feat: adds crawl resource cache

Merge pull request 'feat: add dataset checkpoint sync for MCP' (#23 ) from feature/dataset-version-sync into main
Reviewed-on: #23
2026-05-14 12:21:44 +03:00 · 2026-05-14 08:01:26 +00:00 · 2026-05-14 11:00:46 +03:00 · 2026-05-13 13:12:06 +00:00 · 2026-05-13 16:11:13 +03:00 · 2026-05-13 09:46:48 +00:00
36 changed files with 2057 additions and 462 deletions
--- a/.env.example
+++ b/.env.example
@@ -14,13 +14,5 @@ PARSER_USE_PLAYWRIGHT=false
 ADMIN_USERNAME=admin
 ADMIN_PASSWORD=change-me
 SESSION_SECRET=change-me-session-secret
-MCP_TOKEN=change-me-mcp-token
-MCP_AUTH_MODE=oauth
-MCP_RESOURCE_URL=http://localhost:8001/mcp
-MCP_OAUTH_ISSUER=
-MCP_OAUTH_AUDIENCE=
-MCP_OAUTH_JWKS_URL=
-MCP_OAUTH_REQUIRED_SCOPE=mcp:tools
-
 API_PORT=8000
 MCP_PORT=8001
--- a/MCP_DESCRIPTION.md
+++ b/MCP_DESCRIPTION.md
@@ -0,0 +1,592 @@
+# MCP: описание работы, структуры и тулзов
+
+Документ описывает MCP endpoint сервиса `miem-employees` по текущей реализации в `app/mcp.py`.
+
+## Где находится MCP
+
+- FastAPI router: `app.mcp.router`
+- Подключение к приложению: `app/main.py`
+- HTTP endpoint: `POST /mcp`
+- Локально при обычном запуске API: `http://localhost:8000/mcp`
+- В Docker Compose через отдельный сервис `mcp`: `http://localhost:8001/mcp`
+- Авторизация на уровне приложения: отсутствует. Заголовок `Authorization` не проверяется и не влияет на ответ.
+
+Если доступ к MCP нужно ограничить, это должно делаться внешним контуром: bind на localhost, VPN, firewall, reverse proxy или отдельная сетевая политика.
+
+## Протокол
+
+Endpoint принимает JSON-RPC 2.0 over HTTP.
+
+Общий формат запроса:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "method": "tools/list",
+  "params": {}
+}
+```
+
+Общий формат успешного ответа:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "result": {}
+}
+```
+
+Общий формат ошибки:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "error": {
+    "code": -32601,
+    "message": "Method not found"
+  }
+}
+```
+
+Поддерживаемая версия MCP-протокола:
+
+```text
+2024-11-05
+```
+
+Имя сервиса:
+
+```text
+miem-employees
+```
+
+Версия сервера берется из `app.version.BACKEND_VERSION`.
+
+## Поддерживаемые JSON-RPC методы
+
+### initialize
+
+Возвращает метаданные MCP-сервера и capabilities.
+
+Запрос:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "method": "initialize",
+  "params": {}
+}
+```
+
+Ответ:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "result": {
+    "protocolVersion": "2024-11-05",
+    "serverInfo": {
+      "name": "miem-employees",
+      "version": "0.5.0"
+    },
+    "capabilities": {
+      "tools": {}
+    }
+  }
+}
+```
+
+### tools/list
+
+Возвращает список доступных tools с JSON Schema для аргументов.
+
+Запрос:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "method": "tools/list",
+  "params": {}
+}
+```
+
+Ответ содержит массив `result.tools`.
+
+### tools/call
+
+Вызывает один tool по имени.
+
+Запрос:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "method": "tools/call",
+  "params": {
+    "name": "search_employees",
+    "arguments": {
+      "query": "Сергеев",
+      "limit": 20
+    }
+  }
+}
+```
+
+Ответ tool всегда заворачивается в MCP content-массив:
+
+```json
+{
+  "jsonrpc": "2.0",
+  "id": 1,
+  "result": {
+    "content": [
+      {
+        "type": "text",
+        "text": "{\"items\":[]}"
+      }
+    ]
+  }
+}
+```
+
+Поле `text` содержит сериализованный JSON с `ensure_ascii=false`. Клиент должен распарсить это поле как JSON, если ему нужна структурированная нагрузка.
+
+## Ошибки
+
+- Неизвестный JSON-RPC метод: `code = -32601`, `message = "Method not found"`.
+- Исключения при обработке tool: `code = -32000`, `message` содержит текст исключения.
+- Если сущность не найдена внутри отдельных tools, HTTP и JSON-RPC ответ остаются успешными, а полезная нагрузка содержит `{"error": "not_found"}`.
+
+## Источники данных
+
+MCP читает данные из основной базы через SQLAlchemy session из `app.db.get_db`.
+
+Основные таблицы и модели:
+
+- `employees`: текущая карточка сотрудника, статус, профиль, `current_data`, checksum.
+- `crawl_runs`: история запусков парсинга.
+- `crawl_run_employee_changes`: детальные изменения сотрудников в рамках запуска.
+- `crawl_errors`: ошибки парсинга в рамках запуска.
+- `dataset_versions`: версии полного набора сотрудников.
+- `dataset_version_items`: состав конкретной версии набора сотрудников.
+
+## Общая структура employee payload
+
+Краткая карточка сотрудника:
+
+```json
+{
+  "profile_key": "staff:avsergeev",
+  "profile_id": "avsergeev",
+  "full_name": "Сергеев Алексей Викторович",
+  "status": "active",
+  "canonical_url": "https://www.hse.ru/staff/avsergeev",
+  "last_seen_at": "2026-05-14T10:00:00+00:00",
+  "dismissed_at": null
+}
+```
+
+В sync payload дополнительно отдается `checksum`.
+
+Полная карточка дополнительно содержит:
+
+```json
+{
+  "data": {
+    "contacts": {},
+    "sections": []
+  }
+}
+```
+
+`data` соответствует распарсенному JSON профиля сотрудника. Внутри `sections` могут быть секции с публикациями, курсами, ВКР, таблицами, ссылками и произвольными текстовыми блоками.
+
+## Tools
+
+### get_service_info
+
+Назначение: вернуть метаданные сервиса, список tools и текущую версию набора сотрудников.
+
+Аргументы: отсутствуют.
+
+Возвращает:
+
+```json
+{
+  "service_name": "miem-employees",
+  "backend_version": "0.5.0",
+  "protocolVersion": "2024-11-05",
+  "tools": [],
+  "dataset": {
+    "hash": "sha256",
+    "previous_hash": "sha256 или null",
+    "created_at": "2026-05-14T10:00:00+00:00",
+    "crawl_run_id": 123,
+    "employee_count": 100,
+    "active_count": 95,
+    "dismissed_count": 5
+  }
+}
+```
+
+Особенность: перед ответом сервис создает актуальную `dataset_version`, если текущий набор сотрудников еще не имеет версии.
+
+### sync_employees
+
+Назначение: синхронизировать клиентский кэш сотрудников по hash набора данных.
+
+Аргументы:
+
+```json
+{
+  "client_hash": "sha256 или null",
+  "include_data": true
+}
+```
+
+- `client_hash`: hash версии, которая уже есть у клиента. Если не передан, отдается полный snapshot.
+- `include_data`: управляет включением полного `data` в карточки сотрудников. По умолчанию `true`.
+
+Полный ответ без `client_hash`:
+
+```json
+{
+  "mode": "full",
+  "from_hash": null,
+  "to_hash": "current-sha256",
+  "dataset": {},
+  "items": []
+}
+```
+
+Если клиентский hash совпадает с текущим:
+
+```json
+{
+  "mode": "delta",
+  "from_hash": "current-sha256",
+  "to_hash": "current-sha256",
+  "dataset": {},
+  "changes": {
+    "added": [],
+    "updated": [],
+    "dismissed": [],
+    "removed": []
+  }
+}
+```
+
+Если `client_hash` неизвестен серверу:
+
+```json
+{
+  "mode": "full",
+  "from_hash": "missing",
+  "to_hash": "current-sha256",
+  "dataset": {},
+  "items": [],
+  "reason": "unknown_client_hash"
+}
+```
+
+Если `client_hash` найден и отличается от текущего:
+
+```json
+{
+  "mode": "delta",
+  "from_hash": "old-sha256",
+  "to_hash": "current-sha256",
+  "dataset": {},
+  "changes": {
+    "added": [],
+    "updated": [],
+    "dismissed": [],
+    "removed": []
+  }
+}
+```
+
+Логика delta:
+
+- `added`: сотрудник появился в новой версии.
+- `updated`: изменился checksum или статус, и сотрудник активен.
+- `dismissed`: сотрудник есть в новой версии, но получил статус `dismissed`.
+- `removed`: `profile_key` был в старой версии, но отсутствует в новой.
+
+Hash набора считается по отсортированному списку `{profile_key, status, checksum}`.
+
+### search_employees
+
+Назначение: найти сотрудников по ФИО или canonical URL.
+
+Аргументы:
+
+```json
+{
+  "query": "Сергеев",
+  "status": "active",
+  "limit": 20
+}
+```
+
+- `query`: обязательный по schema, но в коде пустая строка означает поиск без текстового фильтра.
+- `status`: опционально, только `active` или `dismissed`.
+- `limit`: максимум 100, по умолчанию 20.
+
+Возвращает массив кратких employee payload без `data`:
+
+```json
+[
+  {
+    "profile_key": "staff:avsergeev",
+    "profile_id": "avsergeev",
+    "full_name": "Сергеев Алексей Викторович",
+    "status": "active",
+    "canonical_url": "https://www.hse.ru/staff/avsergeev",
+    "last_seen_at": "2026-05-14T10:00:00+00:00",
+    "dismissed_at": null
+  }
+]
+```
+
+### get_employee
+
+Назначение: получить одну карточку сотрудника.
+
+Аргументы:
+
+```json
+{
+  "profile_id_or_url": "avsergeev"
+}
+```
+
+Поиск выполняется по:
+
+- `profile_key`
+- `profile_id`
+- точному `canonical_url`
+- частичному совпадению `canonical_url`
+
+Возвращает полный employee payload с `data`.
+
+Если сотрудник не найден:
+
+```json
+{
+  "error": "not_found"
+}
+```
+
+### list_employee_publications
+
+Назначение: вернуть публикации сотрудника из распарсенных секций профиля.
+
+Аргументы:
+
+```json
+{
+  "profile_id_or_url": "avsergeev"
+}
+```
+
+Сервис ищет секции `current_data.sections` с `type = "publications"` и объединяет массивы `publications`.
+
+Ответ:
+
+```json
+{
+  "employee": {},
+  "items": [
+    {
+      "title": "Название публикации",
+      "text": "Полное описание",
+      "url": "https://..."
+    }
+  ]
+}
+```
+
+Если сотрудник или данные профиля отсутствуют:
+
+```json
+{
+  "items": []
+}
+```
+
+### list_employee_courses
+
+Назначение: вернуть курсы преподавания сотрудника из распарсенных секций профиля.
+
+Аргументы:
+
+```json
+{
+  "profile_id_or_url": "avsergeev"
+}
+```
+
+Сервис ищет секции `current_data.sections` с `type = "courses_by_year"` и объединяет массивы `courses`.
+
+Ответ:
+
+```json
+{
+  "employee": {},
+  "items": [
+    {
+      "title": "Название курса",
+      "url": "https://..."
+    }
+  ]
+}
+```
+
+Если сотрудник или данные профиля отсутствуют:
+
+```json
+{
+  "items": []
+}
+```
+
+### get_crawl_status
+
+Назначение: вернуть последний запуск парсинга.
+
+Аргументы: отсутствуют.
+
+Ответ:
+
+```json
+{
+  "id": 123,
+  "status": "completed",
+  "source_url": "https://miem.hse.ru/persons",
+  "started_at": "2026-05-14T10:00:00+00:00",
+  "finished_at": "2026-05-14T10:10:00+00:00",
+  "found_count": 100,
+  "parsed_count": 98,
+  "error_count": 2,
+  "dismissed_count": 1
+}
+```
+
+Если запусков еще не было:
+
+```json
+{
+  "status": "never_run"
+}
+```
+
+### get_crawl_run_details
+
+Назначение: вернуть детальную информацию по конкретному запуску парсинга: summary, изменения сотрудников и ошибки.
+
+Аргументы:
+
+```json
+{
+  "run_id": 123
+}
+```
+
+Ответ:
+
+```json
+{
+  "id": 123,
+  "source_url": "https://miem.hse.ru/persons",
+  "status": "completed",
+  "status_display": "Завершен",
+  "started_at": "2026-05-14T10:00:00+00:00",
+  "finished_at": "2026-05-14T10:10:00+00:00",
+  "started_display": "14.05.2026 13:00",
+  "finished_display": "14.05.2026 13:10",
+  "found_count": 100,
+  "parsed_count": 98,
+  "new_count": 3,
+  "error_count": 2,
+  "dismissed_count": 1,
+  "processed_count": 100,
+  "progress_percent": 100.0,
+  "message": null,
+  "changes_detail_available": true,
+  "changes": {
+    "new": [],
+    "missing_from_source": [],
+    "dismissed": []
+  },
+  "errors": []
+}
+```
+
+Если запуск не найден:
+
+```json
+{
+  "error": "not_found"
+}
+```
+
+## Примеры curl
+
+Список tools:
+
+```bash
+curl http://localhost:8001/mcp \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'
+```
+
+Поиск сотрудника:
+
+```bash
+curl http://localhost:8001/mcp \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"search_employees","arguments":{"query":"Сергеев","limit":5}}}'
+```
+
+Полная синхронизация:
+
+```bash
+curl http://localhost:8001/mcp \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"sync_employees","arguments":{"include_data":false}}}'
+```
+
+Delta-синхронизация:
+
+```bash
+curl http://localhost:8001/mcp \
+  -H "Content-Type: application/json" \
+  -d '{"jsonrpc":"2.0","id":4,"method":"tools/call","params":{"name":"sync_employees","arguments":{"client_hash":"known-sha256","include_data":true}}}'
+```
+
+## Как MCP используется клиентом
+
+1. Клиент вызывает `initialize` и проверяет `protocolVersion`.
+2. Клиент вызывает `tools/list`, чтобы получить актуальный список tools и input schemas.
+3. Для поиска и точечных запросов клиент вызывает `tools/call` с `search_employees`, `get_employee`, `list_employee_publications`, `list_employee_courses`, `get_crawl_status` или `get_crawl_run_details`.
+4. Для локального кэша клиент вызывает `get_service_info` или `sync_employees`.
+5. Клиент хранит последний `dataset.hash`.
+6. При следующей синхронизации клиент передает hash как `client_hash`.
+7. Сервер возвращает пустую delta, delta с изменениями или полный snapshot, если hash неизвестен.
+
+## Важные особенности реализации
+
+- MCP endpoint read-only: tools не запускают парсинг и не меняют сотрудников напрямую.
+- `get_service_info` и `sync_employees` могут создать новую запись `dataset_versions`, если состояние сотрудников изменилось и новой версии еще нет.
+- Все tool payloads возвращаются как JSON-строка внутри `content[0].text`.
+- `search_employees` ищет через `ilike` по `full_name` и `canonical_url`.
+- `get_employee` допускает частичный URL, поэтому строка `133709486` может найти `https://www.hse.ru/org/persons/133709486`.
+- Временные значения сериализуются через `isoformat()`, display-поля для админских payload формируются в часовом поясе `Europe/Moscow`.
--- a/README.md
+++ b/README.md
@@ -6,7 +6,7 @@

