diff --git a/src/viewer/app.py b/src/viewer/app.py
index e577496..5d8b836 100644
--- a/src/viewer/app.py
+++ b/src/viewer/app.py
@@ -629,4 +629,21 @@ def create_app() -> Flask:
                                metrics=metrics, total_selection=len(selection),
                                groups=groups)
 
+    # --- Health check endpoint ---
+    @app.route("/health")
+    def health():
+        from datetime import datetime, timezone
+        ollama_ok = False
+        try:
+            r = requests.get(f"{OLLAMA_URL}/api/tags", timeout=2)
+            ollama_ok = r.status_code == 200
+        except Exception:
+            pass
+        return jsonify({
+            "status": "ok",
+            "version": "2.1.0",
+            "ollama": ollama_ok,
+            "timestamp": datetime.now(timezone.utc).isoformat(),
+        })
+
     return app
diff --git a/tests/test_decision_engine_ext.py b/tests/test_decision_engine_ext.py
new file mode 100644
index 0000000..90a94aa
--- /dev/null
+++ b/tests/test_decision_engine_ext.py
@@ -0,0 +1,493 @@
+"""Tests étendus pour src/quality/decision_engine.py.
+
+Couvre les règles non couvertes dans test_decision_engine.py :
+- Nettoyage hiérarchique (.9 vs spécifique)
+- D50 sans preuve martiale → downgrade D64.9
+- D69.6 thrombopénie vs plaquettes normales
+- Bio rules génériques (YAML-driven)
+- Fonctions internes (_norm, _first_float, _parse_normal_range, etc.)
+- decision_summaries (DOWNGRADE, REMOVE, RULED_OUT, NEED_INFO)
+"""
+
+from __future__ import annotations
+
+from unittest.mock import patch
+
+import pytest
+
+from src.config import (
+    BiologieCle,
+    CodeDecision,
+    Diagnostic,
+    DossierMedical,
+    PreuveClinique,
+    Sejour,
+    Traitement,
+    VetoReport,
+    VetoIssue,
+)
+from src.quality.decision_engine import (
+    _norm,
+    _first_float,
+    _parse_normal_range,
+    _parse_float,
+    _is_sodium_test,
+    _is_potassium_test,
+    _das_promotion_score,
+    _age_band,
+    apply_decisions,
+    decision_summaries,
+)
+
+
+# ── Helpers ────────────────────────────────────────────────────────────
+
+
+def _dossier(
+    dp_code: str | None = None,
+    das: list[Diagnostic] | None = None,
+    bio: list[BiologieCle] | None = None,
+    traitements: list[Traitement] | None = None,
+    age: int | None = None,
+    sexe: str | None = None,
+) -> DossierMedical:
+    dp = None
+    if dp_code:
+        dp = Diagnostic(texte="DP", cim10_suggestion=dp_code)
+    return DossierMedical(
+        sejour=Sejour(age=age, sexe=sexe),
+        diagnostic_principal=dp,
+        diagnostics_associes=das or [],
+        biologie_cle=bio or [],
+        traitements_sortie=traitements or [],
+    )
+
+
+def _diag(
+    texte: str,
+    code: str,
+    confidence: str = "high",
+    source: str = "trackare",
+    status: str | None = None,
+    source_excerpt: str | None = None,
+    preuves: list[PreuveClinique] | None = None,
+) -> Diagnostic:
+    return Diagnostic(
+        texte=texte,
+        cim10_suggestion=code,
+        cim10_confidence=confidence,
+        source=source,
+        status=status,
+        source_excerpt=source_excerpt,
+        preuves_cliniques=preuves or [],
+    )
+
+
+def _bio(test: str, valeur: str | None = None, valeur_num: float | None = None,
+         quality: str | None = None) -> BiologieCle:
+    return BiologieCle(test=test, valeur=valeur, valeur_num=valeur_num, quality=quality)
+
+
+# ===================================================================
+# Fonctions internes
+# ===================================================================
+
+
+class TestNorm:
+    def test_basic_normalization(self):
+        assert _norm("  Hello   World  ") == "hello world"
+
+    def test_accent_removal(self):
+        assert _norm("Hémoglobine élevée") == "hemoglobine elevee"
+
+    def test_apostrophe_normalization(self):
+        assert _norm("l\u2019anémie") == "l'anemie"
+
+
+class TestFirstFloat:
+    @pytest.mark.parametrize("text,expected", [
+        ("CRP 180 mg/L", 180.0),
+        ("Hb 7,5 g/dL", 7.5),
+        ("-3.2 mmol/L", -3.2),
+        ("pas de nombre", None),
+        ("", None),
+    ])
+    def test_extraction(self, text, expected):
+        assert _first_float(text) == expected
+
+
+class TestParseNormalRange:
+    def test_standard_range(self):
+        lo, hi = _parse_normal_range("Plaquettes 250 G/L [N: 150-450]")
+        assert lo == 150.0
+        assert hi == 450.0
+
+    def test_decimal_range(self):
+        lo, hi = _parse_normal_range("K+ 4.2 mmol/L [N: 3,5 - 5,0]")
+        assert lo == 3.5
+        assert hi == 5.0
+
+    def test_no_range(self):
+        lo, hi = _parse_normal_range("CRP 180 mg/L")
+        assert lo is None
+        assert hi is None
+
+
+class TestParseFloat:
+    @pytest.mark.parametrize("value,expected", [
+        ("4.5", 4.5),
+        ("4,5", 4.5),
+        ("  7.2  ", 7.2),
+        ("abc", None),
+        (None, None),
+        ("", None),
+        ("-3.1", -3.1),
+    ])
+    def test_parsing(self, value, expected):
+        assert _parse_float(value) == expected
+
+
+class TestIsSodiumTest:
+    @pytest.mark.parametrize("test,expected", [
+        ("Sodium", True),
+        ("sodium", True),
+        ("Natrémie", True),
+        ("Na+", True),
+        ("Na", True),
+        ("Potassium", False),
+        ("CRP", False),
+    ])
+    def test_detection(self, test, expected):
+        assert _is_sodium_test(test) == expected
+
+
+class TestIsPotassiumTest:
+    @pytest.mark.parametrize("test,expected", [
+        ("Potassium", True),
+        ("potassium", True),
+        ("Kaliémie", True),
+        ("K+", True),
+        ("K", True),
+        ("Sodium", False),
+        ("CRP", False),
+    ])
+    def test_detection(self, test, expected):
+        assert _is_potassium_test(test) == expected
+
+
+class TestAgeBand:
+    def test_adult(self):
+        d = _dossier(age=65)
+        assert _age_band(d, {"age_bands": {"adult_min_years": 18}}) == "adult"
+
+    def test_child(self):
+        d = _dossier(age=5)
+        assert _age_band(d, {"age_bands": {"adult_min_years": 18}}) == "child"
+
+    def test_unknown_age(self):
+        d = _dossier()
+        assert _age_band(d, {}) == "unknown"
+
+    def test_boundary_18(self):
+        d = _dossier(age=18)
+        assert _age_band(d, {"age_bands": {"adult_min_years": 18}}) == "adult"
+
+
+# ===================================================================
+# Nettoyage hiérarchique (.9 vs spécifique)
+# ===================================================================
+
+
+@patch("src.quality.decision_engine.load_reference_ranges", return_value={})
+@patch("src.quality.decision_engine.load_bio_rules", return_value={})
+@patch("src.quality.decision_engine.rule_enabled", return_value=True)
+@patch("src.quality.decision_engine.cim10_validate", return_value=(True, "label"))
+class TestHierarchyCleanup:
+
+    def test_generic_removed_when_specific_exists(self, mock_val, mock_rule, mock_bio, mock_ref):
+        """K81.9 (générique) retiré quand K81.0 (spécifique) est présent."""
