Qualifier : corrections R2/R6 + router + modèle DB + wizard frontend

Corrections moteur (TDD) : - R2 : within_mandate → record_in_observatory=True (Observatoire des décisions) - R6 : >50 personnes → collective recommandé, pas obligatoire (confidence=recommended) - R3 supprimée : affected_count=1 hors périmètre de l'outil - R9-R12 renommés G1-G4 (garde-fous internes) - 23 tests, 213/213 verts Étape 1 — Router /api/v1/qualify : - POST / → qualify() avec config depuis DB ou defaults - GET /protocol → protocole actif - POST /protocol → créer/remplacer (auth requise) Étape 2 — Modèle QualificationProtocol : - Table qualification_protocols (seuils configurables via admin) - Migration Alembic + seed du protocole par défaut Étape 3 — Wizard frontend decisions/new.vue : - Étape 1 : formulaire de qualification (mandat, affected_count, structurant, contexte) - Étape 2 : résultat (type, raisons, modalités, observatoire, on-chain) - Étape 3 : formulaire de décision (titre, description, protocole si collectif) Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-04-23 19:12:01 +02:00
parent 428299c9c8
commit 5c51cffc93
11 changed files with 1060 additions and 519 deletions
--- a/backend/alembic/env.py
+++ b/backend/alembic/env.py
@@ -39,6 +39,7 @@ from app.models import (  # noqa: F401
    MandateStep,
    VotingProtocol,
    FormulaConfig,
+    QualificationProtocol,
    SanctuaryEntry,
    BlockchainCache,
 )
--- a/backend/alembic/versions/2026_04_23_1708-b78571ae9e00_add_qualification_protocol.py
+++ b/backend/alembic/versions/2026_04_23_1708-b78571ae9e00_add_qualification_protocol.py
@@ -0,0 +1,39 @@
+"""add_qualification_protocol
+
+Revision ID: b78571ae9e00
+Revises: 70914b334cfb
+Create Date: 2026-04-23 17:08:07.161306+00:00
+"""
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = 'b78571ae9e00'
+down_revision: Union[str, None] = '70914b334cfb'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        'qualification_protocols',
+        sa.Column('id', sa.Uuid(), nullable=False),
+        sa.Column('name', sa.String(128), nullable=False),
+        sa.Column('description', sa.Text(), nullable=True),
+        sa.Column('small_group_max', sa.Integer(), nullable=False, server_default='5'),
+        sa.Column('collective_wot_min', sa.Integer(), nullable=False, server_default='50'),
+        sa.Column(
+            'default_modalities_json',
+            sa.Text(),
+            nullable=False,
+            server_default='["vote_wot","vote_smith","consultation_avis","election"]',
+        ),
+        sa.Column('is_active', sa.Boolean(), nullable=False, server_default='1'),
+        sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.func.now()),
+        sa.PrimaryKeyConstraint('id'),
+    )
+
+
+def downgrade() -> None:
+    op.drop_table('qualification_protocols')
--- a/backend/app/engine/qualifier.py
+++ b/backend/app/engine/qualifier.py
@@ -18,7 +18,7 @@ class DecisionType(str, Enum):


 # ---------------------------------------------------------------------------
-# Configuration (thresholds — stored as a QualificationProtocol in DB)
+# Configuration (thresholds — stored as QualificationProtocol in DB)
 # ---------------------------------------------------------------------------


@@ -26,16 +26,15 @@ class DecisionType(str, Enum):
 class QualificationConfig:
    """Configurable thresholds for the qualification engine.

-    These defaults will be seeded as a QualificationProtocol record so they
-    can be adjusted through the admin interface without code changes.
+    Seeded as a QualificationProtocol record so they can be adjusted
+    through the admin interface without code changes.