 - `api`: FastAPI, REST API, HTML-админка, healthcheck.
 - `worker`: weekly scheduler, который запускает парсинг по `CRAWL_CRON`.
- `mcp`: HTTP MCP endpoint с OAuth/OIDC access token для внешних агентов или legacy static token для локального режима.
+- `mcp`: открытый HTTP MCP endpoint для ИИ-агентов.
 - `postgres`: основная БД.

 Парсер использует фиксированный источник сотрудников, по умолчанию `https://miem.hse.ru/persons`. Для каждой карточки сохраняются ФИО, должности, год начала работы, контакты, идентификаторы, вкладки профиля, секции, публикации, курсы, ВКР, JSON-снапшот и сжатый HTML-снапшот. Ссылки обходятся только из меню профиля самого сотрудника (`person-menu`), например `#sci`, `#teaching`, `#main`.
@@ -27,13 +27,6 @@ cp .env.example .env
 - `CRAWL_LIMIT`: опциональный лимит профилей для тестового запуска.
 - `ADMIN_USERNAME`, `ADMIN_PASSWORD`: логин и пароль админки.
 - `SESSION_SECRET`: секрет подписи cookie.
- `MCP_TOKEN`: статический bearer token для legacy/local режима `MCP_AUTH_MODE=token`.
- `MCP_AUTH_MODE`: режим авторизации MCP: `oauth` для внешних агентов или `token` для локальной отладки.
- `MCP_RESOURCE_URL`: публичный URL MCP endpoint, например `https://example.com/mcp`.
- `MCP_OAUTH_ISSUER`: issuer внешнего OIDC-провайдера.
- `MCP_OAUTH_AUDIENCE`: ожидаемый `aud` в OAuth access token.
- `MCP_OAUTH_JWKS_URL`: JWKS endpoint; если не задан, используется `<issuer>/.well-known/jwks.json`.
- `MCP_OAUTH_REQUIRED_SCOPE`: scope для доступа к MCP tools, по умолчанию `mcp:tools`.
 - `PARSER_USE_PLAYWRIGHT`: включение Playwright-рендера динамических вкладок.

 ## Локальный запуск
@@ -51,7 +44,6 @@ uvicorn app.main:app --reload

 - `Dashboard`: общая статистика, последний добавленный сотрудник, прогресс текущего/последнего парсинга и ручной запуск.
 - `Directory`: настраиваемая таблица сотрудников с фильтрами, сортировкой, пагинацией и выбором колонок.
- `Employees`: простая legacy-таблица сотрудников.
 - `Runs`: история запусков, ошибки и progress bar.