+        dp = Diagnostic(texte="Cholécystite", cim10_suggestion="K81.0")
+        das1 = _diag("Cholécystite SAI", "K81.9")
+        das2 = _diag("HTA", "I10")
+        dossier = _dossier(dp_code=None, das=[das1, das2])
+        dossier.diagnostic_principal = dp
+
+        apply_decisions(dossier)
+
+        # K81.9 doit être supprimé
+        k81_9 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "K81.9"][0]
+        assert k81_9.cim10_decision is not None
+        assert k81_9.cim10_decision.action == "REMOVE"
+        assert "RULE-HIERARCHY-CLEANUP" in k81_9.cim10_decision.applied_rules
+        assert k81_9.cim10_final is None
+
+    def test_no_cleanup_when_no_specific(self, mock_val, mock_rule, mock_bio, mock_ref):
+        """K81.9 seul (pas de K81.x spécifique) → conservé."""
+        das1 = _diag("Cholécystite SAI", "K81.9")
+        das2 = _diag("HTA", "I10")
+        dp = Diagnostic(texte="DP", cim10_suggestion="J18.9")
+        dossier = _dossier(das=[das1, das2])
+        dossier.diagnostic_principal = dp
+
+        apply_decisions(dossier)
+
+        k81_9 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "K81.9"][0]
+        # Pas de décision REMOVE
+        assert k81_9.cim10_decision is None or k81_9.cim10_decision.action != "REMOVE"
+
+
+# ===================================================================
+# D50 sans preuve martiale → downgrade D64.9
+# ===================================================================
+
+
+@patch("src.quality.decision_engine.load_reference_ranges", return_value={})
+@patch("src.quality.decision_engine.load_bio_rules", return_value={})
+class TestD50NeedsIron:
+
+    @patch("src.quality.decision_engine.rule_enabled", return_value=True)
+    @patch("src.quality.decision_engine.cim10_validate", return_value=(True, "label"))
+    def test_d50_downgraded_without_iron_evidence(self, mock_val, mock_rule, mock_bio, mock_ref):
+        """D50 sans preuve ferritine/martial → downgrade D64.9."""
+        das = _diag(
+            "Anémie ferriprive",
+            "D50",
+            preuves=[
+                PreuveClinique(type="biologie", element="Hb 9.5 g/dL", interpretation="Anémie confirmée"),
+            ],
+        )
+        dossier = _dossier(dp_code="K85.9", das=[das], age=65)
+
+        apply_decisions(dossier)
+
+        d50 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "D50"][0]
+        assert d50.cim10_final == "D64.9"
+        assert d50.cim10_decision.action == "DOWNGRADE"
+        assert "RULE-D50-NEEDS-IRON" in d50.cim10_decision.applied_rules
+
+    @patch("src.quality.decision_engine.rule_enabled", return_value=True)
+    @patch("src.quality.decision_engine.cim10_validate", return_value=(True, "label"))
+    def test_d50_kept_with_ferritin_evidence(self, mock_val, mock_rule, mock_bio, mock_ref):
+        """D50 avec preuve de ferritine → conservé."""
+        das = _diag(
+            "Anémie ferriprive",
+            "D50",
+            preuves=[
+                PreuveClinique(type="biologie", element="Hb 9.5 g/dL", interpretation="Anémie confirmée"),
+                PreuveClinique(type="biologie", element="Ferritine 8 ng/mL", interpretation="Carence martiale"),
+            ],
+        )
+        dossier = _dossier(dp_code="K85.9", das=[das], age=65)
+
+        apply_decisions(dossier)
+
+        d50 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "D50"][0]
+        assert d50.cim10_decision is None or d50.cim10_decision.action != "DOWNGRADE"
+
+    @patch("src.quality.decision_engine.rule_enabled", return_value=True)
+    @patch("src.quality.decision_engine.cim10_validate", return_value=(True, "label"))
+    def test_d50_kept_with_martial_treatment(self, mock_val, mock_rule, mock_bio, mock_ref):
+        """D50 avec traitement martial → conservé (preuve par le traitement)."""
+        das = _diag(
+            "Anémie ferriprive",
+            "D50",
+            preuves=[
+                PreuveClinique(type="biologie", element="Hb 8.0 g/dL", interpretation="Anémie confirmée"),
+            ],
+        )
+        dossier = _dossier(
+            dp_code="K85.9",
+            das=[das],
+            age=65,
+            traitements=[Traitement(medicament="Fer intraveineux", posologie="200mg")],
+        )
+
+        apply_decisions(dossier)
+
+        d50 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "D50"][0]
+        assert d50.cim10_decision is None or d50.cim10_decision.action != "DOWNGRADE"
+
+    def test_d50_rule_disabled(self, mock_bio, mock_ref):
+        """RULE-D50-NEEDS-IRON désactivée → D50 conservé."""
+        def _rule_selective(rule_id):
+            return rule_id != "RULE-D50-NEEDS-IRON"
+
+        with patch("src.quality.decision_engine.rule_enabled", side_effect=_rule_selective):
+            with patch("src.quality.decision_engine.cim10_validate", return_value=(True, "label")):
+                das = _diag(
+                    "Anémie ferriprive",
+                    "D50",
+                    preuves=[
+                        PreuveClinique(type="biologie", element="Hb 9.5", interpretation="Anémie confirmée"),
+                    ],
+                )
+                dossier = _dossier(dp_code="K85.9", das=[das], age=65)
+                apply_decisions(dossier)
+                d50 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "D50"][0]
+                assert d50.cim10_decision is None or d50.cim10_decision.action != "DOWNGRADE"
+
+
+# ===================================================================
+# D69.6 thrombopénie vs plaquettes normales
+# ===================================================================
+
+
+@patch("src.quality.decision_engine.load_bio_rules", return_value={})
+class TestD696PltNormal:
+
+    @patch("src.quality.decision_engine.load_reference_ranges", return_value={
+        "age_bands": {"adult_min_years": 18},
+        "fallback_ranges": {"adult": {"platelets": {"low": 150, "high": 450}}},
+        "safe_zones_unknown_age": {"platelets_ruled_out_low": 170},
+    })
+    @patch("src.quality.decision_engine.rule_enabled", return_value=True)
+    @patch("src.quality.decision_engine.cim10_validate", return_value=(True, "label"))
+    def test_d696_ruled_out_with_normal_platelets(self, mock_val, mock_rule, mock_ref, mock_bio):
+        """D69.6 avec plaquettes normales → RULED_OUT."""
