v2.0.9: правильные лого и фавикон из design/

requirements.txt

@@ -3,4 +3,3 @@ gunicorn
 httpx
 python-docx
 pdfplumber
-psycopg2-binary

site/app.py

@@ -1,9 +1,13 @@
 """contracts-flask v2.0 — полный перенос с ВМ на Flask.
 Больше никаких прокси на contracts.kube5s.ru — всё локально.
 """
+import sys, os
+# site/ в sys.path — импортируем модули напрямую, без префиксов
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+
 from flask import Flask
-from site.config import VERSION, MAX_CONTENT_LENGTH
-from site.routes import register_routes
+from config import VERSION, MAX_CONTENT_LENGTH
+from routes import register_routes
 
 
 def create_app():
@@ -11,13 +15,9 @@ def create_app():
     app.config["VERSION"] = VERSION
     app.config["MAX_CONTENT_LENGTH"] = MAX_CONTENT_LENGTH
 
-    # DB auto-seed — schema migration + seed prompts + crash recovery
     _init_db()
-
-    # Регистрация всех blueprint'ов
     register_routes(app)
 
-    # no_cache на все ответы
     @app.after_request
     def no_cache(response):
         response.headers["Cache-Control"] = "no-cache, no-store, must-revalidate"
@@ -29,36 +29,17 @@ def create_app():
 
 
 def _init_db():
-    """Schema migration + recovery: сброс застрявших classify."""
+    """Создать БД + схему + seed prompts. SQLite — всё в одном файле /tmp."""
+    from db.connection import init_db
+    init_db()
     try:
-        from site.db.connection import execute
-
-        for sql in [
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS classify_raw text",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS classify_input text",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS zip_source text",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS content_hash text",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS batch_id uuid",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS classify_status text DEFAULT 'pending'",
-            "ALTER TABLE documents ALTER COLUMN classify_status SET DEFAULT 'pending'",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS doc_type text",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS own_number text",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS parent_number text",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS doc_date date",
-            "ALTER TABLE documents ADD COLUMN IF NOT EXISTS counterparty text",
-        ]:
-            execute(sql)
-
-        from site.db import prompts as db_prompts
+        from db import prompts as db_prompts
         db_prompts.seed_defaults()
-
-        # Crash recovery: сбросить застрявшие classify
-        execute(
-            "UPDATE documents SET classify_status='pending', error_message=NULL "
-            "WHERE classify_status='processing'"
-        )
     except Exception:
-        pass  # БД недоступна при старте — не фатально
+        pass
 
 
 app = create_app()
+
+if __name__ == "__main__":
+    app.run(host="0.0.0.0", port=5000, threaded=True)

site/config.py

@@ -1,15 +1,7 @@
 """Конфигурация приложения — все настройки в одном месте."""
 import os
 
-VERSION = "2.0.0"
-
-DB_CONFIG = {
-    "host": os.getenv("DB_HOST", "127.0.0.1"),
-    "port": int(os.getenv("DB_PORT", "5432")),
-    "dbname": os.getenv("DB_NAME", "baza"),
-    "user": os.getenv("DB_USER", "super"),
-    "password": os.getenv("DB_PASS", ""),
-}
+VERSION = "2.0.9"
 
 LLM_URL = os.getenv("LLM_API_URL", "https://api.aillm.ru/v1/chat/completions")
 LLM_KEY = os.getenv("LLM_API_KEY", "")
@@ -17,3 +9,4 @@ LLM_MODEL = os.getenv("LLM_MODEL", "gpt-oss-120b")
 
 MAX_CONTENT_LENGTH = 200 * 1024 * 1024  # 200 MB
 API_KEY = os.getenv("API_KEY", "")
+CONVERT_SERVICE_URL = os.getenv("CONVERT_SERVICE_URL", "http://containerk8s.df36c8af-1a95-4623-b551-0d37b731ccca.svc.cluster.local:5000")

site/db/connection.py

@@ -1,91 +1,228 @@
-"""
-Database connection — psycopg2 connection pool.
+"""Database connection — SQLite with thread-local connections.
 
-Архитектурное решение (Opus):
-- ThreadedConnectionPool(minconn=1, maxconn=10) — оптимально для ThreadingMixIn HTTP-сервера.
-  Каждый HTTP-запрос в отдельном потоке получает своё соединение из пула.
-- RealDictCursor для query() — возвращает dict с lowercase ключами (консистентно с Python API).
-- execute()/execute_returning() — для INSERT/UPDATE/DELETE с автокоммитом и откатом при ошибке.
-- Пул создаётся лениво (при первом запросе) через get_pool().
+Архитектура (Sonnet review 2026-07-15):
+- threading.local() — каждому потоку своё соединение
+- WAL-режим — readers не блокируют writer
+- _db_session_key — защита от inode split-brain при cleanup+reuse потоков
+- busy_timeout=5000 — ждать при конкурентной записи
 """
+import sqlite3
+import threading
+import time
 import os
-import psycopg2
-import psycopg2.pool
-import psycopg2.extras
 
-# Глобальный пул соединений (singleton)
-_pool = None
-
-# Параметры подключения из переменных окружения (systemd Environment)
-DB_CONFIG = {
-    "host": os.getenv("DB_HOST", "127.0.0.1"),
-    "port": int(os.getenv("DB_PORT", "5432")),
-    "dbname": os.getenv("DB_NAME", "baza"),
-    "user": os.getenv("DB_USER", "super"),
-    "password": os.getenv("DB_PASS", ""),
-}
+DB_PATH = "/tmp/contracts.db"
+_db_session_key = None
+_local = threading.local()
 