-    individual_max:      affected_count <= this → always individual
    small_group_max:    affected_count <= this → individual recommended, collective available
-    collective_wot_min:  affected_count > this  → collective required (WoT formula applies)
+    collective_wot_min: affected_count > this  → WoT formula applicable (still recommended, not required)

-    Default modalities shown when collective is chosen (ordered by relevance).
+    affected_count must be >= 2 — decisions affecting only the author
+    have no place in this tool.
    """
-    individual_max: int = 1
    small_group_max: int = 5
    collective_wot_min: int = 50

@@ -55,7 +54,7 @@ class QualificationConfig:
@dataclass
 class QualificationInput:
    within_mandate: bool = False
-    affected_count: int | None = None
+    affected_count: int | None = None   # must be >= 2 when within_mandate=False
    is_structural: bool = False
    context_description: str | None = None  # reserved for LLM suggestion

@@ -66,10 +65,11 @@ class QualificationResult:
    process: str
    recommended_modalities: list[str]
    recommend_onchain: bool
+    onchain_reason: str | None
    confidence: str              # "required" | "recommended" | "optional"
    collective_available: bool
+    record_in_observatory: bool  # True → decision must be logged in Observatoire
    reasons: list[str]
-    onchain_reason: str | None = None


 # ---------------------------------------------------------------------------
@@ -83,9 +83,9 @@ def suggest_modalities_from_context(
 ) -> list[str]:
    """Suggest voting modalities based on a natural-language context description.

-    Stub — returns empty list until local Qwen (qwen3.6) is integrated.
-    When implemented, this will call the LLM API and return an ordered list
-    of modality slugs from config.default_modalities.
+    Stub — returns empty list until local Qwen (qwen3.6, MacStudio) is integrated.
+    When implemented, will call the LLM API and return an ordered subset of
+    config.default_modalities ranked by contextual relevance.
    """
    return []

@@ -99,11 +99,11 @@ def qualify(inp: QualificationInput, config: QualificationConfig) -> Qualificati
    """Qualify a decision and recommend a type, process, and modalities.

    Rules (in priority order):
-      R1/R2  within_mandate → individual + consultation_avis, no modalities
-      R3     affected_count == 1 → individual + personal
-      R4     affected_count ≤ small_group_max → individual recommended, collective available
+      R1/R2  within_mandate → individual + consultation_avis, no vote modalities,
+             decision must be recorded in Observatoire des décisions
+      R4     2 ≤ affected_count ≤ small_group_max → individual recommended, collective available
      R5     small_group_max < affected_count ≤ collective_wot_min → collective recommended
-      R6     affected_count > collective_wot_min → collective required (WoT)
+      R6     affected_count > collective_wot_min → collective recommended (WoT formula applicable)
      R7/R8  is_structural → recommend_onchain with reason
    """
    reasons: list[str] = []
@@ -116,27 +116,14 @@ def qualify(inp: QualificationInput, config: QualificationConfig) -> Qualificati
            process="consultation_avis",
            recommended_modalities=[],
            recommend_onchain=_onchain(inp, reasons),
+            onchain_reason=_onchain_reason(inp),
            confidence="required",
            collective_available=False,
+            record_in_observatory=True,
            reasons=reasons,
-            onchain_reason=_onchain_reason(inp),
        )

-    count = inp.affected_count if inp.affected_count is not None else 1
-
-    # ── R3: single person ───────────────────────────────────────────────────
-    if count <= config.individual_max:
-        reasons.append("Une seule personne concernée.")
-        return QualificationResult(
-            decision_type=DecisionType.INDIVIDUAL,
-            process="personal",
-            recommended_modalities=[],
-            recommend_onchain=_onchain(inp, reasons),
-            confidence="required",
-            collective_available=False,
-            reasons=reasons,
-            onchain_reason=_onchain_reason(inp),
-        )
+    count = inp.affected_count if inp.affected_count is not None else 2

    # ── R4: small group → individual recommended, collective available ───────
    if count <= config.small_group_max:
@@ -144,32 +131,30 @@ def qualify(inp: QualificationInput, config: QualificationConfig) -> Qualificati
            f"{count} personnes concernées : décision individuelle recommandée, "
            "vote collectif possible."
        )
-        modalities = _collect_modalities(inp, config)
        return QualificationResult(
            decision_type=DecisionType.INDIVIDUAL,
            process="personal",
            recommended_modalities=[],
            recommend_onchain=_onchain(inp, reasons),
+            onchain_reason=_onchain_reason(inp),
            confidence="recommended",
            collective_available=True,
+            record_in_observatory=False,
            reasons=reasons,
-            onchain_reason=_onchain_reason(inp),
        )