+        das = _diag("Thrombopénie", "D69.6")
+        bio = [_bio("Plaquettes", "250 G/L [N: 150-450]", valeur_num=250.0)]
+        dossier = _dossier(dp_code="K85.9", das=[das], bio=bio, age=65)
+
+        apply_decisions(dossier)
+
+        d696 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "D69.6"][0]
+        assert d696.cim10_decision is not None
+        assert d696.cim10_decision.action == "RULED_OUT"
+        assert d696.status == "ruled_out"
+        assert d696.cim10_final is None
+
+    @patch("src.quality.decision_engine.load_reference_ranges", return_value={
+        "age_bands": {"adult_min_years": 18},
+        "fallback_ranges": {"adult": {"platelets": {"low": 150, "high": 450}}},
+        "safe_zones_unknown_age": {"platelets_ruled_out_low": 170},
+    })
+    @patch("src.quality.decision_engine.rule_enabled", return_value=True)
+    @patch("src.quality.decision_engine.cim10_validate", return_value=(True, "label"))
+    def test_d696_kept_with_low_platelets(self, mock_val, mock_rule, mock_ref, mock_bio):
+        """D69.6 avec plaquettes basses → conservé."""
+        das = _diag("Thrombopénie", "D69.6")
+        bio = [_bio("Plaquettes", "80 G/L", valeur_num=80.0)]
+        dossier = _dossier(dp_code="K85.9", das=[das], bio=bio, age=65)
+
+        apply_decisions(dossier)
+
+        d696 = [d for d in dossier.diagnostics_associes if d.cim10_suggestion == "D69.6"][0]
+        assert d696.cim10_decision is None or d696.cim10_decision.action != "RULED_OUT"
+
+
+# ===================================================================
+# _das_promotion_score — cas supplémentaires
+# ===================================================================
+
+
+class TestDasPromotionScoreExtra:
+    def test_empty_code(self):
+        """Un diagnostic sans code final a un score minimal."""
+        diag = Diagnostic(texte="Test", cim10_final="")
+        score = _das_promotion_score(diag)
+        # pas de lettre → pertinence=2, mais specificite=0
+        assert score[2] == 0
+
+    def test_none_confidence(self):
+        """Confiance None → score minimal."""
+        diag = Diagnostic(texte="Test", cim10_final="K85.9", cim10_confidence=None)
+        score = _das_promotion_score(diag)
+        assert score[1] == 1  # default low
+
+
+# ===================================================================
+# decision_summaries — toutes les actions
+# ===================================================================
+
+
+class TestDecisionSummaries:
+
+    def test_keep_no_output(self):
+        """KEEP ne produit pas de résumé."""
+        das = Diagnostic(texte="HTA", cim10_suggestion="I10", cim10_decision=CodeDecision(action="KEEP"))
+        dossier = DossierMedical(diagnostics_associes=[das])
+        lines = decision_summaries(dossier)
+        assert len(lines) == 0
+
+    def test_downgrade_summary(self):
+        das = Diagnostic(
+            texte="Anémie",
+            cim10_suggestion="D50",
+            cim10_decision=CodeDecision(
+                action="DOWNGRADE",
+                final_code="D64.9",
+                downgraded_from="D50",
+                applied_rules=["RULE-D50-NEEDS-IRON"],
+                needs_info=["Bilan martial disponible ?"],
+            ),
+        )
+        dossier = DossierMedical(diagnostics_associes=[das])
+        lines = decision_summaries(dossier)
+        assert any("D50" in l and "D64.9" in l for l in lines)
+        assert any("besoin_info" in l for l in lines)
+
+    def test_remove_summary(self):
+        das = Diagnostic(
+            texte="Cholécystite SAI",
+            cim10_suggestion="K81.9",
+            cim10_decision=CodeDecision(
+                action="REMOVE",
+                applied_rules=["RULE-HIERARCHY-CLEANUP"],
+            ),
+        )
+        dossier = DossierMedical(diagnostics_associes=[das])
+        lines = decision_summaries(dossier)
+        assert any("K81.9" in l and "supprimé" in l for l in lines)
+
+    def test_ruled_out_summary(self):
+        das = Diagnostic(
+            texte="Thrombopénie",
+            cim10_suggestion="D69.6",
+            cim10_decision=CodeDecision(
+                action="RULED_OUT",
+                applied_rules=["RULE-D69.6-PLT-NORMAL"],
+                reason="Plaquettes normales",
+            ),
+        )
+        dossier = DossierMedical(diagnostics_associes=[das])
+        lines = decision_summaries(dossier)
+        assert any("D69.6" in l and "écarté" in l for l in lines)
+        assert any("raison" in l for l in lines)
+
+    def test_need_info_summary(self):
+        das = Diagnostic(
+            texte="Hyponatrémie",
+            cim10_suggestion="E87.1",
+            cim10_decision=CodeDecision(
+                action="NEED_INFO",
+                applied_rules=["RULE-HYPONATREMIA-MISSING"],
+                reason="Sodium non extrait",
+                needs_info=["Valeurs de sodium ?"],
+            ),
+        )
+        dossier = DossierMedical(diagnostics_associes=[das])
+        lines = decision_summaries(dossier)
+        assert any("NEED_INFO" in l for l in lines)
+        assert any("besoin_info" in l for l in lines)
+
+    def test_no_diagnostics(self):
+        """Dossier sans diagnostics → liste vide."""
+        dossier = DossierMedical()
+        lines = decision_summaries(dossier)
+        assert lines == []
+
+    def test_dp_decision_included(self):
+        """Décision sur le DP est incluse dans le résumé."""
+        dp = Diagnostic(
+            texte="DP",
+            cim10_suggestion="D50",
+            cim10_decision=CodeDecision(
+                action="DOWNGRADE",
+                final_code="D64.9",
+                downgraded_from="D50",
+                applied_rules=["RULE-D50-NEEDS-IRON"],
+            ),
+        )
+        dossier = DossierMedical(diagnostic_principal=dp)
+        lines = decision_summaries(dossier)
+        assert any("diagnostic_principal" in l for l in lines)
diff --git a/tests/test_dp_finalizer_ext.py b/tests/test_dp_finalizer_ext.py
new file mode 100644
index 0000000..ae13ead
--- /dev/null
+++ b/tests/test_dp_finalizer_ext.py
@@ -0,0 +1,340 @@
+"""Tests étendus pour src/medical/dp_finalizer.py.