 ## Docker Compose
@@ -82,45 +74,37 @@ curl -X POST http://localhost:8000/api/crawl-runs --cookie "miem_admin_session=.
 - новые сотрудники добавляются в `employees`;
 - количество новых сотрудников за запуск сохраняется в `crawl_runs.new_count`;
 - активные сотрудники, исчезнувшие из текущего списка источника, получают статус `dismissed` и `dismissed_at`;
- каждый успешный разбор сохраняет запись в `employee_snapshots`.
+- каждый успешный новый или измененный разбор сохраняет запись в `employee_snapshots`;
+- неизмененные профили учитываются в `crawl_runs.skipped_count` и не получают новый snapshot.

-Во время выполнения парсинга `found_count`, `parsed_count` и `error_count` обновляются в базе. Админка опрашивает `/api/crawl-runs/latest` и показывает прогресс как `parsed_count + error_count / found_count`.
+Во время выполнения парсинга `found_count`, `parsed_count`, `skipped_count` и `error_count` обновляются в базе. Админка опрашивает `/api/crawl-runs/latest` и показывает прогресс как `(parsed_count + skipped_count + error_count) / found_count`.

 ## MCP

-Endpoint: `POST /mcp`, авторизация `Authorization: Bearer <token>`.
-
-Для внешних ИИ-агентов используйте `MCP_AUTH_MODE=oauth`. В этом режиме статический `MCP_TOKEN` не принимается: клиент должен передать OAuth/OIDC access token с нужным scope.
+Endpoint: `POST /mcp`, без авторизации на уровне приложения.

 Поддерживаемые tools:

+- `get_service_info()`
+- `sync_employees(client_hash?, include_data?)`
 - `search_employees(query, status?, limit?)`
 - `get_employee(profile_id_or_url)`
 - `list_employee_publications(profile_id_or_url)`
 - `list_employee_courses(profile_id_or_url)`
 - `get_crawl_status()`
+- `get_crawl_run_details(run_id)`

-Пример локального legacy-режима со статическим токеном:
+`get_service_info` возвращает метаданные сервиса, список tools и текущую версию набора сотрудников. `sync_employees` отдает полный snapshot или delta по `client_hash`; checksum набора строится по сотрудникам, их статусам и текущим checksums. Ответы tools возвращаются как JSON-строка внутри MCP `content[0].text`.
+
+Пример локального запроса списка tools:

 ```bash
 curl http://localhost:8001/mcp \
-  -H "Authorization: Bearer change-me-mcp-token" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'
 ```

-Для production OAuth/OIDC настройте внешний authorization server и включите режим `oauth`:
-
-```env
-MCP_AUTH_MODE=oauth
-MCP_RESOURCE_URL=https://example.com/mcp
-MCP_OAUTH_ISSUER=https://auth.example.com
-MCP_OAUTH_AUDIENCE=miem-mcp
-MCP_OAUTH_JWKS_URL=https://auth.example.com/.well-known/jwks.json
-MCP_OAUTH_REQUIRED_SCOPE=mcp:tools
-```
-
-MCP server работает как OAuth protected resource: он не выдает токены, а проверяет JWT access token по JWKS, `issuer`, `audience`, сроку действия и scope. Metadata для MCP-клиентов доступна по `GET /.well-known/oauth-protected-resource`.
+Если MCP нужно ограничить, делайте это на сетевом уровне: localhost binding, VPN, firewall, reverse proxy или другой внешний контур доступа.