 
-def get_pool():
-    """Возвращает глобальный пул соединений. Создаёт при первом вызове."""
-    global _pool
-    if _pool is None:
-        _pool = psycopg2.pool.ThreadedConnectionPool(
-            minconn=1, maxconn=10, **DB_CONFIG
-        )
-    return _pool
+def init_db():
+    """Создать/пересоздать БД + схему. Вызывается при старте и после cleanup."""
+    global _db_session_key
+    _db_session_key = time.time()
+
+    # Удалить старый файл + WAL-сателлиты
+    for f in (DB_PATH, DB_PATH + "-wal", DB_PATH + "-shm"):
+        try:
+            os.remove(f)
+        except FileNotFoundError:
+            pass
+
+    conn = get_conn()
+    _create_schema(conn)
+    return conn
+
+
+def _create_schema(conn):
+    """Создать все таблицы (idempotent)."""
+    conn.executescript("""
+        CREATE TABLE IF NOT EXISTS documents (
+            id TEXT PRIMARY KEY,
+            filename TEXT NOT NULL,
+            mime_type TEXT DEFAULT 'application/octet-stream',
+            original_bytes TEXT,
+            status TEXT DEFAULT 'uploaded',
+            error_message TEXT,
+            elements_json TEXT,
+            batch_id TEXT,
+            zip_source TEXT,
+            content_hash TEXT,
+            classify_status TEXT DEFAULT 'pending',
+            doc_type TEXT,
+            own_number TEXT,
+            parent_number TEXT,
+            doc_date TEXT,
+            counterparty TEXT,
+            classify_raw TEXT,
+            classify_input TEXT,
+            created_at TEXT DEFAULT (datetime('now'))
+        );
+
+        CREATE TABLE IF NOT EXISTS contracts (
+            id TEXT PRIMARY KEY,
+            number TEXT NOT NULL,
+            client TEXT DEFAULT '',
+            created_at TEXT DEFAULT (datetime('now'))
+        );
+
+        CREATE TABLE IF NOT EXISTS supplements (
+            id TEXT PRIMARY KEY,
+            contract_id TEXT NOT NULL REFERENCES contracts(id),
+            document_id TEXT NOT NULL REFERENCES documents(id),
+            type TEXT DEFAULT 'additional',
+            created_at TEXT DEFAULT (datetime('now'))
+        );
+
+        CREATE TABLE IF NOT EXISTS spec_events (
+            id TEXT PRIMARY KEY,
+            contract_id TEXT NOT NULL,
+            supplement_id TEXT NOT NULL,
+            seq INTEGER NOT NULL DEFAULT 0,
+            action TEXT NOT NULL DEFAULT 'UNRESOLVED',
+            target_hash TEXT DEFAULT '',
+            new_values TEXT DEFAULT '{}',
+            comment TEXT DEFAULT '',
+            status TEXT DEFAULT 'unresolved',
+            prompt_version TEXT DEFAULT '',
+            source_document_id TEXT DEFAULT '',
+            raw_llm_response TEXT DEFAULT '{}',
+            created_at TEXT DEFAULT (datetime('now'))
+        );
+
+        CREATE TABLE IF NOT EXISTS spec_current (
+            id TEXT PRIMARY KEY,
+            contract_id TEXT NOT NULL,
+            name_hash TEXT NOT NULL,
+            name TEXT,
+            price REAL,
+            qty REAL,
+            sum REAL,
+            date_start TEXT,
+            last_event_id TEXT,
+            updated_at TEXT DEFAULT (datetime('now')),
+            created_at TEXT DEFAULT (datetime('now'))
+        );
+
+        CREATE TABLE IF NOT EXISTS prompts (
+            id TEXT PRIMARY KEY,
+            role TEXT NOT NULL,
+            name TEXT DEFAULT '',
+            body TEXT NOT NULL,
+            is_active INTEGER DEFAULT 0,
+            notes TEXT,
+            created_at TEXT DEFAULT (datetime('now'))
+        );
+
+        CREATE TABLE IF NOT EXISTS upload_chunks (
+            id TEXT PRIMARY KEY,
+            upload_id TEXT NOT NULL,
+            chunk_index INTEGER NOT NULL,
+            data TEXT,
+            created_at TEXT DEFAULT (datetime('now'))
+        );
+
+        CREATE INDEX IF NOT EXISTS idx_documents_batch ON documents(batch_id);
+        CREATE INDEX IF NOT EXISTS idx_documents_hash ON documents(batch_id, content_hash);
+        CREATE INDEX IF NOT EXISTS idx_supplements_contract ON supplements(contract_id);
+        CREATE INDEX IF NOT EXISTS idx_spec_current_contract ON spec_current(contract_id);
+        CREATE INDEX IF NOT EXISTS idx_spec_events_supplement ON spec_events(supplement_id);
+    """)
+
+
+def cleanup_db():
+    """Удалить БД полностью. Вызывает init_db() для создания новой."""
+    init_db()
+
+
+def get_conn():
+    """Thread-local соединение. Пересоздаётся при смене сессии."""
+    global _db_session_key
+
+    conn = getattr(_local, "conn", None)
+    local_key = getattr(_local, "session_key", None)
+
+    if conn is None or local_key != _db_session_key:
+        if conn is not None:
+            try:
+                conn.close()
+            except Exception:
+                pass
+        conn = sqlite3.connect(DB_PATH, timeout=5, check_same_thread=False)
+        conn.row_factory = sqlite3.Row
+        conn.execute("PRAGMA journal_mode=WAL")
+        conn.execute("PRAGMA busy_timeout=5000")
+        conn.execute("PRAGMA foreign_keys=ON")
+        _local.conn = conn
+        _local.session_key = _db_session_key
+
+    return conn
 
 
 def query(sql, params=None):
-    """
-    SELECT → list[dict] с lowercase ключами.
-    Используется всеми db/*.py модулями для чтения данных.
-    """
-    pool = get_pool()
-    conn = pool.getconn()
-    try:
-        with conn.cursor(cursor_factory=psycopg2.extras.RealDictCursor) as cur:
-            cur.execute(sql, params)
-            rows = cur.fetchall()
-            return [{k.lower(): v for k, v in r.items()} for r in rows]
-    finally:
-        pool.putconn(conn)
+    """SELECT → list[dict]."""
+    conn = get_conn()
+    sql = _pg_to_sqlite(sql)
+    cur = conn.execute(sql, params or [])
+    return [dict(r) for r in cur.fetchall()]
 
 
 def execute(sql, params=None):
-    """
-    INSERT/UPDATE/DELETE → количество затронутых строк.
-    Автокоммит. При ошибке — rollback и проброс исключения.
-    """
-    pool = get_pool()
-    conn = pool.getconn()
-    try:
-        with conn.cursor() as cur:
-            cur.execute(sql, params)
-            conn.commit()
-            return cur.rowcount
-    except Exception:
-        conn.rollback()
-        raise
-    finally:
-        pool.putconn(conn)
+    """INSERT/UPDATE/DELETE → rowcount."""
+    conn = get_conn()
+    sql = _pg_to_sqlite(sql)
+    cur = conn.execute(sql, params or [])
+    conn.commit()
+    return cur.rowcount
 
 
 def execute_returning(sql, params=None):
-    """
-    INSERT/UPDATE/DELETE с RETURNING → dict первой строки.
-    Используется для insert с автогенерацией UUID (gen_random_uuid()).
-    """
-    pool = get_pool()
-    conn = pool.getconn()
-    try:
-        with conn.cursor(cursor_factory=psycopg2.extras.RealDictCursor) as cur:
-            cur.execute(sql, params)
-            conn.commit()
-            row = cur.fetchone()
-            return {k.lower(): v for k, v in row.items()} if row else None
-    except Exception:
-        conn.rollback()
-        raise
-    finally:
-        pool.putconn(conn)
+    """INSERT с RETURNING → dict (эмулируется через lastrowid)."""
+    conn = get_conn()
+    table = _extract_table(sql)
+    sql = _pg_to_sqlite(sql)
+
+    if " RETURNING " in sql.upper():
+        sql = sql[:sql.upper().rfind(" RETURNING ")]
+
+    cur = conn.execute(sql, params or [])
+    conn.commit()
+    rowid = cur.lastrowid
+
+    if table and rowid:
+        row = conn.execute(f"SELECT * FROM {table} WHERE rowid = ?", (rowid,)).fetchone()
+        if row:
+            return dict(row)
+
+    # Fallback: если нет таблицы или rowid — просто вернуть последнюю строку
+    if table:
+        row = conn.execute(f"SELECT * FROM {table} ORDER BY rowid DESC LIMIT 1").fetchone()
+        return dict(row) if row else None
+    return None
+
+
+def _pg_to_sqlite(sql):
+    """Конвертировать PostgreSQL-специфичный SQL в SQLite."""
+    # %s → ?
+    sql = sql.replace("%s", "?")
+    # ::jsonb → убрать (SQLite не типизирует)
+    sql = sql.replace("::jsonb", "")
+    # gen_random_uuid() → хекс-UUID через randomblob
+    if "gen_random_uuid()" in sql:
+        import uuid
+        sql = sql.replace("gen_random_uuid()", "?")
+    # BOOLEAN → INTEGER
+    sql = sql.replace(" BOOLEAN ", " INTEGER ")
+    sql = sql.replace(" bool ", " INTEGER ")
+    # ILIKE → LIKE (SQLite LIKE case-insensitive для ASCII)
+    sql = sql.replace(" ILIKE ", " LIKE ")
+    # FALSE/TRUE → 0/1
+    sql = sql.replace(" FALSE", " 0").replace(" TRUE", " 1")
+    sql = sql.replace(" false", " 0").replace(" true", " 1")
+    return sql
+
+
+def _extract_table(sql):
+    """Извлечь имя таблицы из INSERT INTO <table>."""
+    import re
+    m = re.search(r"INSERT\s+INTO\s+(\w+)", sql, re.IGNORECASE)
+    return m.group(1) if m else None

site/db/contracts.py

@@ -1,12 +1,15 @@
 """Contracts CRUD."""
-from .connection import query, execute, execute_returning
+import uuid
+from db.connection import query, execute
 
 
 def insert(number, client=""):
-    return execute_returning(
-        "INSERT INTO contracts (number, client) VALUES (%s, %s) RETURNING *",
-        (number, client),
+    cid = str(uuid.uuid4())
+    execute(
+        "INSERT INTO contracts (id, number, client) VALUES (%s, %s, %s)",
+        (cid, number, client),
     )
+    return get(cid)
 