    # ── R5/R6: medium or large group → collective ────────────────────────────
    modalities = _collect_modalities(inp, config)

    if count <= config.collective_wot_min:
-        reasons.append(
-            f"{count} personnes concernées : vote collectif recommandé."
-        )
+        reasons.append(f"{count} personnes concernées : vote collectif recommandé.")
        confidence = "recommended"
    else:
        reasons.append(
-            f"{count} personnes concernées : vote collectif obligatoire "
-            "(formule WoT applicable)."
+            f"{count} personnes concernées : vote collectif recommandé "
+            "(formule WoT applicable à cette échelle)."
        )
-        confidence = "required"
+        confidence = "recommended"
        if "vote_wot" not in modalities:
            modalities = ["vote_wot"] + modalities

@@ -178,10 +163,11 @@ def qualify(inp: QualificationInput, config: QualificationConfig) -> Qualificati
        process="vote_collective",
        recommended_modalities=modalities,
        recommend_onchain=_onchain(inp, reasons),
+        onchain_reason=_onchain_reason(inp),
        confidence=confidence,
        collective_available=True,
+        record_in_observatory=False,
        reasons=reasons,
-        onchain_reason=_onchain_reason(inp),
    )


@@ -210,10 +196,7 @@ def _onchain_reason(inp: QualificationInput) -> str | None:
    )


-def _collect_modalities(
-    inp: QualificationInput,
-    config: QualificationConfig,
-) -> list[str]:
+def _collect_modalities(inp: QualificationInput, config: QualificationConfig) -> list[str]:
    """Combine default modalities with any LLM suggestions (stub for now)."""
    llm_suggestions = []
    if inp.context_description:
--- a/backend/app/main.py
+++ b/backend/app/main.py
@@ -14,6 +14,7 @@ from app.middleware.security_headers import SecurityHeadersMiddleware
 from app.routers import auth, documents, decisions, votes, mandates, protocols, sanctuary, websocket
 from app.routers import public
 from app.routers import organizations
+from app.routers import qualify


 # ── Structured logging setup ───────────────────────────────────────────────
@@ -133,6 +134,7 @@ app.include_router(sanctuary.router, prefix="/api/v1/sanctuary", tags=["sanctuar
 app.include_router(websocket.router, prefix="/api/v1/ws", tags=["websocket"])
 app.include_router(public.router, prefix="/api/v1/public", tags=["public"])
 app.include_router(organizations.router, prefix="/api/v1/organizations", tags=["organizations"])
+app.include_router(qualify.router, prefix="/api/v1/qualify", tags=["qualify"])


 # ── Health check ─────────────────────────────────────────────────────────
--- a/backend/app/models/init.py
+++ b/backend/app/models/init.py
@@ -5,6 +5,7 @@ from app.models.decision import Decision, DecisionStep
 from app.models.vote import VoteSession, Vote
 from app.models.mandate import Mandate, MandateStep
 from app.models.protocol import VotingProtocol, FormulaConfig
+from app.models.qualification import QualificationProtocol
 from app.models.sanctuary import SanctuaryEntry
 from app.models.cache import BlockchainCache