+
+Couvre les cas non couverts dans test_dp_finalizer.py :
+- R6 — Z-code non whitelisté remplacé par DAS
+- Fonctions internes (_family3, _code_in_candidates, _has_strong_evidence)
+- finalize_dp avec source Trackare détectée par evidence/reason
+- Cas limites : codes None, candidats vides, Z whitelistés en R6
+"""
+
+from __future__ import annotations
+
+import pytest
+
+from src.config import DossierMedical, DPSelection, DPCandidate, Diagnostic, Sejour
+from src.medical.dp_finalizer import (
+    finalize_dp,
+    decide_dp_final,
+    _family3,
+    _code_in_candidates,
+    _has_strong_evidence,
+    _apply_r5,
+    _make_selection,
+)
+
+
+# ── Helpers ────────────────────────────────────────────────────────────
+
+
+def _sel(
+    code: str | None = None,
+    term: str = "",
+    verdict: str = "REVIEW",
+    confidence: str = "medium",
+    evidence: list[str] | None = None,
+    reason: str = "",
+    candidates: list[DPCandidate] | None = None,
+) -> DPSelection:
+    return DPSelection(
+        chosen_code=code,
+        chosen_term=term or (code or ""),
+        verdict=verdict,
+        confidence=confidence,
+        evidence=evidence or [],
+        reason=reason,
+        candidates=candidates or [],
+    )
+
+
+def _cand(code: str, term: str = "", score: float = 0.0) -> DPCandidate:
+    return DPCandidate(index=0, term=term or code, code=code, score=score)
+
+
+def _dossier(
+    doc_type: str = "crh",
+    dp_selection: DPSelection | None = None,
+    das: list[Diagnostic] | None = None,
+) -> DossierMedical:
+    return DossierMedical(
+        document_type=doc_type,
+        sejour=Sejour(),
+        dp_selection=dp_selection,
+        diagnostics_associes=das or [],
+    )
+
+
+# ===================================================================
+# Fonctions internes
+# ===================================================================
+
+
+class TestFamily3:
+    @pytest.mark.parametrize("code,expected", [
+        ("K81.0", "K81"),
+        ("K81", "K81"),
+        ("I26.9", "I26"),
+        ("Z51.1", "Z51"),
+        (None, ""),
+        ("", ""),
+    ])
+    def test_extraction(self, code, expected):
+        assert _family3(code) == expected
+
+
+class TestCodeInCandidates:
+    def test_exact_match(self):
+        sel = _sel(candidates=[_cand("I26.9"), _cand("K85.1")])
+        assert _code_in_candidates("I26.9", sel) is True
+
+    def test_family3_match(self):
+        sel = _sel(candidates=[_cand("I26.9")])
+        assert _code_in_candidates("I26.0", sel) is True  # same family
+
+    def test_no_match(self):
+        sel = _sel(candidates=[_cand("I26.9")])
+        assert _code_in_candidates("K85.1", sel) is False
+
+    def test_none_code(self):
+        sel = _sel(candidates=[_cand("I26.9")])
+        assert _code_in_candidates(None, sel) is False
+
+    def test_empty_candidates(self):
+        sel = _sel(candidates=[])
+        assert _code_in_candidates("I26.9", sel) is False
+
+
+class TestHasStrongEvidence:
+    def test_no_evidence(self):
+        sel = _sel(evidence=[])
+        assert _has_strong_evidence(sel) is False
+
+    def test_only_trackare_source(self):
+        """'Source: Trackare' seul n'est pas une preuve forte."""
+        sel = _sel(evidence=["Source: Trackare (codage établissement)"])
+        assert _has_strong_evidence(sel) is False
+
+    def test_strong_evidence(self):
+        sel = _sel(evidence=["Diagnostic de sortie: «Embolie pulmonaire»"])
+        assert _has_strong_evidence(sel) is True
+
+    def test_mixed_evidence(self):
+        sel = _sel(evidence=[
+            "Source: Trackare (codage établissement)",
+            "Conclusion CRH: «Pancréatite biliaire»",
+        ])
+        assert _has_strong_evidence(sel) is True
+
+
+class TestMakeSelection:
+    def test_preserves_source_candidates(self):
+        source = _sel(candidates=[_cand("I26.9", "EP", 5.0)])
+        result = _make_selection(
+            code="I26.9", term="EP", verdict="CONFIRMED",
+            confidence="high", evidence=[], reason="test",
+            source_sel=source,
+        )
+        assert len(result.candidates) == 1
+        assert result.candidates[0].code == "I26.9"
+
+    def test_no_source(self):
+        result = _make_selection(
+            code="I26.9", term="EP", verdict="CONFIRMED",
+            confidence="high", evidence=[], reason="test",
+        )
+        assert result.candidates == []
+        assert result.debug_scores is None
+
+
+# ===================================================================
+# R5 — _apply_r5 (direct call)
+# ===================================================================
+
+
+class TestApplyR5Direct:
+    def test_z_code_non_whitelisted_forced_review(self):
+        dp = _sel(code="Z95.5", verdict="CONFIRMED", confidence="high", evidence=[])
+        dp, flags, alertes = _apply_r5(dp, None, allow_symptom_dp=False)
+        assert dp.verdict == "REVIEW"
+        assert flags.get("z_code_dp_review") is True
+
+    def test_z_code_whitelisted_untouched(self):
+        dp = _sel(code="Z51.1", verdict="CONFIRMED", confidence="high", evidence=[])
+        dp, flags, alertes = _apply_r5(dp, None, allow_symptom_dp=False)
+        assert dp.verdict == "CONFIRMED"
+        assert "z_code_dp_review" not in flags
+
+    def test_r_code_with_non_r_candidate(self):
+        crh = _sel(candidates=[_cand("R06.0"), _cand("J18.9")])
+        dp = _sel(code="R06.0", verdict="CONFIRMED", confidence="high", evidence=[])
+        dp, flags, alertes = _apply_r5(dp, crh, allow_symptom_dp=False)
+        assert dp.verdict == "REVIEW"
+        assert flags.get("r_code_dp_with_non_r_candidate") is True
+
+    def test_r_code_allowed_by_flag(self):
+        crh = _sel(candidates=[_cand("R06.0"), _cand("J18.9")])
+        dp = _sel(code="R06.0", verdict="CONFIRMED", confidence="high", evidence=[])
+        dp, flags, alertes = _apply_r5(dp, crh, allow_symptom_dp=True)
+        assert dp.verdict == "CONFIRMED"
+
+    def test_review_dp_not_changed(self):
+        """R5 ne touche que les CONFIRMED."""