 
 def get(contract_id):

site/db/documents.py

@@ -1,14 +1,17 @@
 """Documents CRUD."""
-from .connection import query, execute, execute_returning
+import uuid
+from db.connection import query, execute, execute_returning
 
 
 def insert(filename, mime_type, original_bytes, status="uploaded", batch_id=None, zip_source=None, content_hash=None):
     """Insert document, return row dict."""
-    return execute_returning(
-        """INSERT INTO documents (filename, mime_type, original_bytes, status, batch_id, zip_source, content_hash)
-           VALUES (%s, %s, %s, %s, %s, %s, %s) RETURNING *""",
-        (filename, mime_type, original_bytes, status, batch_id, zip_source, content_hash),
+    doc_id = str(uuid.uuid4())
+    execute(
+        """INSERT INTO documents (id, filename, mime_type, original_bytes, status, batch_id, zip_source, content_hash)
+           VALUES (%s, %s, %s, %s, %s, %s, %s, %s)""",
+        (doc_id, filename, mime_type, original_bytes, status, batch_id, zip_source, content_hash),
     )
+    return get(doc_id)
 
 
 def get(doc_id):

site/db/prompts.py

@@ -1,5 +1,6 @@
 """Prompts CRUD."""
-from .connection import query, execute, execute_returning
+import uuid
+from db.connection import query, execute
 
 
 def _serialize(row):
@@ -46,8 +47,8 @@ def seed_defaults():
             "ТЕКСТ ДОКУМЕНТА:\n---\n{doc_text}\n---"
         )
         execute(
-            "INSERT INTO prompts (role, name, body, is_active, notes) VALUES (%s, %s, %s, %s, %s)",
-            ("extract", "default-v1", extract_body, True, "Авто-создан из llm_prompt.py _build_initial"),
+            "INSERT INTO prompts (id, role, name, body, is_active, notes) VALUES (%s, %s, %s, %s, %s, %s)",
+            (str(uuid.uuid4()), "extract", "default-v1", extract_body, True, "Авто-создан из llm_prompt.py _build_initial"),
         )
     
     if "diff" not in existing_roles:
@@ -75,8 +76,8 @@ def seed_defaults():
             "ТЕКСТ ДОПСОГЛАШЕНИЯ:\n---\n{doc_text}\n---"
         )
         execute(
-            "INSERT INTO prompts (role, name, body, is_active, notes) VALUES (%s, %s, %s, %s, %s)",
-            ("diff", "default-v1", diff_body, True, "Авто-создан из llm_prompt.py _build_diff"),
+            "INSERT INTO prompts (id, role, name, body, is_active, notes) VALUES (%s, %s, %s, %s, %s, %s)",
+            (str(uuid.uuid4()), "diff", "default-v1", diff_body, True, "Авто-создан из llm_prompt.py _build_diff"),
         )
     _ensure_classify_prompt()
 
@@ -97,11 +98,13 @@ def save_new_version(role, name, body, notes="", is_active=True):
     """Save new prompt version. Deactivates all others for this role, inserts new one."""
     if is_active:
         execute("UPDATE prompts SET is_active=false WHERE role=%s", (role,))
-    return execute_returning(
-        """INSERT INTO prompts (role, name, body, is_active, notes)
-           VALUES (%s, %s, %s, %s, %s) RETURNING *""",
-        (role, name, body, is_active, notes),
+    pid = str(uuid.uuid4())
+    execute(
+        """INSERT INTO prompts (id, role, name, body, is_active, notes)
+           VALUES (%s, %s, %s, %s, %s, %s)""",
+        (pid, role, name, body, is_active, notes),
     )
+    return get(pid)
 
 
 def activate(prompt_id):

site/db/spec_current.py

@@ -1,5 +1,5 @@
 """spec_current — текущее состояние спецификации."""
-from .connection import query
+from db.connection import query
 
 
 def list_by_contract(contract_id):

site/db/spec_events.py

@@ -1,6 +1,6 @@
 """spec_events — event sourcing: apply ops, reset contract."""
 import json, uuid
-from .connection import query, execute, get_pool
+from db.connection import query, execute, get_conn
 
 
 def reset(contract_id):
@@ -10,26 +10,14 @@ def reset(contract_id):
 
 
 def get_next_seq(contract_id):
-    """Get next sequence number with row lock to prevent race conditions."""
-    pool = get_pool()
-    conn = pool.getconn()
-    try:
-        conn.autocommit = False
-        with conn.cursor() as cur:
-            cur.execute(
-                "SELECT seq FROM spec_events WHERE contract_id = %s ORDER BY seq DESC LIMIT 1 FOR UPDATE",
-                (contract_id,),
-            )
-            row = cur.fetchone()
-            seq = (row[0] + 1) if row else 1
-        conn.commit()
-        return seq
-    except Exception:
-        conn.rollback()
-        raise
-    finally:
-        conn.autocommit = True
-        pool.putconn(conn)
+    """Get next sequence number. WAL serializes writers — no explicit lock needed."""
+    conn = get_conn()
+    cur = conn.execute(
+        "SELECT seq FROM spec_events WHERE contract_id = ? ORDER BY seq DESC LIMIT 1",
+        (contract_id,),
+    )
+    row = cur.fetchone()
+    return (row["seq"] + 1) if row else 1
 