@@ -16,6 +17,7 @@ __all__ = [
    "VoteSession", "Vote",
    "Mandate", "MandateStep",
    "VotingProtocol", "FormulaConfig",
+    "QualificationProtocol",
    "SanctuaryEntry",
    "BlockchainCache",
 ]
--- a/backend/app/models/qualification.py
+++ b/backend/app/models/qualification.py
@@ -0,0 +1,38 @@
+import json
+import uuid
+from datetime import datetime
+
+from sqlalchemy import Boolean, DateTime, Integer, String, Text, Uuid, func
+from sqlalchemy.orm import Mapped, mapped_column
+
+from app.database import Base
+
+
+class QualificationProtocol(Base):
+    """Active configuration for the decision qualification engine.
+
+    Thresholds stored here override the engine defaults and can be updated
+    through the admin interface (meta-governance).
+    Only one record should be active at a time (is_active=True).
+    """
+    __tablename__ = "qualification_protocols"
+
+    id: Mapped[uuid.UUID] = mapped_column(Uuid, primary_key=True, default=uuid.uuid4)
+    name: Mapped[str] = mapped_column(String(128), nullable=False)
+    description: Mapped[str | None] = mapped_column(Text)
+
+    small_group_max: Mapped[int] = mapped_column(Integer, default=5)
+    collective_wot_min: Mapped[int] = mapped_column(Integer, default=50)
+
+    # JSON array of modality slugs, e.g. '["vote_wot","vote_smith","election"]'
+    default_modalities_json: Mapped[str] = mapped_column(
+        Text,
+        default='["vote_wot","vote_smith","consultation_avis","election"]',
+    )
+
+    is_active: Mapped[bool] = mapped_column(Boolean, default=True)
+    created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), server_default=func.now())
+
+    @property
+    def default_modalities(self) -> list[str]:
+        return json.loads(self.default_modalities_json)
--- a/backend/app/routers/qualify.py
+++ b/backend/app/routers/qualify.py
@@ -0,0 +1,110 @@
+"""Qualify router: decision qualification engine endpoint."""
+
+from __future__ import annotations
+
+from dataclasses import asdict
+
+from fastapi import APIRouter, Depends
+from sqlalchemy import select
+from sqlalchemy.ext.asyncio import AsyncSession
+
+from app.database import get_db
+from app.engine.qualifier import QualificationConfig, QualificationInput, qualify
+from app.models.qualification import QualificationProtocol
+from app.schemas.qualification import (
+    QualificationProtocolCreate,
+    QualificationProtocolOut,
+    QualifyRequest,
+    QualifyResponse,
+)
+from app.services.auth_service import get_current_identity
+
+router = APIRouter()
+
+
+async def _load_config(db: AsyncSession) -> QualificationConfig:
+    """Load the active QualificationProtocol from DB, or fall back to defaults."""
+    result = await db.execute(
+        select(QualificationProtocol)
+        .where(QualificationProtocol.is_active == True)  # noqa: E712
+        .order_by(QualificationProtocol.created_at.desc())
+        .limit(1)
+    )
+    proto = result.scalar_one_or_none()
+    if proto is None:
+        return QualificationConfig()
+    return QualificationConfig(
+        small_group_max=proto.small_group_max,
+        collective_wot_min=proto.collective_wot_min,
+        default_modalities=proto.default_modalities,
+    )
+
+
+@router.post("/", response_model=QualifyResponse)
+async def qualify_decision(
+    payload: QualifyRequest,
+    db: AsyncSession = Depends(get_db),
+) -> QualifyResponse:
+    """Qualify a decision: determine type, process, and modalities.
+
+    No authentication required — this is an advisory endpoint that helps
+    users understand which decision pathway fits their situation.
+    """
+    config = await _load_config(db)
+    inp = QualificationInput(
+        within_mandate=payload.within_mandate,
+        affected_count=payload.affected_count,
+        is_structural=payload.is_structural,
+        context_description=payload.context_description,
+    )
+    result = qualify(inp, config)
+    return QualifyResponse(**asdict(result))
+
+
+@router.get("/protocol", response_model=QualificationProtocolOut | None)
+async def get_active_protocol(
+    db: AsyncSession = Depends(get_db),
+) -> QualificationProtocolOut | None:
+    """Return the currently active qualification protocol (thresholds)."""
+    result = await db.execute(
+        select(QualificationProtocol)
+        .where(QualificationProtocol.is_active == True)  # noqa: E712
+        .order_by(QualificationProtocol.created_at.desc())
+        .limit(1)
+    )
+    proto = result.scalar_one_or_none()
+    if proto is None:
+        return None
+    return QualificationProtocolOut.model_validate(proto)
+
+
+@router.post("/protocol", response_model=QualificationProtocolOut, status_code=201)
+async def create_protocol(
+    payload: QualificationProtocolCreate,
+    db: AsyncSession = Depends(get_db),
+    _identity=Depends(get_current_identity),
+) -> QualificationProtocolOut:
+    """Create a new qualification protocol (requires auth).
+
+    Deactivates the current active protocol before saving the new one.
+    """
+    # Deactivate current
+    current = await db.execute(
+        select(QualificationProtocol).where(QualificationProtocol.is_active == True)  # noqa: E712
+    )
+    for proto in current.scalars().all():
+        proto.is_active = False
+
+    import json
+    proto = QualificationProtocol(
+        name=payload.name,
+        description=payload.description,
+        small_group_max=payload.small_group_max,
+        collective_wot_min=payload.collective_wot_min,
+        default_modalities_json=json.dumps(payload.default_modalities),
+        is_active=True,
+    )
+    db.add(proto)
+    await db.commit()
+    await db.refresh(proto)
+    return QualificationProtocolOut.model_validate(proto)
--- a/backend/app/schemas/qualification.py
+++ b/backend/app/schemas/qualification.py
@@ -0,0 +1,48 @@
+from __future__ import annotations
+
+from datetime import datetime
+from uuid import UUID
+
+from pydantic import BaseModel, ConfigDict, Field
+
+
+class QualifyRequest(BaseModel):
+    within_mandate: bool = False
+    affected_count: int | None = Field(default=None, ge=2, description="Nombre de personnes concernées (minimum 2)")
+    is_structural: bool = False
+    context_description: str | None = Field(default=None, max_length=2000)
+
+
+class QualifyResponse(BaseModel):
+    decision_type: str
+    process: str
+    recommended_modalities: list[str]
+    recommend_onchain: bool
+    onchain_reason: str | None
+    confidence: str
+    collective_available: bool
+    record_in_observatory: bool
+    reasons: list[str]
+
+
+class QualificationProtocolOut(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: UUID
+    name: str
+    description: str | None = None
+    small_group_max: int
+    collective_wot_min: int
+    default_modalities: list[str]
+    is_active: bool
+    created_at: datetime
+
+
+class QualificationProtocolCreate(BaseModel):
+    name: str = Field(..., min_length=1, max_length=128)
+    description: str | None = None
+    small_group_max: int = Field(default=5, ge=1)
+    collective_wot_min: int = Field(default=50, ge=1)
+    default_modalities: list[str] = Field(
+        default=["vote_wot", "vote_smith", "consultation_avis", "election"]
+    )
--- a/backend/app/tests/test_qualifier.py
+++ b/backend/app/tests/test_qualifier.py
@@ -1,21 +1,21 @@
 """TDD — Moteur de qualification des décisions.

