Dom/t2a
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: mode hybride Ollama — gemma3:27b pour CPAM, 12b pour codage

Browse Source 
Le pipeline utilise désormais gemma3:12b (rapide) pour le codage CIM-10
et gemma3:27b (meilleur raisonnement) pour la contre-argumentation CPAM.
Configurable via OLLAMA_MODEL_CPAM et OLLAMA_TIMEOUT_CPAM.

Inclut aussi : traçabilité source/page DAS, niveaux CMA ATIH, sévérité,
page tracker PDF, améliorations fusion et filtres DAS.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
dom
2026-02-17 17:53:53 +01:00
parent 4ef42dd3d3
commit 01d47f3c4b
20 changed files with 1025 additions and 98 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
src/viewer/app.py
View File

@@ -305,6 +305,13 @@ _SEVERITY_STYLES = {
"leger": ("Léger", "#065f46", "#d1fae5"),
}
_CMA_LEVEL_STYLES = {
1: ("1", "#6b7280", "#f3f4f6"), # gris — pas CMA
2: ("2", "#065f46", "#d1fae5"), # vert
3: ("3", "#92400e", "#fef3c7"), # jaune/orange
4: ("4", "#dc2626", "#fee2e2"), # rouge
}
def format_duration(seconds: float | None) -> str:
"""Formate une durée en secondes vers un format lisible (ex: 2min 30s)."""
@@ -330,13 +337,24 @@ def severity_badge(value: str | None) -> Markup:
)
def cma_level_badge(value: int | None) -> Markup:
"""Badge CMA niveau 1-4 avec couleurs graduées."""
if value is None or value < 1:
return Markup("")
level = min(value, 4)
label, fg, bg = _CMA_LEVEL_STYLES.get(level, _CMA_LEVEL_STYLES[1])
title = {1: "Pas CMA", 2: "CMA niveau 2", 3: "CMA niveau 3", 4: "CMA niveau 4"}.get(level, "")
return Markup(
f'<span title="{title}" style="display:inline-block;padding:2px 8px;border-radius:9999px;'
f'font-size:0.75rem;font-weight:600;color:{fg};background:{bg}">'
f'CMA {label}</span>'
)
def format_dossier_name(name: str) -> str:
"""Transforme un nom de dossier en nom lisible (ex: 15_23096332 → Dossier 23096332)."""
"""Retourne le nom complet du dossier (ex: 1_23096332)."""
if name == "racine":
return "Non classés"
m = re.match(r"^\d+_(\d+)$", name)
if m:
return f"Dossier {m.group(1)}"
return name
@@ -364,6 +382,7 @@ def create_app() -> Flask:
app.jinja_env.filters["confidence_badge"] = confidence_badge
app.jinja_env.filters["confidence_label"] = confidence_label
app.jinja_env.filters["severity_badge"] = severity_badge
app.jinja_env.filters["cma_level_badge"] = cma_level_badge
app.jinja_env.filters["format_duration"] = format_duration
app.jinja_env.filters["format_dossier_name"] = format_dossier_name
app.jinja_env.filters["format_doc_name"] = format_doc_name
@@ -445,13 +464,16 @@ def create_app() -> Flask:
return jsonify({"error": f"PDF source '{source_file}' introuvable"}), 404
try:
anonymized_text, new_dossier, report = process_pdf(pdf_path)
pdf_results = process_pdf(pdf_path)
stem = pdf_path.stem.replace(" ", "_")
subdir = None
if pdf_path.parent != input_dir:
subdir = pdf_path.parent.name
write_outputs(stem, anonymized_text, new_dossier, report, subdir=subdir)
return jsonify({"ok": True, "message": "Traitement terminé"})
multi = len(pdf_results) > 1
for part_idx, (anonymized_text, new_dossier, report) in enumerate(pdf_results):
part_stem = f"{stem}_part{part_idx + 1}" if multi else stem
write_outputs(part_stem, anonymized_text, new_dossier, report, subdir=subdir)
return jsonify({"ok": True, "message": f"Traitement terminé ({len(pdf_results)} dossier(s))"})
except Exception as e:
logger.exception("Erreur lors du retraitement")
return jsonify({"error": str(e)}), 500