 ## Обслуживание

@@ -131,4 +115,4 @@ docker compose exec postgres pg_dump -U miem miem_workers > backup.sql
 docker compose down
 ```

-Версия сервиса: `0.4.0`. Админка всегда показывает версии backend и frontend в footer.
+Версия сервиса: `0.6.0`. Админка всегда показывает версии backend и frontend в footer.
--- a/app/admin.py
+++ b/app/admin.py
@@ -17,6 +17,7 @@ from app.services.admin_data import (
    stats_payload,
 )
 from app.services.crawl_control import get_running_run, run_crawl_if_idle
+from app.services.crawler import refresh_employee
 from app.version import BACKEND_VERSION, FRONTEND_VERSION

 router = APIRouter(prefix="/admin")
@@ -144,10 +145,31 @@ def employee_detail(
    return _render(
        request,
        "employee_detail.html",
-        {"employee": employee, "employee_view": employee_detail_payload(employee), "snapshots": snapshots},
+        {
+            "employee": employee,
+            "employee_view": employee_detail_payload(employee),
+            "snapshots": snapshots,
+            "refresh_status": request.query_params.get("refresh_status"),
+        },
    )


+@router.post("/employees/{employee_id}/refresh")
+def refresh_employee_detail(
+    employee_id: int,
+    request: Request,
+    db: Session = Depends(get_db),
+    settings: Settings = Depends(get_settings),
+):
+    require_admin(request, settings)
+    employee = db.get(Employee, employee_id)
+    if not employee:
+        return RedirectResponse("/admin/directory", status_code=303)
+    run = refresh_employee(db, employee, settings)
+    status = "success" if run.status == "completed" else "error"
+    return RedirectResponse(f"/admin/employees/{employee_id}?refresh_status={status}", status_code=303)
+
+
@router.get("/runs", response_class=HTMLResponse)
 def runs(request: Request, db: Session = Depends(get_db), settings: Settings = Depends(get_settings)):
    require_admin(request, settings)
--- a/app/config.py
+++ b/app/config.py
@@ -1,6 +1,4 @@
 from functools import lru_cache
-from typing import Literal
-
 from pydantic import Field, field_validator
 from pydantic_settings import BaseSettings, SettingsConfigDict

@@ -19,13 +17,6 @@ class Settings(BaseSettings):
    admin_username: str = "admin"
    admin_password: str = "admin"
    session_secret: str = Field(default="dev-session-secret", min_length=8)
-    mcp_token: str = "dev-mcp-token"
-    mcp_auth_mode: Literal["token", "oauth"] = "oauth"
-    mcp_resource_url: str = "http://localhost:8001/mcp"
-    mcp_oauth_issuer: str = ""
-    mcp_oauth_audience: str = ""
-    mcp_oauth_jwks_url: str = ""
-    mcp_oauth_required_scope: str = "mcp:tools"

    @field_validator("crawl_limit", mode="before")
    @classmethod
@@ -34,15 +25,6 @@ class Settings(BaseSettings):
            return None
        return value

-    def oauth_jwks_url(self) -> str:
-        if self.mcp_oauth_jwks_url:
-            return self.mcp_oauth_jwks_url
-        issuer = self.mcp_oauth_issuer.rstrip("/")
-        if not issuer:
-            return ""
-        return f"{issuer}/.well-known/jwks.json"
-
-
@lru_cache
 def get_settings() -> Settings:
    return Settings()
--- a/app/main.py
+++ b/app/main.py
@@ -4,7 +4,6 @@ from fastapi.staticfiles import StaticFiles
 from app.admin import router as admin_router
 from app.api import router as api_router
 from app.db import init_db
-from app.mcp import metadata_router as mcp_metadata_router
 from app.mcp import router as mcp_router
 from app.version import BACKEND_VERSION

@@ -13,7 +12,6 @@ app.mount("/static", StaticFiles(directory="app/static"), name="static")
 app.include_router(api_router)
 app.include_router(admin_router)
 app.include_router(mcp_router)
-app.include_router(mcp_metadata_router)


@app.on_event("startup")
--- a/app/mcp.py
+++ b/app/mcp.py
@@ -4,18 +4,34 @@ from fastapi import APIRouter, Depends, Request
 from sqlalchemy import desc, or_, select
 from sqlalchemy.orm import Session

-from app.config import Settings, get_settings
 from app.db import get_db
 from app.models import CrawlRun, Employee
-from app.security import mcp_protected_resource_metadata, require_mcp_auth
 from app.services.admin_data import run_detail_payload
+from app.services.dataset_versions import service_info_payload, sync_employees_payload
 from app.version import BACKEND_VERSION

 router = APIRouter(prefix="/mcp")
-metadata_router = APIRouter()
+PROTOCOL_VERSION = "2024-11-05"
+SERVICE_NAME = "miem-employees"


 TOOLS = [
+    {
+        "name": "get_service_info",
+        "description": "Return service metadata, supported tools, and current dataset version.",
+        "inputSchema": {"type": "object", "properties": {}},
+    },
+    {
+        "name": "sync_employees",
+        "description": "Synchronize employees by dataset hash. Returns a full snapshot or a delta from client_hash.",
+        "inputSchema": {
+            "type": "object",
+            "properties": {
+                "client_hash": {"type": "string"},
+                "include_data": {"type": "boolean", "default": True},
+            },
+        },
+    },
    {
        "name": "search_employees",
        "description": "Search MIEM employees by name or profile URL.",
@@ -65,9 +81,7 @@ TOOLS = [
 async def mcp_http(
    request: Request,
    db: Session = Depends(get_db),
-    settings: Settings = Depends(get_settings),
 ) -> dict:
-    require_mcp_auth(request, settings)
    payload = await request.json()
    method = payload.get("method")
    request_id = payload.get("id")
@@ -76,8 +90,8 @@ async def mcp_http(
    try:
        if method == "initialize":
            result = {
-                "protocolVersion": "2024-11-05",
-                "serverInfo": {"name": "miem-employees", "version": BACKEND_VERSION},
+                "protocolVersion": PROTOCOL_VERSION,
+                "serverInfo": {"name": SERVICE_NAME, "version": BACKEND_VERSION},
                "capabilities": {"tools": {}},
            }
        elif method == "tools/list":
@@ -92,6 +106,24 @@ async def mcp_http(


 def _call_tool(db: Session, name: str, arguments: dict) -> dict:
+    if name == "get_service_info":
+        return _tool_response(
+            service_info_payload(
+                db,
+                tools=TOOLS,
+                service_name=SERVICE_NAME,
+                backend_version=BACKEND_VERSION,
+                protocol_version=PROTOCOL_VERSION,
+            )
+        )
+    if name == "sync_employees":
+        return _tool_response(
+            sync_employees_payload(
+                db,
+                client_hash=arguments.get("client_hash"),
+                include_data=bool(arguments.get("include_data", True)),
+            )
+        )
    if name == "search_employees":
        return _tool_response(_search_employees(db, arguments))
    if name == "get_employee":
@@ -176,6 +208,7 @@ def _run_payload(run: CrawlRun) -> dict:
        "finished_at": run.finished_at.isoformat() if run.finished_at else None,
        "found_count": run.found_count,
        "parsed_count": run.parsed_count,
+        "skipped_count": run.skipped_count,
        "error_count": run.error_count,
        "dismissed_count": run.dismissed_count,
    }
@@ -183,8 +216,3 @@ def _run_payload(run: CrawlRun) -> dict:

 def _tool_response(data: object) -> dict:
    return {"content": [{"type": "text", "text": json.dumps(data, ensure_ascii=False, default=str)}]}
-
-
-@metadata_router.get("/.well-known/oauth-protected-resource")
-def oauth_protected_resource(settings: Settings = Depends(get_settings)) -> dict:
-    return mcp_protected_resource_metadata(settings)
--- a/app/models.py
+++ b/app/models.py
@@ -70,12 +70,14 @@ class CrawlRun(Base):
    finished_at: Mapped[datetime | None] = mapped_column(DateTime(timezone=True))
    found_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
    parsed_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
+    skipped_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
    new_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
    error_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
    dismissed_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
    message: Mapped[str | None] = mapped_column(Text)

    employee_changes: Mapped[list["CrawlRunEmployeeChange"]] = relationship(back_populates="crawl_run")
+    dataset_versions: Mapped[list["DatasetVersion"]] = relationship(back_populates="crawl_run")


 class CrawlRunEmployeeChange(Base):
@@ -134,3 +136,63 @@ class ParserSource(Base):
    source_url: Mapped[str] = mapped_column(Text, nullable=False)
    enabled: Mapped[bool] = mapped_column(default=True, nullable=False)
    created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=utcnow, nullable=False)
+
+
+class ParseResourceCache(Base):
+    __tablename__ = "parse_resource_cache"
+    __table_args__ = (
+        UniqueConstraint("profile_key", "resource_key", "request_fingerprint", name="uq_parse_resource_cache_resource"),
+        Index("ix_parse_resource_cache_profile_key", "profile_key"),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True)
+    profile_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    resource_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    method: Mapped[str] = mapped_column(String(16), nullable=False)
+    url: Mapped[str] = mapped_column(Text, nullable=False)
+    request_fingerprint: Mapped[str] = mapped_column(String(64), nullable=False)
+    etag: Mapped[str | None] = mapped_column(Text)
+    last_modified: Mapped[str | None] = mapped_column(Text)
+    body_hash: Mapped[str] = mapped_column(String(64), nullable=False)
+    body_snapshot: Mapped[bytes] = mapped_column(LargeBinary, nullable=False)
+    parser_version: Mapped[str | None] = mapped_column(String(32))
+    fetched_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=utcnow, nullable=False)
+
+
+class DatasetVersion(Base):
+    __tablename__ = "dataset_versions"
+    __table_args__ = (
+        UniqueConstraint("hash", name="uq_dataset_versions_hash"),
+        Index("ix_dataset_versions_created_at", "created_at"),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True)
+    hash: Mapped[str] = mapped_column(String(64), nullable=False)
+    previous_hash: Mapped[str | None] = mapped_column(String(64))
+    crawl_run_id: Mapped[int | None] = mapped_column(ForeignKey("crawl_runs.id"))
+    employee_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
+    active_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
+    dismissed_count: Mapped[int] = mapped_column(Integer, default=0, nullable=False)