-Ce fichier est la source de vérité exécutable des règles métier du tunnel "Décider".
-Chaque test est nommé d'après la règle qu'il encode.
+Source de vérité exécutable des règles métier du tunnel "Décider".

-Règles implémentées dans ce fichier (RED → GREEN au fur et à mesure) :
+Règles testées :
  R1   within_mandate → individual + consultation_avis
-  R2  within_mandate → aucune modalité de vote n'est proposée
-  R3  affected_count == 1 (hors mandat) → individual
-  R4  affected_count ≤ small_group_max → individual recommandé / collective possible
+  R2   within_mandate → aucune modalité de vote + consignation Observatoire
+  R4   2 ≤ affected_count ≤ small_group_max → individual recommandé, collectif disponible
  R5   small_group_max < affected_count ≤ collective_wot_min → collective recommandé
-  R6  affected_count > collective_wot_min → collective WoT obligatoire
-  R7  is_structural=True → recommend_onchain=True + raison explicite
-  R8  is_structural=False → recommend_onchain=False (on ne pollue pas avec l'option)
-  R9  decision_type est toujours dans l'enum autorisé
-  R10 individual n'expose jamais de modalités de vote
-  R11 collective expose au moins une modalité
-  R12 les seuils sont lus depuis QualificationConfig (configurables)
+  R6   affected_count > collective_wot_min → collective recommandé (WoT applicable, non obligatoire)
+  R7   is_structural → recommend_onchain + raison explicite
+  R8   is_structural=False → recommend_onchain=False
+
+  GARDE-FOUS (invariants internes qui ne doivent jamais régresser) :
+  G1   decision_type est toujours dans l'enum autorisé
+  G2   individual n'expose jamais de modalités de vote
+  G3   collective expose au moins une modalité
+  G4   les seuils sont lus depuis QualificationConfig (configurables)
 """

 from __future__ import annotations
@@ -30,11 +30,7 @@ from app.engine.qualifier import (
    qualify,
 )

-# ---------------------------------------------------------------------------
-# Fixtures
-# ---------------------------------------------------------------------------
-
-DEFAULT_CONFIG = QualificationConfig()  # valeurs par défaut
+DEFAULT_CONFIG = QualificationConfig()