+        dp = _sel(code="Z95.5", verdict="REVIEW", confidence="medium", evidence=[])
+        dp, flags, alertes = _apply_r5(dp, None, allow_symptom_dp=False)
+        assert dp.verdict == "REVIEW"
+        assert "z_code_dp_review" not in flags  # pas de flag car déjà REVIEW
+
+
+# ===================================================================
+# R6 — Z-code remplacé par DAS (via finalize_dp)
+# ===================================================================
+
+
+class TestR6ZcodeReplacedByDas:
+    def test_z_code_replaced_by_das(self):
+        """Z-code non whitelisté en DP → DAS high/medium non-R non-Z promu."""
+        sel = _sel(
+            code="Z96.6",  # non whitelisté
+            verdict="CONFIRMED",
+            evidence=["Source: Trackare (codage établissement)"],
+            reason="DP Trackare",
+        )
+        das = [
+            Diagnostic(texte="Pancréatite", cim10_suggestion="K85.9", cim10_confidence="high"),
+        ]
+        d = _dossier(doc_type="trackare", dp_selection=sel, das=das)
+        finalize_dp(d)
+
+        assert d.dp_final is not None
+        assert d.dp_final.chosen_code == "K85.9"
+        assert d.quality_flags.get("z_code_replaced_by_das") is True
+        assert any("K85.9" in a for a in d.alertes_codage)
+
+    def test_z_code_whitelisted_not_replaced(self):
+        """Z-code whitelisté en DP → pas de remplacement."""
+        sel = _sel(
+            code="Z51.1",  # whitelisté
+            verdict="CONFIRMED",
+            evidence=["Source: Trackare (codage établissement)"],
+            reason="DP Trackare",
+        )
+        das = [
+            Diagnostic(texte="Pancréatite", cim10_suggestion="K85.9", cim10_confidence="high"),
+        ]
+        d = _dossier(doc_type="trackare", dp_selection=sel, das=das)
+        finalize_dp(d)
+
+        assert d.dp_final.chosen_code == "Z51.1"
+        assert "z_code_replaced_by_das" not in d.quality_flags
+
+    def test_z_code_no_eligible_das(self):
+        """Z-code non whitelisté mais aucun DAS éligible → pas de remplacement."""
+        sel = _sel(
+            code="Z96.6",
+            verdict="CONFIRMED",
+            evidence=["Source: Trackare (codage établissement)"],
+            reason="DP Trackare",
+        )
+        das = [
+            Diagnostic(texte="Symptôme R", cim10_suggestion="R10.4", cim10_confidence="high"),
+            Diagnostic(texte="Z-code", cim10_suggestion="Z87.1", cim10_confidence="low"),
+        ]
+        d = _dossier(doc_type="trackare", dp_selection=sel, das=das)
+        finalize_dp(d)
+
+        # R10.4 est R-code, Z87.1 est Z-code → aucun éligible
+        assert "z_code_replaced_by_das" not in d.quality_flags
+
+
+# ===================================================================
+# finalize_dp — détection source Trackare
+# ===================================================================
+
+
+class TestFinalizeDpSourceDetection:
+    def test_trackare_detected_by_document_type(self):
+        sel = _sel(code="K81.0", verdict="CONFIRMED", evidence=[], reason="CRH DP")
+        d = _dossier(doc_type="trackare", dp_selection=sel)
+        finalize_dp(d)
+        assert d.dp_trackare is not None
+
+    def test_trackare_detected_by_reason(self):
+        sel = _sel(code="K81.0", verdict="CONFIRMED", evidence=[],
+                   reason="DP Trackare — source d'autorité")
+        d = _dossier(doc_type="crh", dp_selection=sel)
+        finalize_dp(d)
+        assert d.dp_trackare is not None
+
+    def test_trackare_detected_by_evidence(self):
+        sel = _sel(code="K81.0", verdict="CONFIRMED",
+                   evidence=["Source: Trackare (codage établissement)"],
+                   reason="DP")
+        d = _dossier(doc_type="crh", dp_selection=sel)
+        finalize_dp(d)
+        assert d.dp_trackare is not None
+
+    def test_crh_only_when_no_trackare_signal(self):
+        sel = _sel(code="K81.0", verdict="CONFIRMED",
+                   evidence=["Diagnostic de sortie CRH"],
+                   reason="CRH CONFIRMED")
+        d = _dossier(doc_type="crh", dp_selection=sel)
+        finalize_dp(d)
+        assert d.dp_trackare is None
+        assert d.dp_crh_only is not None
+
+
+# ===================================================================
+# finalize_dp — pas de dp_selection
+# ===================================================================
+
+
+class TestFinalizeDpNoSelection:
+    def test_no_dp_selection(self):
+        """Dossier sans dp_selection → REVIEW, flag no_dp_source."""
+        d = _dossier(doc_type="crh")
+        finalize_dp(d)
+        assert d.dp_final is not None
+        assert d.dp_final.verdict == "REVIEW"
+        assert d.quality_flags.get("no_dp_source") is True
+
+
+# ===================================================================
+# decide_dp_final — edge cases
+# ===================================================================
+
+
+class TestDecideDpFinalEdgeCases:
+    def test_trackare_z_code_whitelisted(self):
+        """Z03 (whitelisté) en Trackare-only → CONFIRMED."""
+        trackare = _sel(code="Z03.8", verdict="CONFIRMED",
+                        evidence=["Source: Trackare"])
+        dp, flags, _ = decide_dp_final(trackare, None)
+        # Z03 is whitelisted
+        assert dp.verdict == "CONFIRMED"
+
+    def test_trackare_empty_evidence_gets_default(self):
+        """Trackare-only sans evidence → evidence par défaut ajoutée."""
+        trackare = _sel(code="K81.0", verdict="CONFIRMED", evidence=[])
+        dp, flags, _ = decide_dp_final(trackare, None)
+        assert any("Trackare" in e for e in dp.evidence)
+
+    def test_both_sources_trackare_review_from_selector(self):
+        """Si dp_selector a déjà mis REVIEW, le finalizer le respecte."""
+        trackare = _sel(code="K81.0", verdict="REVIEW",
+                        evidence=["Source: Trackare"])
+        crh = _sel(code="K81.0", verdict="REVIEW",
+                   evidence=["Diagnostic CRH"],
+                   candidates=[_cand("K81.0")])
+        dp, flags, _ = decide_dp_final(trackare, crh)
+        # R2 corrobore → CONFIRMED
+        assert dp.verdict == "CONFIRMED"
+        assert flags.get("trackare_confirmed_by_crh") is True
+
+    def test_code_none_handling(self):
+        """Trackare avec code None → pas de crash."""