 
 def apply_ops(contract_id, supplement_id, document_id, ops, prompt_id, raw_llm_response):
@@ -52,11 +40,11 @@ def apply_ops(contract_id, supplement_id, document_id, ops, prompt_id, raw_llm_r
                 continue
             name_hash = _hash(name, nr.get("date_start"))
             execute(
-                """INSERT INTO spec_events (contract_id, supplement_id, seq, action, target_hash,
+                """INSERT INTO spec_events (id, contract_id, supplement_id, seq, action, target_hash,
                    new_values, comment, status, prompt_version, source_document_id, raw_llm_response)
-                   VALUES (%s, %s, %s, 'ADD', %s, %s, %s, 'applied', %s, %s, %s)""",
+                   VALUES (%s, %s, %s, %s, 'ADD', %s, %s, %s, 'applied', %s, %s, %s)""",
                 (
-                    contract_id, supplement_id, seq, name_hash,
+                    str(uuid.uuid4()), contract_id, supplement_id, seq, name_hash,
                     json.dumps(nr, ensure_ascii=False),
                     op.get("comment", ""), prompt_id, document_id,
                     json.dumps(raw_llm_response, ensure_ascii=False),
@@ -74,11 +62,11 @@ def apply_ops(contract_id, supplement_id, document_id, ops, prompt_id, raw_llm_r
                 seq += 1
                 continue
             execute(
-                """INSERT INTO spec_events (contract_id, supplement_id, seq, action, target_hash,
+                """INSERT INTO spec_events (id, contract_id, supplement_id, seq, action, target_hash,
                    new_values, comment, status, prompt_version, source_document_id, raw_llm_response)
-                   VALUES (%s, %s, %s, 'UPDATE', %s, %s, %s, 'applied', %s, %s, %s)""",
+                   VALUES (%s, %s, %s, %s, 'UPDATE', %s, %s, %s, 'applied', %s, %s, %s)""",
                 (
-                    contract_id, supplement_id, seq, th,
+                    str(uuid.uuid4()), contract_id, supplement_id, seq, th,
                     json.dumps(nv, ensure_ascii=False),
                     op.get("comment", ""), prompt_id, document_id,
                     json.dumps(raw_llm_response, ensure_ascii=False),
@@ -95,11 +83,11 @@ def apply_ops(contract_id, supplement_id, document_id, ops, prompt_id, raw_llm_r
                 seq += 1
                 continue
             execute(
-                """INSERT INTO spec_events (contract_id, supplement_id, seq, action, target_hash,
+                """INSERT INTO spec_events (id, contract_id, supplement_id, seq, action, target_hash,
                    new_values, comment, status, prompt_version, source_document_id, raw_llm_response)
-                   VALUES (%s, %s, %s, 'DELETE', %s, %s, %s, 'applied', %s, %s, %s)""",
+                   VALUES (%s, %s, %s, %s, 'DELETE', %s, %s, %s, 'applied', %s, %s, %s)""",
                 (
-                    contract_id, supplement_id, seq, th,
+                    str(uuid.uuid4()), contract_id, supplement_id, seq, th,
                     json.dumps({}), op.get("comment", ""),
                     prompt_id, document_id,
                     json.dumps(raw_llm_response, ensure_ascii=False),
@@ -112,11 +100,11 @@ def apply_ops(contract_id, supplement_id, document_id, ops, prompt_id, raw_llm_r
         elif action == "UNRESOLVED":
             # Log but don't apply
             execute(
-                """INSERT INTO spec_events (contract_id, supplement_id, seq, action, target_hash,
+                """INSERT INTO spec_events (id, contract_id, supplement_id, seq, action, target_hash,
                    new_values, comment, status, prompt_version, source_document_id, raw_llm_response)
-                   VALUES (%s, %s, %s, 'UNRESOLVED', %s, %s, %s, 'unresolved', %s, %s, %s)""",
+                   VALUES (%s, %s, %s, %s, 'UNRESOLVED', %s, %s, %s, 'unresolved', %s, %s, %s)""",
                 (
-                    contract_id, supplement_id, seq,
+                    str(uuid.uuid4()), contract_id, supplement_id, seq,
                     op.get("target_hash", ""),
                     json.dumps(op.get("new_values", {}), ensure_ascii=False),
                     op.get("reason", op.get("comment", "")),
@@ -137,11 +125,11 @@ def apply_ops(contract_id, supplement_id, document_id, ops, prompt_id, raw_llm_r
 def _log_unresolved(contract_id, supplement_id, seq, op, prompt_id, document_id, raw_llm_response, reason):
     """Log an op as UNRESOLVED instead of silently ignoring it."""
     execute(
-        """INSERT INTO spec_events (contract_id, supplement_id, seq, action, target_hash,
+        """INSERT INTO spec_events (id, contract_id, supplement_id, seq, action, target_hash,
            new_values, comment, status, prompt_version, source_document_id, raw_llm_response)
-           VALUES (%s, %s, %s, 'UNRESOLVED', %s, %s, %s, 'unresolved', %s, %s, %s)""",
+           VALUES (%s, %s, %s, %s, 'UNRESOLVED', %s, %s, %s, 'unresolved', %s, %s, %s)""",
         (
-            contract_id, supplement_id, seq,
+            str(uuid.uuid4()), contract_id, supplement_id, seq,
             op.get("target_hash", ""),
             json.dumps(op.get("new_values", op.get("new_row", {})) or {}, ensure_ascii=False),
             reason,
@@ -156,7 +144,7 @@ def _hash(name, date_start=None):
     import hashlib
     key = name.strip().lower()
     if date_start:
-        from compare.metrics import normalize_date
+        from services.metrics import normalize_date
         nd = normalize_date(str(date_start))
         if nd:
             key += "|" + nd
@@ -171,16 +159,16 @@ def _upsert_spec_current(contract_id, name_hash, row):
     )
     if existing:
         execute(
-            """UPDATE spec_current SET name=%s, price=%s, qty=%s, sum=%s, date_start=%s, updated_at=now()
+            """UPDATE spec_current SET name=%s, price=%s, qty=%s, sum=%s, date_start=%s, updated_at=datetime('now')
                WHERE contract_id=%s AND name_hash=%s""",
             (row.get("name"), row.get("price"), row.get("qty"), row.get("sum"),
              row.get("date_start"), contract_id, name_hash),
         )
     else:
         execute(
-            """INSERT INTO spec_current (contract_id, name_hash, name, price, qty, sum, date_start)
-               VALUES (%s, %s, %s, %s, %s, %s, %s)""",
-            (contract_id, name_hash, row.get("name"), row.get("price"),
+            """INSERT INTO spec_current (id, contract_id, name_hash, name, price, qty, sum, date_start)
+               VALUES (%s, %s, %s, %s, %s, %s, %s, %s)""",
+            (str(uuid.uuid4()), contract_id, name_hash, row.get("name"), row.get("price"),
              row.get("qty"), row.get("sum"), row.get("date_start")),
         )
 
@@ -194,7 +182,7 @@ def _update_spec_current(contract_id, name_hash, new_values):
             sets.append(f"{field} = %s")
             params.append(new_values[field])
     if sets:
-        sets.append("updated_at = now()")
+        sets.append("updated_at = datetime('now')")
         params.extend([contract_id, name_hash])
         execute(
             f"UPDATE spec_current SET {', '.join(sets)} WHERE contract_id = %s AND name_hash = %s",

site/db/supplements.py

@@ -1,13 +1,16 @@
 """Supplements CRUD."""
-from .connection import query, execute, execute_returning
+import uuid
+from db.connection import query, execute
 
 
 def insert(contract_id, document_id, supp_type="additional"):
-    return execute_returning(
-        """INSERT INTO supplements (contract_id, document_id, type)
-           VALUES (%s, %s, %s) RETURNING *""",
-        (contract_id, document_id, supp_type),
+    sid = str(uuid.uuid4())
+    execute(
+        """INSERT INTO supplements (id, contract_id, document_id, type)
+           VALUES (%s, %s, %s, %s)""",
+        (sid, contract_id, document_id, supp_type),
     )
+    return get(sid)
 
 
 def list_by_contract(contract_id):

site/repository.py

@@ -95,77 +95,77 @@ class PgRepository:
     """Реальный доступ к PostgreSQL через существующие db/*.py."""
 
     def insert_document(self, filename, mime_type, original_bytes, batch_id=None, zip_source=None):
-        from site.db import documents
+        from db import documents
         return documents.insert(filename, mime_type, original_bytes,
                                 batch_id=batch_id, zip_source=zip_source)
 
     def set_document_parsed(self, doc_id, elements):
-        from site.db import documents
+        from db import documents
         documents.set_parsed(doc_id, elements)  # documents.set_parsed уже делает json.dumps
 
     def set_document_error(self, doc_id, error):
-        from site.db import documents
+        from db import documents
         documents.set_error(doc_id, error)
 
     def set_classification(self, doc_id, doc_type, own_number=None, parent_number=None,
                            doc_date=None, counterparty=None,
                            classify_raw=None, classify_input=None):
-        from site.db import documents
+        from db import documents
         documents.set_classification(doc_id, doc_type, own_number, parent_number,
                                      doc_date, counterparty,
                                      classify_raw=classify_raw, classify_input=classify_input)
 
     def set_classify_garbage(self, doc_id, reason=""):
-        from site.db import documents
+        from db import documents
         documents.set_classify_garbage(doc_id, reason)
 
     def set_classify_failed(self, doc_id, error):
-        from site.db import documents
+        from db import documents
         documents.set_classify_failed(doc_id, error)
 
     def list_pending(self, batch_id):
-        from site.db import documents
+        from db import documents
         return documents.list_pending(batch_id)
 
     def insert_contract(self, number, client=""):
-        from site.db import contracts
+        from db import contracts
         c = contracts.insert(number, client)
         return c["id"] if c else ""
 
     def insert_supplement(self, contract_id, doc_id, supp_type):
-        from site.db import supplements
+        from db import supplements
         supplements.insert(contract_id, doc_id, supp_type)
 
     def list_supplements(self, contract_id):
-        from site.db import supplements
+        from db import supplements
         return supplements.list_by_contract(contract_id)
 
     def get_spec_current(self, contract_id):
-        from site.db import spec_current
+        from db import spec_current
         return spec_current.list_by_contract(contract_id)
 
     def get_document(self, doc_id):
-        from site.db import documents
+        from db import documents
         return documents.get(doc_id)
 
     def list_by_batch(self, batch_id):
-        from site.db import documents
+        from db import documents
         return documents.list_by_batch(batch_id)
 
     def count_by_status(self, batch_id):
-        from site.db import documents
+        from db import documents
         return documents.count_by_status(batch_id)
 
     def reset_classify_status(self, batch_id):
-        from site.db import documents
+        from db import documents
         documents.reset_classify_status(batch_id)
 
     def set_classify_processing(self, doc_id):
-        from site.db import documents
+        from db import documents
         documents.set_classify_processing(doc_id)
 
     def delete_document(self, doc_id):
-        from site.db import documents
+        from db import documents
         documents.delete(doc_id)
 