 # ---------------------------------------------------------------------------
@@ -43,19 +39,17 @@ DEFAULT_CONFIG = QualificationConfig()  # valeurs par défaut


 def test_r1_within_mandate_gives_individual():
-    """Une décision dans le périmètre d'un mandat est toujours individuelle."""
    result = qualify(QualificationInput(within_mandate=True), DEFAULT_CONFIG)
    assert result.decision_type == DecisionType.INDIVIDUAL


 def test_r1_within_mandate_gives_consultation_avis():
-    """La décision dans un mandat utilise le processus 'consultation_avis'."""
    result = qualify(QualificationInput(within_mandate=True), DEFAULT_CONFIG)
    assert result.process == "consultation_avis"


 def test_r1_within_mandate_overrides_large_affected_count():
-    """Même si de nombreuses personnes sont concernées, un mandat impose 'individual'."""
+    """Même si de nombreuses personnes sont concernées, un mandat impose individual."""
    result = qualify(
        QualificationInput(within_mandate=True, affected_count=500),
        DEFAULT_CONFIG,
@@ -65,35 +59,29 @@ def test_r1_within_mandate_overrides_large_affected_count():


 # ---------------------------------------------------------------------------
-# R2 — within_mandate → aucune modalité de vote
+# R2 — within_mandate → aucune modalité de vote + consignation Observatoire
 # ---------------------------------------------------------------------------


-def test_r2_within_mandate_no_modalities():
+def test_r2_within_mandate_no_vote_modalities():
    """Le mandataire décide seul après consultation — pas de vote collectif."""
    result = qualify(QualificationInput(within_mandate=True), DEFAULT_CONFIG)
    assert result.recommended_modalities == []


-# ---------------------------------------------------------------------------
-# R3 — affected_count == 1 hors mandat → individual
-# ---------------------------------------------------------------------------
+def test_r2_within_mandate_records_in_observatory():
+    """Une décision dans un mandat doit être consignée dans l'Observatoire."""
+    result = qualify(QualificationInput(within_mandate=True), DEFAULT_CONFIG)
+    assert result.record_in_observatory is True


-def test_r3_single_person_is_individual():
+def test_r2_out_of_mandate_does_not_force_observatory():
+    """Hors mandat, la consignation dans l'Observatoire n'est pas imposée."""
    result = qualify(
-        QualificationInput(within_mandate=False, affected_count=1),
+        QualificationInput(within_mandate=False, affected_count=10),
        DEFAULT_CONFIG,
    )
-    assert result.decision_type == DecisionType.INDIVIDUAL
-
-
-def test_r3_single_person_process_is_personal():
-    result = qualify(
-        QualificationInput(within_mandate=False, affected_count=1),
-        DEFAULT_CONFIG,
-    )
-    assert result.process == "personal"
+    assert result.record_in_observatory is False


 # ---------------------------------------------------------------------------
@@ -102,7 +90,6 @@ def test_r3_single_person_process_is_personal():


 def test_r4_small_group_recommends_individual():
-    """2 à small_group_max personnes : individual recommandé, collective possible."""
    for count in range(2, DEFAULT_CONFIG.small_group_max + 1):
        result = qualify(
            QualificationInput(within_mandate=False, affected_count=count),
@@ -111,9 +98,14 @@ def test_r4_small_group_recommends_individual():
        assert result.decision_type == DecisionType.INDIVIDUAL, (
            f"affected_count={count} devrait recommander individual"
        )
-        assert result.collective_available is True, (
-            f"affected_count={count} : collective doit rester disponible"
+
+
+def test_r4_small_group_collective_is_available():
+    result = qualify(
+        QualificationInput(within_mandate=False, affected_count=3),
+        DEFAULT_CONFIG,
    )
+    assert result.collective_available is True


 def test_r4_small_group_confidence_is_recommended():
@@ -148,11 +140,11 @@ def test_r5_medium_group_confidence_is_recommended():


 # ---------------------------------------------------------------------------
-# R6 — affected_count > collective_wot_min → collective WoT requis
+# R6 — affected_count > collective_wot_min → collective recommandé (pas obligatoire)
 # ---------------------------------------------------------------------------


-def test_r6_large_group_requires_collective():
+def test_r6_large_group_recommends_collective():
    result = qualify(
        QualificationInput(within_mandate=False, affected_count=DEFAULT_CONFIG.collective_wot_min + 1),
        DEFAULT_CONFIG,
@@ -160,12 +152,13 @@ def test_r6_large_group_requires_collective():
    assert result.decision_type == DecisionType.COLLECTIVE