+        trackare = _sel(code=None, verdict="REVIEW", evidence=["Source: Trackare"])
+        crh = _sel(code="K81.0", verdict="CONFIRMED",
+                   evidence=["Diagnostic de sortie CRH"])
+        dp, flags, _ = decide_dp_final(trackare, crh)
+        # Code None ne commence pas par R, ne commence pas par Z
+        # Il ne matche pas CRH → R4 ambigu
+        assert dp is not None
diff --git a/tests/test_health_endpoint.py b/tests/test_health_endpoint.py
new file mode 100644
index 0000000..3eb46df
--- /dev/null
+++ b/tests/test_health_endpoint.py
@@ -0,0 +1,79 @@
+"""Tests pour le endpoint /health du viewer Flask."""
+
+import json
+import os
+from unittest.mock import patch, MagicMock
+
+import pytest
+import requests
+
+from src.viewer.app import create_app
+
+
+@pytest.fixture
+def app():
+    with patch.dict(os.environ, {"T2A_DEMO_USER": "", "T2A_DEMO_PASS": ""}):
+        app = create_app()
+        app.config["TESTING"] = True
+        yield app
+
+
+@pytest.fixture
+def client(app):
+    return app.test_client()
+
+
+class TestHealthEndpoint:
+
+    def test_health_returns_200(self, client):
+        response = client.get("/health")
+        assert response.status_code == 200
+
+    def test_health_returns_json(self, client):
+        response = client.get("/health")
+        data = response.get_json()
+        assert data is not None
+        assert data["status"] == "ok"
+
+    def test_health_contains_required_fields(self, client):
+        response = client.get("/health")
+        data = response.get_json()
+        assert "status" in data
+        assert "version" in data
+        assert "ollama" in data
+        assert "timestamp" in data
+
+    def test_health_version_format(self, client):
+        response = client.get("/health")
+        data = response.get_json()
+        assert data["version"] == "2.1.0"
+
+    def test_health_timestamp_is_iso(self, client):
+        from datetime import datetime
+        response = client.get("/health")
+        data = response.get_json()
+        # Should parse without error
+        ts = datetime.fromisoformat(data["timestamp"])
+        assert ts is not None
+
+    @patch("src.viewer.app.requests.get")
+    def test_health_ollama_reachable(self, mock_get, client):
+        """Quand Ollama répond, ollama=True."""
+        mock_resp = MagicMock()
+        mock_resp.status_code = 200
+        mock_get.return_value = mock_resp
+        response = client.get("/health")
+        data = response.get_json()
+        assert data["ollama"] is True
+
+    @patch("src.viewer.app.requests.get", side_effect=requests.ConnectionError("refused"))
+    def test_health_ollama_unreachable(self, mock_get, client):
+        """Quand Ollama est injoignable, ollama=False."""
+        response = client.get("/health")
+        data = response.get_json()
+        assert data["ollama"] is False
+
+    def test_health_ollama_is_bool(self, client):
+        response = client.get("/health")
+        data = response.get_json()
+        assert isinstance(data["ollama"], bool)
diff --git a/tests/test_viewer_helpers.py b/tests/test_viewer_helpers.py
new file mode 100644
index 0000000..4e3d1a2
--- /dev/null
+++ b/tests/test_viewer_helpers.py
@@ -0,0 +1,526 @@
+"""Tests unitaires pour src/viewer/helpers.py.
+
+Couvre : filtres Jinja2, helpers de statistiques, fonctions utilitaires.
+Pas de dépendance à Ollama ni au système de fichiers (sauf tmpdir).
+"""
+
+from __future__ import annotations
+
+import re
+from collections import Counter
+from unittest.mock import patch, MagicMock
+
+import pytest
+from markupsafe import Markup
+
+from src.config import (
+    DossierMedical,
+    Diagnostic,
+    DPSelection,
+    DPCandidate,
+    ActeCCAM,
+    CodeDecision,
+    GHMEstimation,
+    VetoReport,
+    VetoIssue,
+    CompletudeDossier,
+    ControleCPAM,
+    FinancialImpact,
+    Sejour,
+)
+from src.viewer.helpers import (
+    compute_group_stats,
+    compute_dashboard_stats,
+    compute_dim_synthesis,
+    confidence_badge,
+    confidence_label,
+    severity_badge,
+    cma_level_badge,
+    format_duration,
+    format_dossier_name,
+    format_doc_name,
+    decision_badge,
+    format_cpam_text,
+    human_where,
+    _date_to_iso,
+    _sort_qc_alerts,
+    _compute_jours_restants,
+)
+
+
+# ── Helpers ────────────────────────────────────────────────────────────
+
+
+def _item(dossier: DossierMedical, path_rel: str = "test/file.json") -> dict:
+    """Crée un item de scan_dossiers minimal."""
+    return {"name": "file", "path_rel": path_rel, "dossier": dossier}
+
+
+def _diag(
+    texte: str = "Diagnostic",
+    code: str | None = None,
+    confidence: str | None = None,
+    est_cma: bool | None = None,
+    decision: CodeDecision | None = None,
+    cim10_final: str | None = None,
+) -> Diagnostic:
+    return Diagnostic(
+        texte=texte,
+        cim10_suggestion=code,
+        cim10_confidence=confidence,
+        est_cma=est_cma,
+        cim10_decision=decision,
+        cim10_final=cim10_final,
+    )
+
+
+# ===================================================================
+# confidence_badge / confidence_label
+# ===================================================================
+
+
+class TestConfidenceBadge:
+    def test_high(self):
+        result = confidence_badge("high")
+        assert isinstance(result, Markup)
+        assert "Haute" in result
+        assert "#16a34a" in result
+
+    def test_medium(self):
+        result = confidence_badge("medium")
+        assert "Moyenne" in result
+
+    def test_low(self):
+        result = confidence_badge("low")
+        assert "Basse" in result
+        assert "#dc2626" in result
+
+    def test_none_returns_empty(self):
+        assert confidence_badge(None) == ""
+
+    def test_empty_string_returns_empty(self):
+        assert confidence_badge("") == ""
+
+    def test_unknown_value_uses_default_colors(self):
+        result = confidence_badge("unknown")
+        assert "unknown" in result
+        assert "#6b7280" in result  # default foreground
+
+
+class TestConfidenceLabel:
+    @pytest.mark.parametrize("value,expected", [
+        ("high", "Haute"),
+        ("medium", "Moyenne"),
+        ("low", "Basse"),
+        (None, ""),
+        ("", ""),
+        ("autre", "autre"),
+    ])
+    def test_labels(self, value, expected):
+        assert confidence_label(value) == expected
+
+
+# ===================================================================
+# cma_level_badge
+# ===================================================================
+
+
+class TestCmaLevelBadge:
+    def test_none_returns_empty(self):
+        assert cma_level_badge(None) == ""
+
+    def test_zero_returns_empty(self):
+        assert cma_level_badge(0) == ""
+
+    def test_negative_returns_empty(self):
+        assert cma_level_badge(-1) == ""
+
+    @pytest.mark.parametrize("level", [1, 2, 3, 4])
+    def test_valid_levels(self, level):
+        result = cma_level_badge(level)
+        assert isinstance(result, Markup)
+        assert f"CMA {level}" in result
+
+    def test_level_above_4_capped(self):
+        """Un niveau > 4 est cappé à 4."""