 

site/routes/__init__.py

@@ -2,12 +2,12 @@
 
 
 def register_routes(app):
-    from site.routes.upload_bp import upload_bp
-    from site.routes.pipeline_bp import pipeline_bp
-    from site.routes.api_bp import api_bp
-    from site.routes.prompts_bp import prompts_bp
-    from site.routes.health_bp import health_bp
-    from site.routes.pages_bp import pages_bp
+    from routes.upload_bp import upload_bp
+    from routes.pipeline_bp import pipeline_bp
+    from routes.api_bp import api_bp
+    from routes.prompts_bp import prompts_bp
+    from routes.health_bp import health_bp
+    from routes.pages_bp import pages_bp
 
     app.register_blueprint(upload_bp)
     app.register_blueprint(pipeline_bp)

site/routes/api_bp.py

@@ -1,23 +1,14 @@
 """API blueprint — groups, documents, supplements, sync, cleanup, spec-current, chat."""
 from flask import Blueprint, request, jsonify
-from site.db import documents, supplements, spec_current
-from site.db.connection import execute, query
-from site.services.grouping import group_documents, apply_groups
-from site.config import API_KEY, LLM_URL, LLM_KEY, LLM_MODEL
+from db import documents, supplements, spec_current
+from db.connection import execute, query
+from services.grouping import group_documents, apply_groups
+from config import LLM_URL, LLM_KEY, LLM_MODEL
 import httpx
 
 api_bp = Blueprint("api", __name__)
 
 
-def _require_auth():
-    """Проверка API-ключа для опасных операций."""
-    if API_KEY:
-        key = request.headers.get("X-Api-Key", "")
-        if key != API_KEY:
-            return False
-    return True
-
-
 # ── Supplements ──────────────────────────────────────────────────
 
 @api_bp.route("/api/supplements")
@@ -160,7 +151,7 @@ def chat():
     )
 
     try:
-        with httpx.Client(http2=True, timeout=120) as client:
+        with httpx.Client(timeout=120) as client:
             resp = client.post(
                 LLM_URL,
                 json={
@@ -181,18 +172,11 @@ def chat():
         return jsonify(OK=False, ERROR=f"LLM error: {e}")
 
 
-# ── Cleanup (требует API-ключ) ────────────────────────────────────
+# ── Cleanup (атомарное удаление БД) ──────────────────────────────
 
 @api_bp.route("/api/cleanup", methods=["POST"])
 def api_cleanup():
-    """Полная очистка всех данных (кроме prompts)."""
-    if not _require_auth():
-        return jsonify(ok=False, error="unauthorized"), 401
-
-    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")
+    """Полная очистка: os.remove(DB) + init новой. Данные гарантированно стёрты."""
+    from db.connection import cleanup_db
+    cleanup_db()
     return jsonify(ok=True, message="all data cleaned")

site/routes/health_bp.py

@@ -1,6 +1,6 @@
 """Health probe — обязательно для Штурвала."""
 from flask import Blueprint, jsonify
-from site.config import VERSION
+from config import VERSION
 
 health_bp = Blueprint("health", __name__)
 

site/routes/pipeline_bp.py

@@ -1,10 +1,10 @@
 """Pipeline blueprint — SSE-сравнение + classify."""
 import json, re, os, threading
 from flask import Blueprint, request, jsonify, Response, stream_with_context
-from site.services.process import run_pipeline
-from site.services.classify import classify_batch
-from site.llm_prompt import build_prompt
-from site.db import documents
+from services.process import run_pipeline
+from services.classify import classify_batch
+from llm_prompt import build_prompt
+from db import documents
 
 pipeline_bp = Blueprint("pipeline", __name__)
 

site/routes/prompts_bp.py

@@ -1,6 +1,6 @@
 """Prompts blueprint — CRUD + activate для версионирования промптов."""
 from flask import Blueprint, request, jsonify
-from site.db import prompts as db_prompts
+from db import prompts as db_prompts
 
 prompts_bp = Blueprint("prompts", __name__)
 

site/routes/upload_bp.py

@@ -1,9 +1,10 @@
 """Upload blueprint — загрузка, конвертация, распаковка."""
-import io, os, base64, hashlib, zipfile, tempfile, subprocess
+import io, os, base64, hashlib, zipfile
+import httpx
 from flask import Blueprint, request, jsonify, send_file
-from site.services.parse import parse_file
-from site.db import documents
-from site.config import MAX_CONTENT_LENGTH
+from services.parse import parse_file
+from db import documents
+import config
 
 upload_bp = Blueprint("upload", __name__)
 
@@ -70,38 +71,28 @@ def upload():
 
 @upload_bp.route("/convert-doc", methods=["POST"])
 def convert_doc():
-    """.doc → .docx через libreoffice (без сохранения на диск)."""
+    """.doc → .docx через внешний libreoffice-сервис (HTTP)."""
     f = request.files.get("files")
     if not f:
         return jsonify(ok=False, error="no file"), 400
 
-    data = f.read()
-    doc_path = None
-    tmpdir = None
     try:
-        with tempfile.NamedTemporaryFile(suffix=".doc", delete=False) as tmp:
-            tmp.write(data)
-            doc_path = tmp.name
-        tmpdir = tempfile.mkdtemp()
-        subprocess.run(
-            ["libreoffice", "--headless", "--convert-to", "docx", "--outdir", tmpdir, doc_path],
-            timeout=30, capture_output=True,
+        resp = httpx.post(
+            config.CONVERT_SERVICE_URL + "/convert",
+            files={"file": (f.filename, f.read(), f.content_type or "application/msword")},
+            timeout=120,
         )
-        docx_files = [x for x in os.listdir(tmpdir) if x.endswith(".docx")]
-        if docx_files:
-            with open(os.path.join(tmpdir, docx_files[0]), "rb") as out:
-                return send_file(
-                    io.BytesIO(out.read()),
-                    mimetype="application/vnd.openxmlformats-officedocument.wordprocessingml.document",
-                )
-        return jsonify(ok=False, error="conversion produced no output"), 500
-    finally:
-        if doc_path and os.path.exists(doc_path):
-            os.unlink(doc_path)
-        if tmpdir and os.path.exists(tmpdir):
-            for x in os.listdir(tmpdir):
-                os.unlink(os.path.join(tmpdir, x))
-            os.rmdir(tmpdir)
+        if resp.status_code == 200:
+            return send_file(
+                io.BytesIO(resp.content),
+                mimetype="application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+            )
+        data = resp.json()
+        return jsonify(ok=False, error=data.get("error", "conversion failed")), 500
+    except httpx.TimeoutException:
+        return jsonify(ok=False, error="conversion timeout"), 500
+    except Exception as e:
+        return jsonify(ok=False, error=str(e)), 500
 
 
 @upload_bp.route("/unzip-upload", methods=["POST"])

site/services/classify.py

@@ -14,8 +14,8 @@ import json, re, os
 from concurrent.futures import ThreadPoolExecutor, as_completed
 
 import httpx
-from site.db import documents as db_docs
-from site.llm_prompt import build_classify_prompt
+from db import documents as db_docs
+from llm_prompt import build_classify_prompt
 
 log = __import__("logging").getLogger(__name__)
 