-def test_r6_large_group_confidence_is_required():
+def test_r6_large_group_confidence_is_recommended_not_required():
+    """Au-delà du seuil WoT, le vote collectif est recommandé — pas imposé."""
    result = qualify(
        QualificationInput(within_mandate=False, affected_count=DEFAULT_CONFIG.collective_wot_min + 1),
        DEFAULT_CONFIG,
    )
-    assert result.confidence == "required"
+    assert result.confidence == "recommended"


 def test_r6_large_group_includes_vote_wot_modality():
@@ -182,7 +175,6 @@ def test_r6_large_group_includes_vote_wot_modality():


 def test_r7_structural_recommends_onchain():
-    """Décision structurante (force de loi ou action machine) → on-chain proposé."""
    result = qualify(
        QualificationInput(within_mandate=False, affected_count=100, is_structural=True),
        DEFAULT_CONFIG,
@@ -195,14 +187,13 @@ def test_r7_structural_provides_onchain_reason():
        QualificationInput(within_mandate=False, affected_count=100, is_structural=True),
        DEFAULT_CONFIG,
    )
-    assert result.onchain_reason is not None
-    assert len(result.onchain_reason) > 0
+    assert result.onchain_reason is not None and len(result.onchain_reason) > 0


-def test_r7_structural_individual_can_also_recommend_onchain():
-    """Même une décision individuelle structurante peut être gravée on-chain."""
+def test_r7_structural_within_mandate_can_also_recommend_onchain():
+    """Même une décision dans un mandat peut être gravée si structurante."""
    result = qualify(
-        QualificationInput(within_mandate=False, affected_count=1, is_structural=True),
+        QualificationInput(within_mandate=True, is_structural=True),
        DEFAULT_CONFIG,
    )
    assert result.recommend_onchain is True
@@ -214,48 +205,37 @@ def test_r7_structural_individual_can_also_recommend_onchain():


 def test_r8_non_structural_never_recommends_onchain():
-    """On ne pollue pas les décisions ordinaires avec l'option on-chain."""
-    for count in [1, 3, 10, 100]:
+    for count in [2, 3, 10, 100]:
        result = qualify(
            QualificationInput(within_mandate=False, affected_count=count, is_structural=False),
            DEFAULT_CONFIG,
        )
        assert result.recommend_onchain is False, (
-            f"affected_count={count}, is_structural=False : on-chain ne doit pas être proposé"
+            f"affected_count={count} non structurant : on-chain ne doit pas être proposé"
        )


 # ---------------------------------------------------------------------------
-# R9 — decision_type toujours dans l'enum autorisé
+# GARDE-FOUS internes (régressions silencieuses)
 # ---------------------------------------------------------------------------


-def test_r9_decision_type_always_valid():
+def test_g1_decision_type_always_valid():
    valid_types = set(DecisionType)
-    inputs = [
-        QualificationInput(within_mandate=True),
-        QualificationInput(within_mandate=False, affected_count=1),
-        QualificationInput(within_mandate=False, affected_count=3),
-        QualificationInput(within_mandate=False, affected_count=10),
-        QualificationInput(within_mandate=False, affected_count=100),
-        QualificationInput(within_mandate=False, affected_count=100, is_structural=True),
-    ]
-    for inp in inputs:
-        result = qualify(inp, DEFAULT_CONFIG)
-        assert result.decision_type in valid_types, (
-            f"Type invalide '{result.decision_type}' pour input {inp}"
-        )
-
-
-# ---------------------------------------------------------------------------
-# R10 — individual n'expose jamais de modalités
-# ---------------------------------------------------------------------------
-
-
-def test_r10_individual_no_modalities():
    for inp in [
        QualificationInput(within_mandate=True),
-        QualificationInput(within_mandate=False, affected_count=1),
+        QualificationInput(within_mandate=False, affected_count=2),
+        QualificationInput(within_mandate=False, affected_count=10),
+        QualificationInput(within_mandate=False, affected_count=100),
+    ]:
+        result = qualify(inp, DEFAULT_CONFIG)
+        assert result.decision_type in valid_types
+
+
+def test_g2_individual_never_has_vote_modalities():
+    for inp in [
+        QualificationInput(within_mandate=True),
+        QualificationInput(within_mandate=False, affected_count=2),
        QualificationInput(within_mandate=False, affected_count=3),
    ]:
        result = qualify(inp, DEFAULT_CONFIG)
@@ -265,12 +245,7 @@ def test_r10_individual_no_modalities():
            )