+    created_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), default=utcnow, nullable=False)
+
+    crawl_run: Mapped[CrawlRun | None] = relationship(back_populates="dataset_versions")
+    items: Mapped[list["DatasetVersionItem"]] = relationship(back_populates="dataset_version", cascade="all, delete-orphan")
+
+
+class DatasetVersionItem(Base):
+    __tablename__ = "dataset_version_items"
+    __table_args__ = (
+        UniqueConstraint("dataset_version_id", "profile_key", name="uq_dataset_version_items_version_profile"),
+        Index("ix_dataset_version_items_hash", "dataset_version_id"),
+        Index("ix_dataset_version_items_profile_key", "profile_key"),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True)
+    dataset_version_id: Mapped[int] = mapped_column(ForeignKey("dataset_versions.id"), nullable=False)
+    profile_key: Mapped[str] = mapped_column(String(255), nullable=False)
+    employee_id: Mapped[int | None] = mapped_column(ForeignKey("employees.id"))
+    status: Mapped[str] = mapped_column(String(32), nullable=False)
+    checksum: Mapped[str] = mapped_column(String(64), nullable=False)
+
+    dataset_version: Mapped[DatasetVersion] = relationship(back_populates="items")
+    employee: Mapped[Employee | None] = relationship()
--- a/app/parser/profile.py
+++ b/app/parser/profile.py
@@ -1,3 +1,5 @@
+import hashlib
+import json
 import re
 from urllib.parse import urljoin

@@ -149,22 +151,42 @@ def parse_person_profile(
    headers: dict[str, str],
    timeout: int,
    use_playwright: bool = False,
+    resource_cache=None,
 ) -> dict | None:
    normalized_url = normalize_profile_url(source_url)
    if not normalized_url:
        return None
-    response = session.get(normalized_url, headers=headers, timeout=timeout)
-    response.raise_for_status()
-    html = response.text
+    profile_type, profile_id = parse_profile_identity(normalized_url)
+    cache_profile_key = f"{profile_type}:{profile_id}"
+    resource_manifest = []
+    html = _fetch_text(
+        session,
+        normalized_url,
+        headers,
+        timeout,
+        resource_cache=resource_cache,
+        profile_key=cache_profile_key,
+        resource_key="main-html",
+        resource_manifest=resource_manifest,
+    )
    if use_playwright:
        html = _render_with_playwright(normalized_url, html)

    soup = BeautifulSoup(html, "html.parser")
-    profile_type, profile_id = parse_profile_identity(normalized_url)
    header = extract_person_header(soup, normalized_url)
    tabs = extract_person_tabs(soup, normalized_url)
    sections = extract_sections(soup, normalized_url)
-    sections = enrich_sections_from_hse_widgets(session, soup, normalized_url, headers, timeout, sections)
+    sections = enrich_sections_from_hse_widgets(
+        session,
+        soup,
+        normalized_url,
+        headers,
+        timeout,
+        sections,
+        resource_cache=resource_cache,
+        profile_key=cache_profile_key,
+        resource_manifest=resource_manifest,
+    )
    internal_links = [tab["href"] for tab in tabs if tab.get("href")]

    return {
@@ -181,6 +203,7 @@ def parse_person_profile(
        "employee_internal_links": internal_links,
        "parser_version": BACKEND_VERSION,
        "_html": html,
+        "_resource_manifest": resource_manifest,
    }


@@ -191,13 +214,33 @@ def enrich_sections_from_hse_widgets(
    headers: dict[str, str],
    timeout: int,
    sections: list[dict],
+    resource_cache=None,
+    profile_key: str | None = None,
+    resource_manifest: list[dict] | None = None,
 ) -> list[dict]:
    enriched = list(sections)
-    publications = _load_widget_publications(session, soup, headers, timeout)
+    publications = _load_widget_publications(
+        session,
+        soup,
+        headers,
+        timeout,
+        resource_cache=resource_cache,
+        profile_key=profile_key,
+        resource_manifest=resource_manifest,
+    )
    if publications:
        enriched = _upsert_publications_section(enriched, publications)

-    theses = _load_widget_graduation_theses(session, soup, source_url, headers, timeout)
+    theses = _load_widget_graduation_theses(
+        session,
+        soup,
+        source_url,
+        headers,
+        timeout,
+        resource_cache=resource_cache,
+        profile_key=profile_key,
+        resource_manifest=resource_manifest,
+    )
    if theses:
        enriched = _upsert_graduation_theses_section(enriched, theses)
    return enriched
@@ -226,7 +269,16 @@ def _render_with_playwright(source_url: str, fallback_html: str) -> str:
        return fallback_html


-def _load_widget_publications(session: Session, soup: BeautifulSoup, headers: dict[str, str], timeout: int) -> list[dict]:
+def _load_widget_publications(
+    session: Session,
+    soup: BeautifulSoup,
+    headers: dict[str, str],
+    timeout: int,
+    *,
+    resource_cache=None,
+    profile_key: str | None = None,
+    resource_manifest: list[dict] | None = None,
+) -> list[dict]:
    script = soup.select_one('script[data-widget-name="AuthorSearch"][data-author]')
    if not script:
        return []
@@ -251,22 +303,37 @@ def _load_widget_publications(session: Session, soup: BeautifulSoup, headers: di
            },
        }
        try:
-            response = session.post(
-                "https://publications.hse.ru/api/searchPubs",
-                json=payload,
-                headers=headers,
-                timeout=timeout,
-            )
-            response.raise_for_status()
-            data = response.json()
+            if resource_cache and profile_key:
+                text = _fetch_text(
+                    session,
+                    "https://publications.hse.ru/api/searchPubs",
+                    headers,
+                    timeout,
+                    resource_cache=resource_cache,
+                    profile_key=profile_key,
+                    resource_key=f"publications-page-{page_id}",
+                    resource_manifest=resource_manifest,
+                    method="POST",
+                    json_payload=payload,
+                )
+                data = json.loads(text)
+            else:
+                response = session.post(
+                    "https://publications.hse.ru/api/searchPubs",
+                    json=payload,
+                    headers=headers,
+                    timeout=timeout,
+                )
+                response.raise_for_status()
+                data = response.json()
        except Exception:
            return publications

        result = data.get("result") if isinstance(data, dict) else {}
-        items = result.get("items") if isinstance(result, dict) else []
-        if not isinstance(items, list) or not items:
+        items = _extract_publication_items(result)
+        if not items:
            break
-        publications.extend(_normalize_publication_item(item) for item in items if isinstance(item, dict))
+        publications.extend(_normalize_publication_item(item) for item in items)

        total = int(result.get("total") or 0)
        if not result.get("more") and len(publications) >= total:
@@ -275,12 +342,44 @@ def _load_widget_publications(session: Session, soup: BeautifulSoup, headers: di
    return _dedupe_publications(publications)


+def _extract_publication_items(result: object) -> list[dict]:
+    if not isinstance(result, dict):
+        return []
+    return _flatten_publication_items(result.get("items"))
+
+
+def _flatten_publication_items(value: object) -> list[dict]:
+    if isinstance(value, list):
+        return [item for item in value if _is_publication_item(item)]
+    if not isinstance(value, dict):
+        return []
+
+    nested_items = value.get("items")
+    if isinstance(nested_items, list):
+        return [item for item in nested_items if _is_publication_item(item)]
+    if isinstance(nested_items, dict):
+        return _flatten_publication_items(nested_items)
+
+    publications = []
+    for child in value.values():
+        publications.extend(_flatten_publication_items(child))
+    return publications
+
+
+def _is_publication_item(value: object) -> bool:
+    return isinstance(value, dict) and ("id" in value or "title" in value)
+
+
 def _load_widget_graduation_theses(
    session: Session,
    soup: BeautifulSoup,
    source_url: str,
    headers: dict[str, str],
    timeout: int,
+    *,
+    resource_cache=None,
+    profile_key: str | None = None,
+    resource_manifest: list[dict] | None = None,
 ) -> list[dict]:
    script = soup.select_one('script[src*="/n/stat/vkr/app.js"][data-person-id]')
    if not script:
@@ -292,14 +391,30 @@ def _load_widget_graduation_theses(

    request_headers = {**headers, "x-portal-language": "ru"}
    try:
-        response = session.get(
-            urljoin(source_url, api_url),
-            params={"supervisorId": person_id},
-            headers=request_headers,
-            timeout=timeout,
-        )
-        response.raise_for_status()
-        data = response.json()
+        url = urljoin(source_url, api_url)
+        params = {"supervisorId": person_id}
+        if resource_cache and profile_key:
+            text = _fetch_text(
+                session,
+                url,
+                request_headers,
+                timeout,
+                resource_cache=resource_cache,
+                profile_key=profile_key,
+                resource_key="graduation-theses",
+                resource_manifest=resource_manifest,
+                params=params,
+            )
+            data = json.loads(text)
+        else:
+            response = session.get(
+                url,
+                params=params,
+                headers=request_headers,
+                timeout=timeout,
+            )
+            response.raise_for_status()
+            data = response.json()
    except Exception:
        return []

@@ -359,7 +474,7 @@ def _infer_section_type(title: str, nodes: list) -> str:
    lowered = title.lower()
    if _has_table(nodes):
        return "table"
-    if "публикац" in lowered:
+    if _is_publications_title(lowered):
        return "publications"
    if "учебные курсы" in lowered:
        return "courses_by_year"
@@ -370,6 +485,10 @@ def _infer_section_type(title: str, nodes: list) -> str:
    return "generic"


+def _is_publications_title(lowered_title: str) -> bool:
+    return lowered_title.startswith("публикац")
+
+
 def _has_table(nodes: list) -> bool:
    return any(isinstance(node, Tag) and (node.name == "table" or node.find("table")) for node in nodes)

@@ -597,3 +716,62 @@ def _dedupe_dicts(items: list[dict]) -> list[dict]:
