]) -> list[set[UUID]]:
+    """Union-find over (a, b) pairs → list of connected-component sets."""
+    parent: dict[UUID, UUID] = {}
+
+    def find(x: UUID) -> UUID:
+        while parent.get(x, x) != x:
+            parent[x] = parent.get(parent.get(x, x), parent.get(x, x))
+            x = parent.get(x, x)
+        return x
+
+    def union(a: UUID, b: UUID) -> None:
+        ra, rb = find(a), find(b)
+        if ra != rb:
+            parent[rb] = ra
+
+    for a, b in pairs:
+        union(a, b)
+
+    groups: dict[UUID, set[UUID]] = defaultdict(set)
+    all_nodes = {n for pair in pairs for n in pair}
+    for node in all_nodes:
+        groups[find(node)].add(node)
+    return list(groups.values())
+
+
+# ── main ──────────────────────────────────────────────────────────────
+
+async def _run(apply: bool, verbose: bool) -> None:
+    pool = await db.get_pool()
+
+    async with pool.acquire() as conn:
+        # ── 1. Load equivalent_halachot pairs ────────────────────────
+        pair_rows = await conn.fetch(
+            "SELECT halacha_a, halacha_b FROM equivalent_halachot"
+        )
+        pairs: list[tuple[UUID, UUID]] = [(r["halacha_a"], r["halacha_b"]) for r in pair_rows]
+        components = _build_components(pairs)
+        clustered_ids: set[UUID] = {h for c in components for h in c}
+        print(f"equivalent_halachot pairs: {len(pairs)}")
+        print(f"connected components (clusters ≥2): {len(components)}")
+
+        # ── 2. Load all halachot that still need canonical_id ─────────
+        all_rows = await conn.fetch(
+            "SELECT h.id, h.rule_statement, h.rule_type, h.practice_areas, "
+            "       h.subject_tags, h.embedding, h.case_law_id, h.confidence, "
+            "       h.review_status, h.created_at, "
+            "       COALESCE(cor.pos, 0) AS corroboration_count "
+            "FROM halachot h "
+            "LEFT JOIN ("
+            "  SELECT halacha_id, COUNT(DISTINCT source_citation_id) FILTER "
+            "         (WHERE treatment IN ('followed','explained')) AS pos "
+            "  FROM halacha_citation_corroboration GROUP BY halacha_id"
+            ") cor ON cor.halacha_id = h.id "
+            "WHERE h.canonical_id IS NULL"
+        )
+        pending = {r["id"]: dict(r) for r in all_rows}
+        print(f"halachot without canonical_id: {len(pending)}")
+
+        if not pending:
+            print("✅ nothing to backfill — all halachot already have canonical_id.")
+            return
+
+        # ── 3. Process clusters ───────────────────────────────────────
+        def _pick_canonical(members: list[dict]) -> dict:
+            """Best representative: highest corroboration → highest confidence → earliest."""