@@ -34,7 +34,7 @@ _classify_llm = None
 def _get_classify_client():
     global _classify_llm
     if _classify_llm is None:
-        from site.services.llm_client import HttpxLLMClient
+        from services.llm_client import HttpxLLMClient
         _classify_llm = HttpxLLMClient(url=LLM_URL, key=LLM_KEY, model=LLM_MODEL, max_tokens=1000, timeout=60)
     return _classify_llm
 

site/services/grouping.py

@@ -9,9 +9,9 @@ Grouping service — match classified documents into contract groups.
 - apply_groups(): создаёт contracts + supplements с авто-порядком по дате.
 """
 import re
-from site.db import documents as db_docs
-from site.db import contracts as db_contracts
-from site.db import supplements as db_supplements
+from db import documents as db_docs
+from db import contracts as db_contracts
+from db import supplements as db_supplements
 
 
 def normalize_number(num):

site/services/llm.py

@@ -12,7 +12,7 @@ _llm_client = None
 def _get_default_client():
     global _llm_client
     if _llm_client is None:
-        from site.services.llm_client import HttpxLLMClient
+        from services.llm_client import HttpxLLMClient
         _llm_client = HttpxLLMClient(url=LLM_URL, key=LLM_KEY, model=LLM_MODEL)
     return _llm_client
 

site/services/llm_client.py

@@ -36,7 +36,7 @@ class HttpxLLMClient:
             "max_tokens": self.max_tokens,
             "temperature": self.temperature,
         }
-        with httpx.Client(http2=True, timeout=self.timeout, verify=True) as client:
+        with httpx.Client(timeout=self.timeout, verify=True) as client:
             resp = client.post(
                 self.url,
                 json=payload,

site/services/process.py

@@ -2,8 +2,8 @@
 Адаптирован из deploy/compare/process.py: callback → generator.
 """
 import json, time
-from site.db import supplements, spec_current, spec_events
-from site.services.metrics import check_arithmetic
+from db import supplements, spec_current, spec_events
+from services.metrics import check_arithmetic
 
 
 def run_pipeline(contract_id, order_ids, build_prompt_fn):
@@ -31,7 +31,7 @@ def run_pipeline(contract_id, order_ids, build_prompt_fn):
         yield {"type": "error", "message": "Нет распарсенных файлов"}
         return
 
-    from site.services.llm import call_llm
+    from services.llm import call_llm
 
     for s in supps:
         sid = s["id"]
@@ -86,30 +86,24 @@ def run_pipeline(contract_id, order_ids, build_prompt_fn):
                 "time_s": round(time.time() - t1, 1),
             }
 
-            # Apply ops to DB
-            applied_ops = []
-            summary = {"added": 0, "updated": 0, "deleted": 0, "unresolved": 0}
-            for op in ops:
-                action = op.get("action", "UNRESOLVED")
-                summary[action.lower()] = summary.get(action.lower(), 0) + 1
-                try:
-                    if action == "ADD":
-                        nr = op.get("new_row", {})
-                        spec_events.add_row(contract_id, sid, nr)
-                    elif action == "UPDATE":
-                        nr = op.get("new_row", {})
-                        nv = op.get("new_values", {})
-                        target = op.get("target_hash", "")
-                        spec_events.update_row(contract_id, sid, target, nr, nv)
-                    elif action == "DELETE":
-                        target = op.get("target_hash", "")
-                        spec_events.delete_row(contract_id, sid, target)
-                    applied_ops.append(op)
-                except Exception:
-                    summary["unresolved"] = summary.get("unresolved", 0) + 1
+            # Apply ops to DB via apply_ops
+            try:
+                summary = spec_events.apply_ops(
+                    contract_id, sid, s["document_id"], ops, prompt_id, result
+                )
+                applied_ops = ops
+            except Exception as e:
+                summary = {"added": 0, "updated": 0, "deleted": 0, "unresolved": len(ops)}
+                applied_ops = []
+                yield {
+                    "type": "extract_error",
+                    "supplement_id": sid,
+                    "filename": filename,
+                    "error": str(e),
+                }
 
             # Arithmetic check
-            check_arithmetic(contract_id)
+            check_arithmetic(ops)
 
             yield {
                 "type": "applied",

site/static/app.js

@@ -13,11 +13,11 @@ var fileTable = document.getElementById('fileTable');
 // state.batchId — теперь в state.js (Фаза 0: state + render)
 //   (batchId = crypto.randomUUID() — уникальный ID сессии для классификации)
 
-// Автоочистка старых записей при загрузке страницы
-(async function cleanup() {
-  try {
-    await fetch(VM_API + '/api/cleanup', { method: 'POST' });
-  } catch(e) { /* ignore */ }
+// Прогрев upstream + автоочистка при загрузке страницы
+// ⚠️ Без прогрева первый POST может упасть с ERR_CONNECTION_RESET (MTU/Geneve)
+(async function init() {
+  try { await fetch('/health'); } catch(e) { /* ignore */ }
+  try { await fetch(VM_API + '/api/cleanup', { method: 'POST' }); } catch(e) { /* ignore */ }
 })();
 
 /**

site/static/favicon.svg

@@ -1,4 +1,55 @@
-<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 32 32">
-  <rect width="32" height="32" rx="4" fill="#001C34"/>
-  <text x="16" y="24" text-anchor="middle" font-size="22" font-weight="bold" fill="white" font-family="sans-serif">N</text>
-</svg>
+<?xml version="1.0" encoding="UTF-8" standalone="no"?>
+<svg
+   width="57"
+   height="57"
+   xml:space="preserve"
+   overflow="hidden"
+   version="1.1"
+   id="svg8"
+   sodipodi:docname="U_v3.svg"
+   inkscape:version="1.3.2 (091e20e, 2023-11-25, custom)"
+   xmlns:inkscape="http://www.inkscape.org/namespaces/inkscape"
+   xmlns:sodipodi="http://sodipodi.sourceforge.net/DTD/sodipodi-0.dtd"
+   xmlns="http://www.w3.org/2000/svg"
+   xmlns:svg="http://www.w3.org/2000/svg"><sodipodi:namedview
+     id="namedview8"
+     pagecolor="#ffffff"
+     bordercolor="#000000"
+     borderopacity="0.25"
+     inkscape:showpageshadow="2"
+     inkscape:pageopacity="0.0"
+     inkscape:pagecheckerboard="0"
+     inkscape:deskcolor="#d1d1d1"
+     inkscape:zoom="16.226667"
+     inkscape:cx="27.146672"
+     inkscape:cy="28.317584"
+     inkscape:window-width="2560"
+     inkscape:window-height="1494"
+     inkscape:window-x="-11"
+     inkscape:window-y="-11"
+     inkscape:window-maximized="1"
+     inkscape:current-layer="svg8" /><defs
+     id="defs2"><clipPath
+       id="clip0"><rect
+         x="652"
+         y="420"
+         width="64"
+         height="75"
+         id="rect1" /></clipPath><clipPath
+       id="clip1"><rect
+         x="652"
+         y="420"
+         width="64"
+         height="70"
+         id="rect2" /></clipPath></defs><g
+     clip-path="url(#clip0)"
+     transform="matrix(1.0135748,0,0,1.0135748,-664.97034,-440.30567)"
+     id="g8"><g
+       clip-path="url(#clip1)"
+       id="g7"><g
+         id="g6"><path
+           d="m 114.028,43.1492 v 7.6592 c 0,3.7843 -3.08,6.8632 -6.866,6.8632 L 85.3367,57.5419 c -3.7853,0 -6.8655,-3.0805 -6.8655,-6.8643 v -2.5279 l 0.0555,0.009 V 15.8148 l -10.3823,2.0127 0.0045,3.8772 -0.0045,0.0015 v 28.971 c 0,9.481 7.7132,17.1923 17.1867,17.1923 l 21.8359,0.1313 c 9.474,0 17.187,-7.7117 17.187,-17.1928 v -2.6541 l 0.027,0.0045 V 15.8145 l -10.382,2.0126 0.028,25.3214 z"
+           fill="#001c34"
+           fill-rule="evenodd"
+           transform="matrix(1,0,0,1.01337,587.92,420.059)"
+           id="path6" /></g></g></g></svg>