            seen.add(key)
            unique.append(item)
    return unique
+
+
+def _fetch_text(
+    session: Session,
+    url: str,
+    headers: dict[str, str],
+    timeout: int,
+    *,
+    resource_cache=None,
+    profile_key: str | None = None,
+    resource_key: str,
+    resource_manifest: list[dict] | None,
+    method: str = "GET",
+    json_payload: object | None = None,
+    params: dict | None = None,
+) -> str:
+    if resource_cache and profile_key:
+        cached = resource_cache.fetch_text(
+            session,
+            profile_key=profile_key,
+            resource_key=resource_key,
+            method=method,
+            url=url,
+            headers=headers,
+            timeout=timeout,
+            json_payload=json_payload,
+            params=params,
+        )
+        if resource_manifest is not None:
+            resource_manifest.append(
+                {
+                    "resource_key": resource_key,
+                    "method": method,
+                    "url": url,
+                    "body_hash": cached.body_hash,
+                    "from_cache": cached.from_cache,
+                    "status_code": cached.status_code,
+                }
+            )
+        return cached.text
+
+    if method.upper() == "POST":
+        response = session.post(url, json=json_payload, headers=headers, timeout=timeout, params=params)
+    else:
+        response = session.get(url, headers=headers, timeout=timeout, params=params)
+    response.raise_for_status()
+    text = response.text
+    if resource_manifest is not None:
+        resource_manifest.append(
+            {
+                "resource_key": resource_key,
+                "method": method,
+                "url": url,
+                "body_hash": hashlib.sha256(text.encode("utf-8")).hexdigest(),
+                "from_cache": False,
+                "status_code": response.status_code,
+            }
+        )
+    return text
--- a/app/security.py
+++ b/app/security.py
@@ -3,10 +3,7 @@ import hashlib
 import hmac
 import json
 import time
-from functools import lru_cache

-import jwt
-from jwt import PyJWKClient, PyJWTError
 from fastapi import HTTPException, Request, status

 from app.config import Settings
@@ -47,93 +44,3 @@ def require_admin(request: Request, settings: Settings) -> str:
    if not username:
        raise HTTPException(status_code=status.HTTP_303_SEE_OTHER, headers={"Location": "/admin/login"})
    return username
-
-
-def require_mcp_auth(request: Request, settings: Settings) -> None:
-    auth = request.headers.get("authorization", "")
-    if not auth.startswith("Bearer "):
-        raise _mcp_unauthorized(settings, "Missing bearer token")
-
-    token = auth.removeprefix("Bearer ").strip()
-    if _mcp_static_token_allowed(settings) and hmac.compare_digest(token, settings.mcp_token):
-        return
-    if _mcp_oauth_allowed(settings):
-        _validate_mcp_oauth_token(token, settings)
-        return
-    raise _mcp_unauthorized(settings, "Invalid MCP token")
-
-
-def require_mcp_token(request: Request, settings: Settings) -> None:
-    require_mcp_auth(request, settings)
-
-
-def mcp_protected_resource_metadata(settings: Settings) -> dict:
-    authorization_servers = [settings.mcp_oauth_issuer.rstrip("/")] if settings.mcp_oauth_issuer else []
-    return {
-        "resource": settings.mcp_resource_url,
-        "authorization_servers": authorization_servers,
-        "bearer_methods_supported": ["header"],
-        "scopes_supported": [settings.mcp_oauth_required_scope],
-        "resource_documentation": settings.mcp_resource_url,
-    }
-
-
-def _mcp_static_token_allowed(settings: Settings) -> bool:
-    return settings.mcp_auth_mode == "token"
-
-
-def _mcp_oauth_allowed(settings: Settings) -> bool:
-    return settings.mcp_auth_mode == "oauth"
-
-
-def _validate_mcp_oauth_token(token: str, settings: Settings) -> None:
-    if not settings.mcp_oauth_issuer or not settings.mcp_oauth_audience or not settings.oauth_jwks_url():
-        raise _mcp_unauthorized(settings, "MCP OAuth is not configured")
-    try:
-        signing_key = _get_mcp_oauth_signing_key(token, settings).key
-        claims = jwt.decode(
-            token,
-            signing_key,
-            algorithms=["RS256", "RS384", "RS512", "ES256", "ES384", "ES512"],
-            audience=settings.mcp_oauth_audience,
-            issuer=settings.mcp_oauth_issuer.rstrip("/"),
-        )
-    except PyJWTError as exc:
-        raise _mcp_unauthorized(settings, "Invalid OAuth access token") from exc
-    if not _claims_have_scope(claims, settings.mcp_oauth_required_scope):
-        raise HTTPException(status_code=status.HTTP_403_FORBIDDEN, detail="Missing required MCP OAuth scope")
-
-
-def _claims_have_scope(claims: dict, required_scope: str) -> bool:
-    scopes: set[str] = set()
-    scope = claims.get("scope")
-    if isinstance(scope, str):
-        scopes.update(scope.split())
-    scp = claims.get("scp")
-    if isinstance(scp, str):
-        scopes.update(scp.split())
-    elif isinstance(scp, list):
-        scopes.update(str(item) for item in scp)
-    return required_scope in scopes
-
-
-@lru_cache(maxsize=16)
-def _get_jwk_client(jwks_url: str) -> PyJWKClient:
-    return PyJWKClient(jwks_url)
-
-
-def _get_mcp_oauth_signing_key(token: str, settings: Settings):
-    return _get_jwk_client(settings.oauth_jwks_url()).get_signing_key_from_jwt(token)
-
-
-def _mcp_unauthorized(settings: Settings, detail: str) -> HTTPException:
-    headers = {}
-    if _mcp_oauth_allowed(settings):
-        headers["WWW-Authenticate"] = f'Bearer resource_metadata="{_mcp_metadata_url(settings)}"'
-    return HTTPException(status_code=status.HTTP_401_UNAUTHORIZED, detail=detail, headers=headers)
-
-
-def _mcp_metadata_url(settings: Settings) -> str:
-    resource_url = settings.mcp_resource_url.rstrip("/")
-    base_url = resource_url[: -len("/mcp")] if resource_url.endswith("/mcp") else resource_url
-    return f"{base_url}/.well-known/oauth-protected-resource"
--- a/app/services/admin_data.py
+++ b/app/services/admin_data.py
@@ -153,7 +153,7 @@ def stats_payload(db: Session) -> dict[str, Any]:
 def run_payload(run: CrawlRun | None) -> dict[str, Any] | None:
    if not run:
        return None
-    processed = run.parsed_count + run.error_count
+    processed = run.parsed_count + run.skipped_count + run.error_count
    percent = round((processed / run.found_count) * 100, 1) if run.found_count else 0
    return {
        "id": run.id,
@@ -166,6 +166,7 @@ def run_payload(run: CrawlRun | None) -> dict[str, Any] | None:
        "finished_display": format_admin_datetime(run.finished_at),
        "found_count": run.found_count,
        "parsed_count": run.parsed_count,
+        "skipped_count": run.skipped_count,
        "new_count": run.new_count,
        "error_count": run.error_count,
        "dismissed_count": run.dismissed_count,
--- a/app/services/crawler.py
+++ b/app/services/crawler.py
@@ -1,6 +1,7 @@
 import gzip
 import hashlib
 import json
+import re
 import time
 from datetime import datetime, timezone

@@ -13,6 +14,8 @@ from app.models import CrawlError, CrawlRun, CrawlRunEmployeeChange, Employee, E
 from app.parser.collector import collect_profile_links
 from app.parser.profile import parse_person_profile
 from app.parser.profile_url import profile_key
+from app.services.dataset_versions import get_or_create_current_version
+from app.services.resource_cache import ResourceCache

 HEADERS = {
    "User-Agent": "Mozilla/5.0 (compatible; MIEMEmployeesBot/0.1.0; +https://miem.hse.ru/)"
@@ -28,8 +31,10 @@ def run_crawl(db: Session, settings: Settings) -> CrawlRun:

    found_keys: set[str] = set()
    parsed_count = 0
+    skipped_count = 0
    try:
        with requests.Session() as session:
+            resource_cache = ResourceCache(db)
            urls = collect_profile_links(session, source.source_url, HEADERS, settings.request_timeout)
            if settings.crawl_limit:
                urls = urls[: settings.crawl_limit]
@@ -47,12 +52,17 @@ def run_crawl(db: Session, settings: Settings) -> CrawlRun:
                        HEADERS,
                        settings.request_timeout,
                        settings.parser_use_playwright,
+                        resource_cache=resource_cache,
                    )
                    if not parsed:
                        continue
-                    _upsert_employee(db, run, parsed)
-                    parsed_count += 1
+                    _, changed = _upsert_employee(db, run, parsed)
+                    if changed:
+                        parsed_count += 1
+                    else:
+                        skipped_count += 1
                    run.parsed_count = parsed_count
+                    run.skipped_count = skipped_count
                    db.commit()
                except Exception as exc:
                    run.error_count += 1
@@ -68,8 +78,9 @@ def run_crawl(db: Session, settings: Settings) -> CrawlRun:
                finally:
                    time.sleep(settings.request_delay_seconds)

-        run.dismissed_count = _mark_dismissed(db, run, found_keys, session, settings.request_timeout)
+            run.dismissed_count = _mark_dismissed(db, run, found_keys, session, settings.request_timeout)
        run.status = "completed"
+        get_or_create_current_version(db, crawl_run_id=run.id)
    except Exception as exc:
        run.status = "failed"
        run.message = str(exc)
@@ -80,6 +91,54 @@ def run_crawl(db: Session, settings: Settings) -> CrawlRun:
    return run


+def refresh_employee(db: Session, employee: Employee, settings: Settings) -> CrawlRun:
+    run = CrawlRun(source_url=employee.canonical_url, status="running", found_count=1)
+    db.add(run)
+    db.commit()
+    db.refresh(run)
+
+    try:
+        with requests.Session() as session:
+            resource_cache = ResourceCache(db)
+            parsed = parse_person_profile(
+                session,
+                employee.canonical_url,
+                HEADERS,
+                settings.request_timeout,
+                settings.parser_use_playwright,
+                resource_cache=resource_cache,
+            )
+        if not parsed:
+            raise ValueError("Профиль не удалось распарсить.")
+        if _parsed_profile_key(parsed) != employee.profile_key:
+            raise ValueError("Распарсенный профиль не совпадает с обновляемым сотрудником.")
+
+        _, changed = _upsert_employee(db, run, parsed)
+        if changed:
+            run.parsed_count = 1
+        else:
+            run.skipped_count = 1
+        run.status = "completed"
+        get_or_create_current_version(db, crawl_run_id=run.id)
+    except Exception as exc:
+        run.status = "failed"
+        run.error_count = 1
+        run.message = str(exc)
+        db.add(
+            CrawlError(
+                crawl_run_id=run.id,
+                profile_url=employee.canonical_url,
+                error_type=type(exc).__name__,
+                message=str(exc),
+            )
+        )
+    finally:
+        run.finished_at = datetime.now(timezone.utc)
+        db.commit()
+        db.refresh(run)
+    return run
+
+
 def _ensure_source(db: Session, source_url: str) -> ParserSource:
    source = db.scalar(select(ParserSource).where(ParserSource.source_url == source_url))
    if source:
@@ -91,10 +150,15 @@ def _ensure_source(db: Session, source_url: str) -> ParserSource:
    return source


-def _upsert_employee(db: Session, run: CrawlRun, parsed: dict) -> Employee:
+def _parsed_profile_key(parsed: dict) -> str:
+    return f"{parsed.get('profile_type')}:{parsed.get('profile_id')}"
+
+
+def _upsert_employee(db: Session, run: CrawlRun, parsed: dict) -> tuple[Employee, bool]:
    html = parsed.pop("_html", None)
+    parsed.pop("_resource_manifest", None)
    checksum = _checksum(parsed)
-    key = f"{parsed.get('profile_type')}:{parsed.get('profile_id')}"
+    key = _parsed_profile_key(parsed)
    employee = db.scalar(select(Employee).where(Employee.profile_key == key))
    now = datetime.now(timezone.utc)
    if not employee:
@@ -111,12 +175,15 @@ def _upsert_employee(db: Session, run: CrawlRun, parsed: dict) -> Employee:
    else:
        is_new = False

+    parser_version = parsed.get("parser_version")
+    changed = is_new or employee.current_checksum != checksum or employee.parser_version != parser_version
    employee.full_name = parsed.get("full_name")
    employee.status = "active"
    employee.last_seen_at = now
    employee.dismissed_at = None
-    employee.parser_version = parsed.get("parser_version")
-    employee.current_data = parsed
+    employee.parser_version = parser_version
+    if changed:
+        employee.current_data = parsed
    employee.current_checksum = checksum
    db.flush()

@@ -130,28 +197,29 @@ def _upsert_employee(db: Session, run: CrawlRun, parsed: dict) -> Employee:
            message="Сотрудник впервые найден в источнике.",
        )

-    db.query(ProfileTab).filter(ProfileTab.employee_id == employee.id).delete()
-    for tab in parsed.get("tabs") or []:
+    if changed:
+        db.query(ProfileTab).filter(ProfileTab.employee_id == employee.id).delete()
+        for tab in parsed.get("tabs") or []:
+            db.add(
+                ProfileTab(
+                    employee_id=employee.id,
+                    title=tab.get("title") or "",
+                    href=tab.get("href") or "",
+                    data_index=tab.get("data_index"),
+                )
+            )
+
        db.add(
-            ProfileTab(
+            EmployeeSnapshot(
                employee_id=employee.id,
-                title=tab.get("title") or "",
-                href=tab.get("href") or "",
-                data_index=tab.get("data_index"),
+                crawl_run_id=run.id,
+                parsed_data=parsed,
+                html_snapshot=gzip.compress(html.encode("utf-8")) if html else None,
+                checksum=checksum,
+                parser_version=parser_version,
            )
        )
-
-    db.add(
-        EmployeeSnapshot(
-            employee_id=employee.id,
-            crawl_run_id=run.id,
-            parsed_data=parsed,
-            html_snapshot=gzip.compress(html.encode("utf-8")) if html else None,
-            checksum=checksum,
-            parser_version=parsed.get("parser_version"),
-        )
-    )
-    return employee
+    return employee, changed


 def _mark_dismissed(db: Session, run: CrawlRun, found_keys: set[str], session: requests.Session, timeout: int) -> int:
@@ -219,5 +287,23 @@ def _record_employee_change(


 def _checksum(data: dict) -> str:
-    payload = json.dumps(data, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
+    payload = json.dumps(_stable_checksum_payload(data), ensure_ascii=False, sort_keys=True, separators=(",", ":"))
    return hashlib.sha256(payload.encode("utf-8")).hexdigest()
+
+
+def _stable_checksum_payload(value):
+    if isinstance(value, dict):
+        return {key: _stable_checksum_payload(item) for key, item in value.items()}
+    if isinstance(value, list):
+        return [_stable_checksum_payload(item) for item in value]
+    if isinstance(value, str):
+        return _normalize_date_dependent_experience(value)
+    return value
+
+
+def _normalize_date_dependent_experience(value: str) -> str:
+    return re.sub(
+        r"(?i)(стаж(?:\s+работы)?(?:\s+в\s+ниу\s+вшэ|\s+в\s+вшэ)?\s*:?\s*)\d+\s*(?:год(?:а|ов)?|лет)",
+        r"\1<experience-years>",
+        value,
+    )
--- a/app/services/dataset_versions.py
+++ b/app/services/dataset_versions.py
@@ -0,0 +1,227 @@
+import hashlib
+import json
+from dataclasses import dataclass
+
+from sqlalchemy import desc, select
+from sqlalchemy.orm import Session
+
+from app.models import DatasetVersion, DatasetVersionItem, Employee
+
+
+@dataclass(frozen=True)
+class EmployeeMarker:
+    profile_key: str
+    employee_id: int | None
+    status: str
+    checksum: str
+
+
+def get_or_create_current_version(db: Session, *, crawl_run_id: int | None = None) -> DatasetVersion:
+    employees = db.scalars(select(Employee).order_by(Employee.profile_key)).all()
+    markers = [_employee_marker(employee) for employee in employees]
+    dataset_hash = _dataset_hash(markers)
+    latest = get_latest_version(db)
+    if latest and latest.hash == dataset_hash:
+        return latest
+
+    active_count = sum(1 for marker in markers if marker.status == "active")
+    dismissed_count = sum(1 for marker in markers if marker.status == "dismissed")
+    version = DatasetVersion(
+        hash=dataset_hash,
+        previous_hash=latest.hash if latest else None,
+        crawl_run_id=crawl_run_id,
+        employee_count=len(markers),
+        active_count=active_count,
+        dismissed_count=dismissed_count,
+    )
+    db.add(version)
+    db.flush()
+    for marker in markers:
+        db.add(
+            DatasetVersionItem(
+                dataset_version_id=version.id,
+                profile_key=marker.profile_key,
+                employee_id=marker.employee_id,
+                status=marker.status,
+                checksum=marker.checksum,
+            )
+        )
+    db.flush()
+    return version
+
+
+def get_latest_version(db: Session) -> DatasetVersion | None:
+    return db.scalar(select(DatasetVersion).order_by(desc(DatasetVersion.created_at), desc(DatasetVersion.id)).limit(1))
+
+
+def get_version_by_hash(db: Session, dataset_hash: str | None) -> DatasetVersion | None:
+    if not dataset_hash:
+        return None
+    return db.scalar(select(DatasetVersion).where(DatasetVersion.hash == dataset_hash).limit(1))
+
+
+def service_info_payload(db: Session, *, tools: list[dict], service_name: str, backend_version: str, protocol_version: str) -> dict:
+    version = get_or_create_current_version(db)
+    db.commit()
+    return {
+        "service_name": service_name,
+        "backend_version": backend_version,
+        "protocolVersion": protocol_version,
+        "tools": tools,
+        "dataset": _version_payload(version),
+    }
+
+
+def sync_employees_payload(db: Session, *, client_hash: str | None = None, include_data: bool = True) -> dict:
+    current = get_or_create_current_version(db)
+    db.commit()
+    if not client_hash:
+        return _full_sync_payload(db, current, include_data=include_data, reason=None)
+    if client_hash == current.hash:
+        return {
+            "mode": "delta",
+            "from_hash": client_hash,
+            "to_hash": current.hash,
+            "dataset": _version_payload(current),
+            "changes": {"added": [], "updated": [], "dismissed": [], "removed": []},
+        }
+
+    previous = get_version_by_hash(db, client_hash)
+    if not previous:
+        return _full_sync_payload(db, current, include_data=include_data, reason="unknown_client_hash", from_hash=client_hash)
+
+    return _delta_sync_payload(db, previous, current, include_data=include_data)
+
+
+def _full_sync_payload(
+    db: Session,
+    current: DatasetVersion,
+    *,
+    include_data: bool,
+    reason: str | None,
+    from_hash: str | None = None,
+) -> dict:
+    employees = db.scalars(select(Employee).order_by(Employee.profile_key)).all()
+    payload = {
+        "mode": "full",
+        "from_hash": from_hash,
+        "to_hash": current.hash,
+        "dataset": _version_payload(current),
+        "items": [_employee_payload(employee, include_data=include_data) for employee in employees],
+    }
+    if reason:
+        payload["reason"] = reason
+    return payload
+
+
+def _delta_sync_payload(db: Session, previous: DatasetVersion, current: DatasetVersion, *, include_data: bool) -> dict:
+    previous_items = _items_by_profile_key(previous)
+    current_items = _items_by_profile_key(current)
+    employees = {employee.profile_key: employee for employee in db.scalars(select(Employee)).all()}
+    added = []
+    updated = []
+    dismissed = []
+    removed = []
+
+    for profile_key, current_item in sorted(current_items.items()):
+        previous_item = previous_items.get(profile_key)
+        employee = employees.get(profile_key)
+        if not previous_item:
+            if employee:
+                added.append(_employee_payload(employee, include_data=include_data))
+            continue
+        if previous_item.checksum == current_item.checksum and previous_item.status == current_item.status:
+            continue
+        if current_item.status == "dismissed":
+            dismissed.append(_tombstone(profile_key, current_item.status, employee))
+        elif employee:
+            updated.append(_employee_payload(employee, include_data=include_data))
+
+    for profile_key, previous_item in sorted(previous_items.items()):
+        if profile_key not in current_items:
+            removed.append(_tombstone(profile_key, "removed", employees.get(profile_key), checksum=previous_item.checksum))
+
+    return {
+        "mode": "delta",
+        "from_hash": previous.hash,
+        "to_hash": current.hash,
+        "dataset": _version_payload(current),
+        "changes": {