-# ---------------------------------------------------------------------------
-# R11 — collective expose au moins une modalité
-# ---------------------------------------------------------------------------
-
-
-def test_r11_collective_has_at_least_one_modality():
+def test_g3_collective_has_at_least_one_modality():
    result = qualify(
        QualificationInput(within_mandate=False, affected_count=20),
        DEFAULT_CONFIG,
@@ -279,19 +254,9 @@ def test_r11_collective_has_at_least_one_modality():
    assert len(result.recommended_modalities) >= 1


-# ---------------------------------------------------------------------------
-# R12 — les seuils sont lus depuis QualificationConfig
-# ---------------------------------------------------------------------------
-
-
-def test_r12_custom_config_overrides_thresholds():
-    """Les seuils viennent de QualificationConfig, pas de constantes hardcodées."""
-    custom = QualificationConfig(
-        individual_max=1,
-        small_group_max=2,    # au lieu de 5
-        collective_wot_min=10,
-    )
-    # affected_count=3 est au-delà de small_group_max=2 → collective
+def test_g4_custom_config_overrides_thresholds():
+    """Les seuils viennent de QualificationConfig — pas de constantes hardcodées."""
+    custom = QualificationConfig(small_group_max=2, collective_wot_min=10)
    result = qualify(
        QualificationInput(within_mandate=False, affected_count=3),
        custom,
@@ -299,9 +264,7 @@ def test_r12_custom_config_overrides_thresholds():
    assert result.decision_type == DecisionType.COLLECTIVE


-def test_r12_default_config_thresholds_are_expected():
-    """Valeurs par défaut documentées et vérifiables."""
+def test_g4_default_thresholds_are_stable():
    cfg = QualificationConfig()
-    assert cfg.individual_max == 1
    assert cfg.small_group_max == 5
    assert cfg.collective_wot_min == 50
--- a/backend/seed.py
+++ b/backend/seed.py
@@ -33,6 +33,7 @@ from app.models.document import Document, DocumentItem
 from app.models.decision import Decision, DecisionStep
 from app.models.mandate import Mandate, MandateStep
 from app.models.organization import Organization
+from app.models.qualification import QualificationProtocol
 from app.models.user import DuniterIdentity
 from app.models.vote import VoteSession, Vote

@@ -2459,6 +2460,37 @@ async def seed_organizations(session: AsyncSession) -> dict[str, Organization]:
 # Main seed runner
 # ---------------------------------------------------------------------------

+async def seed_qualification_protocol(session: AsyncSession) -> QualificationProtocol:
+    """Seed the default qualification protocol (thresholds for the Décider tunnel)."""
+    stmt = select(QualificationProtocol).where(QualificationProtocol.is_active == True)  # noqa: E712
+    result = await session.execute(stmt)
+    existing = result.scalar_one_or_none()
+    if existing is not None:
+        print("  [skip] Protocole de qualification déjà présent")
+        return existing
+
+    proto = QualificationProtocol(
+        name="Protocole de qualification par défaut",
+        description=(
+            "Seuils utilisés par le tunnel Décider pour router vers "
+            "individual/collective et proposer les modalités de vote."
+        ),
+        small_group_max=5,
+        collective_wot_min=50,
+        default_modalities_json=json.dumps([
+            "vote_wot",
+            "vote_smith",
+            "consultation_avis",
+            "election",
+        ]),
+        is_active=True,
+    )
+    session.add(proto)
+    await session.flush()
+    print("  [ok] Protocole de qualification créé")
+    return proto
+
+
 async def run_seed():
    print("=" * 60)
    print("libreDecision - Seed Database")
@@ -2474,6 +2506,9 @@ async def run_seed():
            orgs = await seed_organizations(session)
            duniter_g1_id = orgs["duniter-g1"].id

+            print("\n[0b] Protocole de qualification...")
+            await seed_qualification_protocol(session)
+
            print("\n[1/10] Formula Configs...")
            formulas = await seed_formula_configs(session)

--- a/frontend/app/pages/decisions/new.vue
+++ b/frontend/app/pages/decisions/new.vue