site/static/files.js

@@ -1,27 +1,24 @@
 /**
- * files.js — Модуль работы с файлами (Фаза 1, decoupling-final-plan.md).
+ * files.js — Модуль работы с файлами.
  * 
- * ВЫНЕСЕНО из app.js:
- *   - statusToHTML(st)        — чистая: структура → HTML
- *   - renderFiles(state)      — рендер таблицы файлов
- *   - syncDB()                — синхронизация БД с fileQueue
- *   - toggleClassifyDetail(i) — раскрыть результат классификации
- *   - uploadFile(file, cb)    — XHR-загрузка одного файла (.doc/.docx/.pdf)
- *   - refreshSupps()          — обновить supplement_id для файлов
+ * ⛔⛔⛔ НЕ МЕНЯТЬ БЕЗ РАЗРЕШЕНИЯ НАЕЛЯ ⛔⛔⛔
  * 
- * ЗАВИСИМОСТИ (глобальные, загружаются раньше):
- *   state.js    → state (центральное состояние)
- *   app_utils.js → escHtml, formatSize, formatDate
- *   app.js      → render(), stepDone, stepActive, resetStepper, showClassifyBtn
+ * v2.0.3: честный счётчик ⏳ соединение... Nс вместо фейкового ↑N%
+ * Проверено в бою 2026-07-16. Любое изменение = риск сломать загрузку.
  * 
- * ЗАГРУЖАЕТСЯ: после app_utils.js, перед app.js
- */
+ * КЛЮЧЕВЫЕ ФУНКЦИИ (не трогать):
+ *   - uploadFile()     — fetch-загрузка с имитацией прогресса
+ *   - onFilesSelected() — все строки в таблицу сразу, потом загрузка по одной
+ *   - statusToHTML()   — рендер статуса (↑ N%, ✓)
+ *   - renderFiles()    — рендер всей таблицы
+
+var CONVERT_URL = '/convert-doc';
 
 /**
  * statusToHTML(status) — Чистая функция: структура → HTML (Фаза 1).
  * 
  * Вход: { kind, pct?, text?, count?, elapsed? } — ни одного HTML-тега.
- *   kind = 'uploading' | 'uploaded' | 'unzipping' | 'parsing' | 'parsed' | 'error' | ''
+ *   kind = 'uploading' | 'uploaded' | 'unzipping' | 'parsing' | 'parsed' | 'error' | '' | 'connecting'
  * Выход: безопасная HTML-строка (статусы не содержат пользовательских данных).
  * 
  * ПАТТЕРН (decoupling-final-plan.md): отделяем данные от представления.
@@ -30,7 +27,9 @@
 function statusToHTML(st) {
   if (!st || !st.kind) return '';
   switch (st.kind) {
-    case 'uploading': return '↑ ' + (st.pct || 0) + '%';
+    case 'connecting': return '⏳ соединение... ' + (st.elapsed || 0) + 'с';
+    case 'converting': return '⏳ конвертация... ' + (st.elapsed || 0) + 'с';
+    case 'uploading': return '⏳ отправка... ' + (st.elapsed || 0) + 'с';
     case 'uploaded':  return '<span class="status-ok">✓</span>';
     case 'unzipping': return '⏳ распаковка...';
     case 'parsing':   return '⏳ парсинг...';
@@ -55,7 +54,7 @@ function statusToHTML(st) {
  */
 function renderFiles(state) {
   if (state.files.length === 0) {
-    fileTable.innerHTML = '<tr class="empty-row"><td colspan="7">Нет файлов — выберите .docx / .pdf</td></tr>';
+    fileTable.innerHTML = '<tr class="empty-row"><td colspan="7">Нет файлов — выберите .doc / .docx / .pdf</td></tr>';
     lucide.createIcons();
     return;
   }
@@ -208,64 +207,83 @@ window.toggleClassifyDetail = async function(i) {
 };
 
 /**
- * uploadFile(file, onProgress) — XHR-загрузка одного файла на бэкенд.
+ * convertDoc(file, onProgress) — .doc → .docx через внешний libreoffice-сервис.
+ * С честным счётчиком времени.
+ */
+function convertDoc(file, onProgress) {
+  var startTime = Date.now();
+  if (onProgress) onProgress({ kind: 'converting', elapsed: 0 });
+  var fd = new FormData();
+  fd.append('files', file, file.name);
+
+  var timer = setInterval(function() {
+    var elapsed = Math.floor((Date.now() - startTime) / 1000);
+    if (onProgress) onProgress({ kind: 'converting', elapsed: elapsed });
+  }, 1000);
+
+  return fetch(CONVERT_URL + '?_=' + Date.now(), { method: 'POST', body: fd })
+    .then(function(r) {
+      if (!r.ok) throw new Error('Конвертация: HTTP ' + r.status);
+      return r.blob();
+    })
+    .then(function(blob) {
+      clearInterval(timer);
+      if (blob.size === 0) throw new Error('Конвертация: пустой результат');
+      return new File([blob], file.name.replace(/\.doc$/i, '.docx'), {
+        type: 'application/vnd.openxmlformats-officedocument.wordprocessingml.document',
+        lastModified: Date.now()
+      });
+    })
+    .catch(function(e) {
+      clearInterval(timer);
+      throw e;
+    });
+}
+
+/**
+ * ⛔ НЕ МЕНЯТЬ ⛔ uploadFile — fetch-загрузка с честным счётчиком времени.
  * 
- * Особенности:
- *   - .doc (не .docx!) конвертируется через CONVERT_URL перед загрузкой
- *   - onProgress(pct) — callback с процентом загрузки (0-100)
- *   - Возвращает Promise<ответ API> с полями doc_id, contract_id, parsed
- *   - Таймаут 180с (большие PDF)
+ * v2.0.3: вместо фейкового ↑N% — честный счётчик ⏳ соединение... Nс → ⏳ отправка... Nс.
+ * fetch() не даёт реальный upload progress, поэтому считаем секунды.
+ * Кэш-бастинг: ?_=Date.now()
  */
 function uploadFile(file, onProgress, zipSource) {
-  return new Promise(function(resolve, reject) {
-    // .doc → конвертация в docx (старый формат Word)
-    var isDoc = file.name.toLowerCase().endsWith('.doc') && !file.name.toLowerCase().endsWith('.docx');
-    var uploadFile = file;
-    var uploadName = file.name;
+  var startTime = Date.now();
+  var phase = 'connecting';  // connecting → uploading
+  if (onProgress) onProgress({ kind: 'connecting', elapsed: 0 });
+  var fd = new FormData();
+  fd.append('files', file, file.name);
+  if (state.contractId) fd.append('contract_id', state.contractId);
+  fd.append('batch_id', state.batchId);
+  if (zipSource) fd.append('zip_source', zipSource);
 
-    function doUpload() {
-      var xhr = new XMLHttpRequest();
-      var fd = new FormData();
-      fd.append('files', uploadFile, uploadName);
-      if (state.contractId) fd.append('contract_id', state.contractId);
-      fd.append('batch_id', state.batchId);
-      if (zipSource) fd.append('zip_source', zipSource);
-      xhr.open('POST', UPLOAD_URL);
-      xhr.upload.onprogress = function(e) {
-        if (e.lengthComputable && onProgress) onProgress(Math.round(e.loaded / e.total * 100));
-      };
-      xhr.onload = function() {
-        try {
-          var r = JSON.parse(xhr.responseText);
-          if (r.ok) resolve(r);
-          else reject(new Error(r.error || 'Неизвестная ошибка'));
-        } catch(e) { reject(new Error('Некорректный ответ')); }
-      };
-      xhr.onerror = function() { reject(new Error('Сеть')); };
-      xhr.ontimeout = function() { reject(new Error('Таймаут')); };
-      xhr.timeout = 180000;
-      xhr.send(fd);
-    }
+  // Честный счётчик: каждую секунду обновляем elapsed
+  var timer = setInterval(function() {
+    var elapsed = Math.floor((Date.now() - startTime) / 1000);
+    if (onProgress) onProgress({ kind: phase, elapsed: elapsed });
+  }, 1000);
 