+        result = cma_level_badge(5)
+        assert "CMA 4" in result
+
+
+# ===================================================================
+# format_dossier_name / format_doc_name
+# ===================================================================
+
+
+class TestFormatDossierName:
+    def test_racine(self):
+        assert format_dossier_name("racine") == "Non classés"
+
+    def test_normal_name(self):
+        assert format_dossier_name("190_23139234") == "190_23139234"
+
+
+class TestFormatDocName:
+    @pytest.mark.parametrize("name,expected", [
+        ("190_fusionne_cim10", "Fusionné"),
+        ("CRH_23139234_cim10", "CRH"),
+        ("CRO_23139234_cim10", "CRO"),
+        ("crh_23139234", "CRH"),
+        ("trackare-01295620", "Trackare"),
+        ("ANAPATH_23103383", "Anapath"),
+        ("some_other_doc", "some_other_doc"),
+    ])
+    def test_doc_names(self, name, expected):
+        assert format_doc_name(name) == expected
+
+
+# ===================================================================
+# decision_badge
+# ===================================================================
+
+
+class TestDecisionBadge:
+    def test_none_returns_empty(self):
+        assert decision_badge(None) == ""
+
+    def test_keep_returns_empty(self):
+        """KEEP n'a pas de badge (cas par défaut)."""
+        dec = CodeDecision(action="KEEP")
+        assert decision_badge(dec) == ""
+
+    @pytest.mark.parametrize("action,label", [
+        ("DOWNGRADE", "Rétrogradé"),
+        ("REMOVE", "Supprimé"),
+        ("RULED_OUT", "Écarté"),
+        ("NEED_INFO", "Preuve manquante"),
+        ("PROMOTE_DP", "Promu en DP"),
+    ])
+    def test_action_labels(self, action, label):
+        dec = CodeDecision(action=action)
+        result = decision_badge(dec)
+        assert label in result
+
+    def test_dict_input(self):
+        """Accepte un dict en plus d'un CodeDecision."""
+        result = decision_badge({"action": "REMOVE"})
+        assert "Supprimé" in result
+
+    def test_dict_keep(self):
+        result = decision_badge({"action": "KEEP"})
+        assert result == ""
+
+    def test_unknown_action(self):
+        """Action inconnue affiche le texte brut."""
+        result = decision_badge({"action": "CUSTOM_ACTION"})
+        assert "CUSTOM_ACTION" in result
+
+
+# ===================================================================
+# human_where
+# ===================================================================
+
+
+class TestHumanWhere:
+    @pytest.mark.parametrize("value,expected", [
+        (None, "Global"),
+        ("", "Global"),
+        ("diagnostic_principal", "Diagnostic Principal"),
+        ("diagnostics_associes", "Diagnostics Associés"),
+        ("sejour", "Séjour"),
+        ("diagnostics_associes[0]", "DAS n°1"),
+        ("diagnostics_associes[5]", "DAS n°6"),
+        ("actes_ccam[0]", "Acte n°1"),
+        ("actes_ccam[2]", "Acte n°3"),
+        ("autre_chose", "autre_chose"),
+    ])
+    def test_conversions(self, value, expected):
+        assert human_where(value) == expected
+
+
+# ===================================================================
+# _date_to_iso
+# ===================================================================
+
+
+class TestDateToIso:
+    def test_valid_date(self):
+        assert _date_to_iso("15/03/2025") == "2025-03-15"
+
+    def test_invalid_format(self):
+        assert _date_to_iso("2025-03-15") == ""
+
+    def test_empty_string(self):
+        assert _date_to_iso("") == ""
+
+    def test_single_part(self):
+        assert _date_to_iso("15") == ""
+
+
+# ===================================================================
+# _sort_qc_alerts
+# ===================================================================
+
+
+class TestSortQcAlerts:
+    def test_critical_first(self):
+        alerts = [
+            "Recommandation : ajouter un DAS",
+            "Erreur critique : code invalide",
+            "Code justifié solidement",
+        ]
+        sorted_alerts = _sort_qc_alerts(alerts)
+        assert sorted_alerts[0] == "Erreur critique : code invalide"
+        assert sorted_alerts[-1] == "Code justifié solidement"
+
+    def test_empty_list(self):
+        assert _sort_qc_alerts([]) == []
+
+    def test_dp_prioritized_within_tier(self):
+        """Les alertes mentionnant le DP sont priorisées dans leur tier."""
+        alerts = [
+            "Redondance dans DAS",
+            "Redondance DP diagnostic principal suspect",
+        ]
+        sorted_alerts = _sort_qc_alerts(alerts)
+        assert "DP" in sorted_alerts[0] or "diagnostic principal" in sorted_alerts[0]
+
+
+# ===================================================================
+# _compute_jours_restants
+# ===================================================================
+
+
+class TestComputeJoursRestants:
+    def test_no_date(self):
+        ctrl = MagicMock()
+        ctrl.date_limite_reponse = None
+        assert _compute_jours_restants(ctrl) is None
+
+    def test_invalid_date(self):
+        ctrl = MagicMock()
+        ctrl.date_limite_reponse = "not-a-date"
+        assert _compute_jours_restants(ctrl) is None
+
+    def test_valid_date_returns_int(self):
+        ctrl = MagicMock()
+        ctrl.date_limite_reponse = "15/03/2030"
+        result = _compute_jours_restants(ctrl)
+        assert isinstance(result, int)
+        assert result > 0  # well into the future
+
+
+# ===================================================================
+# compute_group_stats
+# ===================================================================
+
+
+class TestComputeGroupStats:
+    def test_empty_items(self):
+        stats = compute_group_stats([])
+        assert stats == {"das_count": 0, "alertes_count": 0, "actes_count": 0, "cma_count": 0}
+
+    def test_cma_counted_on_dp(self):
+        """CMA sur le DP est compté."""