+            return max(members, key=lambda r: (
+                1 if r["review_status"] in ("approved", "published") else 0,
+                r["corroboration_count"],
+                float(r["confidence"] or 0),
+                -r["created_at"].timestamp(),
+            ))
+
+        canonical_created = 0
+        halacha_updated = 0
+
+        for component in components:
+            members = [pending[h] for h in component if h in pending]
+            if not members:
+                continue  # cluster fully backfilled already
+
+            rep = _pick_canonical(members)
+            if verbose:
+                print(f"\n  cluster({len(members)}) rep={rep['id']} "
+                      f"corr={rep['corroboration_count']} "
+                      f"status={rep['review_status']}")
+
+            if apply:
+                canonical_id = await conn.fetchval(
+                    "INSERT INTO canonical_halachot "
+                    "(canonical_statement, rule_type, practice_areas, subject_tags, "
+                    " embedding, first_established_in, review_status, instance_count) "
+                    "VALUES ($1,$2,$3,$4,$5,$6,'pending_synthesis',$7) RETURNING id",
+                    rep["rule_statement"] or "",
+                    rep["rule_type"] or "interpretive",
+                    rep["practice_areas"] or [],
+                    rep["subject_tags"] or [],
+                    rep["embedding"],
+                    rep["case_law_id"],
+                    len(members),
+                )
+                canonical_created += 1
+
+                for m in members:
+                    itype = "original" if m["id"] == rep["id"] else "citation"
+                    await conn.execute(
+                        "UPDATE halachot SET canonical_id=$1, instance_type=$2, "
+                        "updated_at=now() WHERE id=$3",
+                        canonical_id, itype, m["id"],
+                    )
+                    halacha_updated += 1
+
+        # ── 4. Process singletons (no equivalent links) ───────────────
+        singletons = [r for r in pending.values() if r["id"] not in clustered_ids]
+        print(f"\nsingletons (no equivalent links): {len(singletons)}")
+
+        for r in singletons:
+            if verbose:
+                print(f"  singleton: {r['id']}")
+            if apply:
+                canonical_id = await conn.fetchval(
+                    "INSERT INTO canonical_halachot "
+                    "(canonical_statement, rule_type, practice_areas, subject_tags, "
+                    " embedding, first_established_in, review_status, instance_count) "
+                    "VALUES ($1,$2,$3,$4,$5,$6,'pending_synthesis',1) RETURNING id",
+                    r["rule_statement"] or "",
+                    r["rule_type"] or "interpretive",
+                    r["practice_areas"] or [],
+                    r["subject_tags"] or [],
+                    r["embedding"],
+                    r["case_law_id"],
+                )
+                canonical_created += 1
+                await conn.execute(
+                    "UPDATE halachot SET canonical_id=$1, instance_type='original', "
+                    "updated_at=now() WHERE id=$2",
+                    canonical_id, r["id"],
+                )
+                halacha_updated += 1
+
+        # ── 5. Backfill halacha_citation_corroboration.canonical_id ───
+        if apply:
+            result = await conn.execute(
+                "UPDATE halacha_citation_corroboration hcc "
+                "SET canonical_id = h.canonical_id "
+                "FROM halachot h "
+                "WHERE hcc.halacha_id = h.id "
+                "  AND hcc.canonical_id IS NULL "
+                "  AND h.canonical_id IS NOT NULL"
+            )
+            corr_updated = int(result.split()[-1])
+            print(f"\ncorroboration rows backfilled: {corr_updated}")
+
+        # ── 6. Summary ────────────────────────────────────────────────
+        if apply:
+            remaining = await conn.fetchval(
+                "SELECT COUNT(*) FROM halachot WHERE canonical_id IS NULL"
+            )
+            canonical_total = await conn.fetchval(
+                "SELECT COUNT(*) FROM canonical_halachot"
+            )
+            print(f"\n✅ backfill complete")
+            print(f"   canonical_halachot rows: {canonical_total}")
+            print(f"   halachot updated: {halacha_updated}")
+            print(f"   halachot still without canonical_id: {remaining}")
+        else:
+            cluster_halachot = sum(
+                len([m for m in c if m in pending]) for c in components
+            )
+            print(f"\n[dry-run] would create:")
+            print(f"   canonical_halachot for {len(components)} clusters "
+                  f"({cluster_halachot} halachot) + {len(singletons)} singletons")
+            print(f"   = ~{len(components) + len(singletons)} canonical principles "
+                  f"from {len(pending)} halachot instances")
+            print("   Run with --apply to execute.")
+
+
+def main() -> None:
+    ap = argparse.ArgumentParser(description=__doc__,
+                                 formatter_class=argparse.RawDescriptionHelpFormatter)
+    ap.add_argument("--apply", action="store_true",
+                    help="Actually write to DB (default: dry-run, prints only)")
+    ap.add_argument("--verbose", "-v", action="store_true",
+                    help="Print each cluster/singleton as it is processed")
+    args = ap.parse_args()
+    asyncio.run(_run(apply=args.apply, verbose=args.verbose))
+
+
+if __name__ == "__main__":
+    main()