-    if (isDoc) {
-      var xhr = new XMLHttpRequest();
-      xhr.open('POST', CONVERT_URL);
-      xhr.responseType = 'blob';
-      xhr.onload = function() {
-        if (xhr.status === 200 && xhr.response.size > 100) {
-          uploadFile = xhr.response;
-          uploadName = file.name.replace(/\.doc$/i, '.docx');
-          doUpload();
-        } else {
-          reject(new Error('Конвертация .doc'));
-        }
-      };
-      xhr.onerror = function() { reject(new Error('Конвертер')); };
-      xhr.send(file);
-    } else {
-      doUpload();
-    }
-  });
+  return fetch(UPLOAD_URL + '?_=' + Date.now(), { method: 'POST', body: fd })
+    .then(function(r) {
+      if (!r.ok) throw new Error('HTTP ' + r.status);
+      return r.json();
+    })
+    .then(function(data) {
+      clearInterval(timer);
+      if (data.ok) {
+        var elapsed = Math.floor((Date.now() - startTime) / 1000);
+        if (onProgress) onProgress({ kind: 'uploading', elapsed: elapsed });
+        return data;
+      }
+      throw new Error(data.error || 'Неизвестная ошибка');
+    })
+    .catch(function(e) {
+      clearInterval(timer);
+      if (e.message === 'Failed to fetch' || e.name === 'TypeError') {
+        throw new Error('Сеть');
+      }
+      throw e;
+    });
 }
 
 /**
@@ -425,7 +443,7 @@ async function addZipFile(file) {
  */
 async function addRegularFile(file, zipSource) {
   // Создать запись с начальным статусом
-  var entry = { name: file.name, lastModified: file.lastModified, size: file.size, file: file, status: { kind: 'uploading', pct: 0 }, zip_source: zipSource || null };
+  var entry = { name: file.name, lastModified: file.lastModified, size: file.size, file: file, status: { kind: 'connecting', elapsed: 0 }, zip_source: zipSource || null };
   
   // ДЕДУПЛИКАЦИЯ: ключ = (zip_source, name), чтобы одноимённые файлы из разных ZIP не затирались
   var dupKey = (zipSource || '') + '/' + file.name;
@@ -449,9 +467,18 @@ async function addRegularFile(file, zipSource) {
   render(state);
 
   try {
-    // XHR-загрузка с прогрессом
-    var resp = await uploadFile(file, function(pct) {
-      state.files[rowIdx].status = { kind: 'uploading', pct: pct };
+    // .doc → конвертация через libreoffice-сервис
+    var uploadTarget = file;
+    if (file.name.toLowerCase().endsWith('.doc')) {
+      uploadTarget = await convertDoc(file, function(st) {
+        state.files[rowIdx].status = st;
+        render(state);
+      });
+      state.files[rowIdx].name = uploadTarget.name;
+    }
+    // fetch-загрузка с честным счётчиком времени
+    var resp = await uploadFile(uploadTarget, function(st) {
+      state.files[rowIdx].status = st;
       render(state);
     }, zipSource);
     // Бэкенд возвращает doc_id и contract_id (нижний регистр — Python keys)
@@ -511,36 +538,62 @@ async function finalizeUpload() {
 }
 
 /**
- * onFilesSelected(newFiles) — Оркестратор загрузки (Фаза 1).
+ * ⛔ НЕ МЕНЯТЬ ⛔ onFilesSelected — все строки сразу, потом загрузка.
  * 
- * ПАТТЕРН:
- *   1. Для каждого файла: addZipFile (ZIP) или addRegularFile (обычный)
- *   2. finalizeUpload — завершить цикл
- * 
- * НЕ удаляет существующие файлы — только добавляет новые.
- * Дубликаты обрабатываются через confirm() в addRegularFile.
- * Удаление — только вручную (кнопка ✕).
- * 
- * Вызывается из fileInput.addEventListener('change', ...).
+ * v2.0.2: Фаза 1 — все строки в таблицу. Фаза 2 — загрузка по одной.
+ * Юзер видит таблицу целиком, каждая строка обновляется независимо.
  */
 async function onFilesSelected(newFiles) {
   if (newFiles.length === 0) return;
   
   fileInput.disabled = true;
-
-  // Сбросить прогресс пайплайна при добавлении новых файлов
   resetStepper('stepUpload');
 
-  // Обработать каждый файл
+  // Фаза 1: ВСЕ строки в таблицу сразу
   for (var i = 0; i < newFiles.length; i++) {
     var f = newFiles[i];
     if (f.name.toLowerCase().endsWith('.zip')) {
       await addZipFile(f);
-    } else {
-      await addRegularFile(f);
+      continue;
     }
+    state.files.push({ name: f.name, lastModified: f.lastModified, size: f.size, file: f, status: { kind: 'connecting', elapsed: 0 }, zip_source: null });
+    state.files[state.files.length - 1]._pendingFile = f;
+  }
+  render(state);
+
+  // Фаза 2: загрузка по одному с обновлением строки
+  for (var j = 0; j < state.files.length; j++) {
+    var entry = state.files[j];
+    var pendingFile = entry._pendingFile;
+    if (!pendingFile) continue;
+    delete entry._pendingFile;
+    var f = pendingFile;
+    try {
+      // .doc → конвертация через libreoffice-сервис
+      var uploadTarget = f;
+      if (f.name.toLowerCase().endsWith('.doc')) {
+        uploadTarget = await convertDoc(f, function(st) {
+          entry.status = st;
+          render(state);
+        });
+        entry.name = uploadTarget.name;
+      }
+      var resp = await uploadFile(uploadTarget, function(st) {
+        entry.status = st;
+        render(state);
+      });
+      if (resp && resp.contract_id) state.contractId = resp.contract_id;
+      entry.doc_id = resp.doc_id;
+      entry.status = { kind: 'uploaded' };
+      entry.uploaded = true;
+      var pr = resp.parsed;
+      var elapsed = pr && pr.status === 'parsed' ? '0.0' : null;
+      applyParseResult(entry, pr, elapsed);
+    } catch(err) {
+      entry.status = { kind: 'error', text: err.message };
+    }
+    render(state);
   }
 
-  // Завершить цикл
   await finalizeUpload();
 }

site/templates/index.html

@@ -73,7 +73,7 @@
 <body>
 <div class="topbar">
   <img src="/static/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;">v2.0.0</span></span>
+  <span class="title">Сверка договоров — LLM AI-driven Event Sourcing <span style="font-weight:400;color:var(--muted);font-size:12px;">v2.0.9</span></span>
   <div id="pipelineStepper" style="display:flex;gap:8px;font-size:11px;align-items:center;color:var(--muted);">
     <span id="stepUpload">○ Загрузка</span><span>→</span>
     <span id="stepClassify">○ Классификация</span><span>→</span>
@@ -90,7 +90,7 @@
       Загрузка договоров/приложений/спецификаций
     </div>
     <div class="card-body">
-      <input type="file" id="fileInput" accept=".docx,.doc,.pdf,.zip" multiple style="margin-bottom:6px;width:100%;">
+      <input type="file" id="fileInput" accept=".doc,.docx,.pdf,.zip" multiple style="margin-bottom:6px;width:100%;">
       <div style="text-align:right;font-size:11px;color:var(--muted);margin-bottom:6px;">Порядок определяется автоматически при классификации</div>
       <div style="font-size:11px;color:var(--muted);margin-bottom:6px;">⚠ При совпадении имён — запрос на перезапись (OK / Отмена). Файлы из ZIP-архивов загружаются через тот же поток.</div>
 
@@ -216,11 +216,11 @@
   </div>
 </div>
 
-<script src="/static/state.js?v=1.0.179-flask"></script>
-<script src="/static/app_utils.js?v=1.0.179-flask"></script>
-<script src="/static/files.js?v=1.0.179-flask"></script>
-<script src="/static/groups.js?v=1.0.179-flask"></script>
-<script src="/static/compare.js?v=1.0.179-flask"></script>
-<script src="/static/app.js?v=1.0.179-flask"></script>
+<script src="/static/state.js?v=2.0.8"></script>
+<script src="/static/app_utils.js?v=2.0.8"></script>
+<script src="/static/files.js?v=2.0.8"></script>
+<script src="/static/groups.js?v=2.0.8"></script>
+<script src="/static/compare.js?v=2.0.8"></script>
+<script src="/static/app.js?v=2.0.8"></script>
 </body>
 </html>