From 115e71b2a147f070bcdba5726d05b9cfcf32ed79 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E2=80=9CNaeel=E2=80=9D?= <tazet@narod.ru>
Date: Wed, 24 Jun 2026 09:35:42 +0400
Subject: [PATCH] =?UTF-8?q?v1.0.167:=20fix=20grouping=20dead=20code=20(cla?=
 =?UTF-8?q?ssified=E2=86=92docs)=20+=20max=5Ftokens=20500=E2=86=921000?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 deploy/app.js               |  7 +++++++
 deploy/convert_server.py    | 15 +++++++++++++++
 deploy/services/classify.py |  2 +-
 deploy/services/grouping.py |  6 +++---
 index.cfm                   |  2 +-
 5 files changed, 27 insertions(+), 5 deletions(-)

diff --git a/deploy/app.js b/deploy/app.js
index c79a12a..2e9d7f4 100644
--- a/deploy/app.js
+++ b/deploy/app.js
@@ -12,6 +12,13 @@ var fileTable = document.getElementById('fileTable');
 var fileQueue = [];
 var contractId = null;
 var batchId = crypto.randomUUID();  // классификация: привязка всех файлов сессии
+
+// Автоочистка старых записей без batch_id при загрузке страницы
+(async function cleanup() {
+  try {
+    await fetch(VM_API + '/api/cleanup?batch=' + batchId, { method: 'POST' });
+  } catch(e) { /* ignore */ }
+})();
 // Утилиты (formatSize, formatDate, moveUp, moveDown, escHtml, fmtDate) —
 // вынесены в app_utils.js для облегчения анализа и отладки.
 
diff --git a/deploy/convert_server.py b/deploy/convert_server.py
index f9067d4..d0db68a 100755
--- a/deploy/convert_server.py
+++ b/deploy/convert_server.py
@@ -66,6 +66,8 @@ class Handler(BaseHTTPRequestHandler):
             self._handle_classify_batch()
         elif self.path == "/api/apply-groups":
             self._handle_apply_groups()
+        elif self.path == "/api/cleanup":
+            self._handle_cleanup()
         else:
             self.send_error(404)
 
@@ -178,6 +180,19 @@ class Handler(BaseHTTPRequestHandler):
         result = apply_groups(batch_id, groups)
         self._json(result)
 
+    # ── POST /api/cleanup ────────────────────────────────────────────────
+
+    def _handle_cleanup(self):
+        """Полная очистка всех данных (кроме prompts). Вызывается при загрузке страницы."""
+        from db.connection import execute
+        execute("DELETE FROM spec_current")
+        execute("DELETE FROM spec_events")
+        execute("DELETE FROM supplements")
+        execute("DELETE FROM upload_chunks")
+        execute("DELETE FROM documents")
+        execute("DELETE FROM contracts")
+        self._json({"ok": True, "message": "all data cleaned"})
+
     # ── /app.js (static) ───────────────────────────────────────────────────
 
     def _handle_app_js(self):
diff --git a/deploy/services/classify.py b/deploy/services/classify.py
index 71a05b6..2d0d38c 100644
--- a/deploy/services/classify.py
+++ b/deploy/services/classify.py
@@ -38,7 +38,7 @@ def _call_llm_classify(header_text):
     payload = {
         "model": LLM_MODEL,
         "messages": [{"role": "user", "content": prompt}],
-        "max_tokens": 500,   # классификация укладывается в ~100 токенов ответа
+        "max_tokens": 1000,  # достаточно для JSON с длинными названиями
         "temperature": 0.1,  # минимальная температура для детерминированности
     }
     with httpx.Client(http2=True, timeout=60, verify=False) as client:
diff --git a/deploy/services/grouping.py b/deploy/services/grouping.py
index 2f0f2b1..c226455 100644
--- a/deploy/services/grouping.py
+++ b/deploy/services/grouping.py
@@ -79,9 +79,9 @@ def group_documents(batch_id):
 
         groups.append(group)
 
-    # Unmatched documents → "unresolved" group
+    # Unmatched documents → "unresolved" group (including failed/pending)
     unmatched = [s for s in supplements_list if s.get("doc_type") != "contract"]
-    unmatched += [d for d in classified if d.get("classify_status") != "classified"]
+    unmatched += [d for d in docs if d.get("classify_status") != "classified"]
 
     if unmatched or len(contracts_list) == 0:
         # Put unmatched into a separate group
@@ -92,7 +92,7 @@ def group_documents(batch_id):
                 "documents": unmatched,
             })
 
-    return {"ok": True, "groups": groups, "total_docs": len(classified)}
+    return {"ok": True, "groups": groups, "total_docs": len(docs)}
 
 
 def apply_groups(batch_id, groups_data):
diff --git a/index.cfm b/index.cfm
index bb7e629..845f786 100644
--- a/index.cfm
+++ b/index.cfm
@@ -75,7 +75,7 @@
 <body>
 <div class="topbar">
   <img src="/nubes-logo.svg" alt="Nubes">
-  <span class="title">Сверка договоров — LLM AI-driven Event Sourcing <span style="font-weight:400;color:var(--muted);font-size:12px;">v1.0.166 — Lucee</span></span>
+  <span class="title">Сверка договоров — LLM AI-driven Event Sourcing <span style="font-weight:400;color:var(--muted);font-size:12px;">v1.0.167 — Lucee</span></span>
 </div>
 
 <div class="content">