+        d = DossierMedical(
+            diagnostic_principal=_diag("DP", "K85.9", est_cma=True),
+        )
+        stats = compute_group_stats([_item(d)])
+        assert stats["cma_count"] == 1
+
+    def test_multiple_items(self):
+        d1 = DossierMedical(
+            diagnostics_associes=[_diag("DAS1", "I10", est_cma=True)],
+            actes_ccam=[ActeCCAM(texte="Acte1")],
+            alertes_codage=["Alerte"],
+        )
+        d2 = DossierMedical(
+            diagnostics_associes=[_diag("DAS2", "E11.9"), _diag("DAS3", "J18.9", est_cma=True)],
+        )
+        stats = compute_group_stats([_item(d1), _item(d2)])
+        assert stats["das_count"] == 3
+        assert stats["actes_count"] == 1
+        assert stats["alertes_count"] == 1
+        assert stats["cma_count"] == 2
+
+
+# ===================================================================
+# compute_dashboard_stats
+# ===================================================================
+
+
+class TestComputeDashboardStats:
+    def test_empty_groups(self):
+        stats = compute_dashboard_stats({})
+        assert stats["total_dossiers"] == 0
+        assert stats["total_fichiers"] == 0
+        assert stats["top_codes"] == []
+        assert stats["processing_time_avg"] == 0
+
+    def test_single_dossier(self):
+        d = DossierMedical(
+            diagnostic_principal=_diag("DP", "K85.9", confidence="high"),
+            diagnostics_associes=[_diag("DAS", code="I10")],
+            actes_ccam=[ActeCCAM(texte="Acte")],
+            alertes_codage=["Alerte"],
+            processing_time_s=10.5,
+        )
+        groups = {"grp1": [_item(d)]}
+        stats = compute_dashboard_stats(groups)
+        assert stats["total_dossiers"] == 1
+        assert stats["total_fichiers"] == 1
+        assert stats["total_das"] == 1
+        assert stats["total_actes"] == 1
+        assert stats["total_alertes"] == 1
+        assert stats["processing_time_avg"] == 10.5
+
+    def test_dp_validity_absent_when_no_dp(self):
+        d = DossierMedical()
+        groups = {"grp": [_item(d)]}
+        stats = compute_dashboard_stats(groups)
+        assert stats["dp_validity"].get("absent", 0) == 1
+
+    def test_ghm_types_counted(self):
+        d = DossierMedical(
+            ghm_estimation=GHMEstimation(type_ghm="C", severite=3),
+        )
+        groups = {"grp": [_item(d)]}
+        stats = compute_dashboard_stats(groups)
+        assert stats["ghm_types"].get("C") == 1
+        assert stats["severity_dist"].get(3) == 1
+
+
+# ===================================================================
+# compute_dim_synthesis
+# ===================================================================
+
+
+class TestComputeDimSynthesis:
+    def test_empty_groups(self):
+        result = compute_dim_synthesis({})
+        assert result["dp"]["total"] == 0
+        assert result["das"]["total"] == 0
+        assert result["veto"]["avg_score"] == 0
+
+    def test_dp_confirmed(self):
+        d = DossierMedical(
+            dp_final=DPSelection(
+                chosen_code="K85.9",
+                verdict="CONFIRMED",
+                confidence="high",
+                evidence=["Test evidence"],
+            ),
+        )
+        groups = {"grp": [_item(d)]}
+        result = compute_dim_synthesis(groups)
+        assert result["dp"]["total"] == 1
+        assert result["dp"]["confirmed"] == 1
+
+    def test_dp_review_creates_alert(self):
+        d = DossierMedical(
+            dp_final=DPSelection(
+                chosen_code="R10.4",
+                verdict="REVIEW",
+                confidence="medium",
+                reason="Ambigu",
+                evidence=[],
+            ),
+        )
+        groups = {"grp": [_item(d)]}
+        result = compute_dim_synthesis(groups)
+        assert result["dp"]["review"] == 1
+        assert len(result["alertes"]["review"]) == 1
+
+    def test_das_decisions_counted(self):
+        d = DossierMedical(
+            diagnostics_associes=[
+                _diag("DAS1", "I10", decision=CodeDecision(action="KEEP")),
+                _diag("DAS2", "D69.6", decision=CodeDecision(action="RULED_OUT")),
+                _diag("DAS3", "D50", decision=CodeDecision(action="DOWNGRADE", final_code="D64.9")),
+                _diag("DAS4", "Z87.1", decision=CodeDecision(action="REMOVE")),
+            ],
+        )
+        groups = {"grp": [_item(d)]}
+        result = compute_dim_synthesis(groups)
+        assert result["das"]["total"] == 4
+        assert result["das"]["kept"] == 1
+        assert result["das"]["ruled_out"] == 1
+        assert result["das"]["downgraded"] == 1
+        assert result["das"]["removed"] == 1
+
+    def test_das_no_decision_counted_as_kept(self):
+        d = DossierMedical(
+            diagnostics_associes=[_diag("DAS", "I10")],  # no decision
+        )
+        groups = {"grp": [_item(d)]}
+        result = compute_dim_synthesis(groups)
+        assert result["das"]["kept"] == 1
+
+    def test_veto_report_aggregated(self):
+        d = DossierMedical(
+            veto_report=VetoReport(
+                verdict="FAIL",
+                score_contestabilite=40,
+                issues=[
+                    VetoIssue(veto="VETO-01", severity="HARD", where="dp", message="test"),
+                ],
+            ),
+        )
+        groups = {"grp": [_item(d)]}
+        result = compute_dim_synthesis(groups)
+        assert result["veto"]["avg_score"] == 40
+        assert result["veto"]["distribution"].get("FAIL") == 1
+        assert len(result["alertes"]["fail"]) == 1
+
+    def test_completude_indefendable(self):
+        d = DossierMedical(
+            completude=CompletudeDossier(
+                verdict_global="indefendable",
+                score_global=20,
+                documents_manquants=["CRO", "Anapath"],
+            ),
+        )
+        groups = {"grp": [_item(d)]}
+        result = compute_dim_synthesis(groups)
+        assert result["completude"]["distribution"].get("indefendable") == 1
+        assert len(result["alertes"]["indefendable"]) == 1
+
+    def test_taux_modification_zero_when_no_das(self):
+        result = compute_dim_synthesis({})
+        assert result["das"]["taux_modification"] == 0
+
+    def test_cpam_impact(self):
+        d = DossierMedical(
+            controles_cpam=[
+                ControleCPAM(
+                    numero_ogc=1,
+                    financial_impact=FinancialImpact(
+                        impact_estime_euros=1500,
+                        priorite="haute",
+                    ),
+                    validation_dim="valide",
+                ),
+            ],
+        )
+        groups = {"grp": [_item(d)]}
+        result = compute_dim_synthesis(groups)
+        assert result["cpam"]["total"] == 1
+        assert result["cpam"]["impact_total"] == 1500
+        assert result["cpam"]["by_priority"].get("haute") == 1
+        assert result["cpam"]["by_status"].get("valide") == 1
+
+
+# ===================================================================
+# format_cpam_text (additional edge cases beyond test_viewer.py)
+# ===================================================================
+
+
+class TestFormatCpamTextExtra:
+    def test_blank_lines_produce_br(self):
+        result = format_cpam_text("Line 1\n\nLine 2")
+        assert "<br>" in result
+
+    def test_list_closed_at_end(self):
+        """A list not followed by text should still be closed."""
+        result = format_cpam_text("- Item 1\n- Item 2")
+        assert result.count("<ul") == 1
+        assert result.count("</ul>") == 1