+            "added": added,
+            "updated": updated,
+            "dismissed": dismissed,
+            "removed": removed,
+        },
+    }
+
+
+def _items_by_profile_key(version: DatasetVersion) -> dict[str, DatasetVersionItem]:
+    return {item.profile_key: item for item in version.items}
+
+
+def _version_payload(version: DatasetVersion) -> dict:
+    return {
+        "hash": version.hash,
+        "previous_hash": version.previous_hash,
+        "created_at": version.created_at.isoformat() if version.created_at else None,
+        "crawl_run_id": version.crawl_run_id,
+        "employee_count": version.employee_count,
+        "active_count": version.active_count,
+        "dismissed_count": version.dismissed_count,
+    }
+
+
+def _employee_marker(employee: Employee) -> EmployeeMarker:
+    return EmployeeMarker(
+        profile_key=employee.profile_key,
+        employee_id=employee.id,
+        status=employee.status,
+        checksum=employee.current_checksum or _payload_hash(employee.current_data or {}),
+    )
+
+
+def _dataset_hash(markers: list[EmployeeMarker]) -> str:
+    payload = [
+        {"profile_key": marker.profile_key, "status": marker.status, "checksum": marker.checksum}
+        for marker in sorted(markers, key=lambda item: item.profile_key)
+    ]
+    return _payload_hash(payload)
+
+
+def _payload_hash(value: object) -> str:
+    payload = json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(",", ":"), default=str)
+    return hashlib.sha256(payload.encode("utf-8")).hexdigest()
+
+
+def _employee_payload(employee: Employee, *, include_data: bool) -> dict:
+    payload = {
+        "profile_key": employee.profile_key,
+        "profile_id": employee.profile_id,
+        "full_name": employee.full_name,
+        "status": employee.status,
+        "canonical_url": employee.canonical_url,
+        "last_seen_at": employee.last_seen_at.isoformat() if employee.last_seen_at else None,
+        "dismissed_at": employee.dismissed_at.isoformat() if employee.dismissed_at else None,
+        "checksum": employee.current_checksum or _payload_hash(employee.current_data or {}),
+    }
+    if include_data:
+        payload["data"] = employee.current_data
+    return payload
+
+
+def _tombstone(profile_key: str, status: str, employee: Employee | None, *, checksum: str | None = None) -> dict:
+    payload = {
+        "profile_key": profile_key,
+        "status": status,
+        "checksum": checksum or (employee.current_checksum if employee else None),
+    }
+    if employee:
+        payload.update(
+            {
+                "profile_id": employee.profile_id,
+                "full_name": employee.full_name,
+                "canonical_url": employee.canonical_url,
+                "dismissed_at": employee.dismissed_at.isoformat() if employee.dismissed_at else None,
+            }
+        )
+    return payload
--- a/app/services/resource_cache.py
+++ b/app/services/resource_cache.py
@@ -0,0 +1,147 @@
+from __future__ import annotations
+
+import gzip
+import hashlib
+import json
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from typing import Any
+
+import requests
+from sqlalchemy import select
+from sqlalchemy.orm import Session
+
+from app.models import ParseResourceCache
+from app.version import BACKEND_VERSION
+
+
+@dataclass(frozen=True)
+class CachedResource:
+    text: str
+    body_hash: str
+    from_cache: bool
+    status_code: int
+
+
+class ResourceCache:
+    def __init__(self, db: Session):
+        self.db = db
+
+    def fetch_text(
+        self,
+        session: requests.Session,
+        *,
+        profile_key: str,
+        resource_key: str,
+        method: str,
+        url: str,
+        headers: dict[str, str],
+        timeout: int,
+        json_payload: Any | None = None,
+        params: dict[str, Any] | None = None,
+    ) -> CachedResource:
+        method = method.upper()
+        fingerprint = _request_fingerprint(method=method, url=url, json_payload=json_payload, params=params)
+        cached = self.db.scalar(
+            select(ParseResourceCache).where(
+                ParseResourceCache.profile_key == profile_key,
+                ParseResourceCache.resource_key == resource_key,
+                ParseResourceCache.request_fingerprint == fingerprint,
+            )
+        )
+
+        request_headers = dict(headers)
+        if cached:
+            if cached.etag:
+                request_headers["If-None-Match"] = cached.etag
+            if cached.last_modified:
+                request_headers["If-Modified-Since"] = cached.last_modified
+
+        response = _send(
+            session,
+            method=method,
+            url=url,
+            headers=request_headers,
+            timeout=timeout,
+            json_payload=json_payload,
+            params=params,
+        )
+        if response.status_code == 304 and cached:
+            cached.fetched_at = datetime.now(timezone.utc)
+            self.db.flush()
+            return CachedResource(
+                text=gzip.decompress(cached.body_snapshot).decode("utf-8"),
+                body_hash=cached.body_hash,
+                from_cache=True,
+                status_code=response.status_code,
+            )
+
+        response.raise_for_status()
+        text = response.text
+        body_hash = _body_hash(text)
+        etag = response.headers.get("ETag") if hasattr(response, "headers") else None
+        last_modified = response.headers.get("Last-Modified") if hasattr(response, "headers") else None
+
+        if cached:
+            cached.method = method
+            cached.url = url
+            cached.etag = etag
+            cached.last_modified = last_modified
+            cached.body_hash = body_hash
+            cached.body_snapshot = gzip.compress(text.encode("utf-8"))
+            cached.parser_version = BACKEND_VERSION
+            cached.fetched_at = datetime.now(timezone.utc)
+        else:
+            self.db.add(
+                ParseResourceCache(
+                    profile_key=profile_key,
+                    resource_key=resource_key,
+                    method=method,
+                    url=url,
+                    request_fingerprint=fingerprint,
+                    etag=etag,
+                    last_modified=last_modified,
+                    body_hash=body_hash,
+                    body_snapshot=gzip.compress(text.encode("utf-8")),
+                    parser_version=BACKEND_VERSION,
+                    fetched_at=datetime.now(timezone.utc),
+                )
+            )
+        self.db.flush()
+        return CachedResource(text=text, body_hash=body_hash, from_cache=False, status_code=response.status_code)
+
+
+def _send(
+    session: requests.Session,
+    *,
+    method: str,
+    url: str,
+    headers: dict[str, str],
+    timeout: int,
+    json_payload: Any | None,
+    params: dict[str, Any] | None,
+) -> requests.Response:
+    if method == "POST":
+        return session.post(url, json=json_payload, headers=headers, timeout=timeout, params=params)
+    return session.get(url, headers=headers, timeout=timeout, params=params)
+
+
+def _request_fingerprint(
+    *,
+    method: str,
+    url: str,
+    json_payload: Any | None,
+    params: dict[str, Any] | None,
+) -> str:
+    payload = {
+        "method": method,
+        "url": url,
+        "json": json_payload,
+        "params": params,
+    }
+    encoded = json.dumps(payload, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
+    return hashlib.sha256(encoded.encode("utf-8")).hexdigest()
+
+
+def _body_hash(text: str) -> str:
+    return hashlib.sha256(text.encode("utf-8")).hexdigest()
--- a/app/static/admin.css
+++ b/app/static/admin.css
@@ -171,6 +171,10 @@
  background: transparent;
 }

+.button--compact {
+  padding: 8px 12px;
+}
+
 .code {
  overflow-x: auto;
  padding: 14px;
@@ -201,11 +205,34 @@
  gap: 10px;
 }

+.employee-card__actions {
+  display: grid;
+  justify-items: end;
+  gap: 10px;
+}
+
 .employee-card__title {
  margin: 0;
  font-size: 24px;
 }

+.employee-card__notice {
+  margin: 0;
+  padding: 12px 14px;
+  border-radius: 8px;
+  font-weight: 700;
+}
+
+.employee-card__notice--success {
+  color: #065f46;
+  background: #d1fae5;
+}
+
+.employee-card__notice--error {
+  color: #991b1b;
+  background: #fee2e2;
+}
+
 .employee-card__section {
  padding: 20px;
  background: #ffffff;
--- a/app/static/admin.js
+++ b/app/static/admin.js
@@ -89,12 +89,14 @@
      const status = document.querySelector("[data-progress-status]");
      const processed = document.querySelector("[data-progress-processed]");
      const found = document.querySelector("[data-progress-found]");
+      const skipped = document.querySelector("[data-progress-skipped]");
      const errors = document.querySelector("[data-progress-errors]");
      const fill = document.querySelector("[data-progress-fill]");
      const percent = document.querySelector("[data-progress-percent]");
      if (status) status.textContent = run.status_display || run.status;
      if (processed) processed.textContent = run.processed_count;
      if (found) found.textContent = run.found_count;
+      if (skipped) skipped.textContent = run.skipped_count;
      if (errors) errors.textContent = run.error_count;
      if (fill) fill.style.width = `${run.progress_percent}%`;
      if (percent) percent.textContent = run.progress_percent;
--- a/app/templates/dashboard.html
+++ b/app/templates/dashboard.html
@@ -37,6 +37,7 @@
    <div class="progress-panel__meta">
      <span data-progress-status>{{ run.status_display if run else "Ожидание" }}</span>
      <span>обработано: <span data-progress-processed>{{ run.processed_count if run else 0 }}</span> / <span data-progress-found>{{ run.found_count if run else 0 }}</span></span>
+      <span>без изменений: <span data-progress-skipped>{{ run.skipped_count if run else 0 }}</span></span>
      <span>ошибок: <span data-progress-errors>{{ run.error_count if run else 0 }}</span></span>
    </div>
    <div class="progress-bar" aria-label="Parsing progress">
@@ -48,10 +49,10 @@
 <section class="panel">
  <h2 class="panel__title">Последние запуски</h2>
  <table class="table">
-    <thead><tr><th class="table__head">ID</th><th class="table__head">Статус</th><th class="table__head">Обработано</th><th class="table__head">Ошибки</th><th class="table__head">Старт</th></tr></thead>
+    <thead><tr><th class="table__head">ID</th><th class="table__head">Статус</th><th class="table__head">Обработано</th><th class="table__head">Без изменений</th><th class="table__head">Ошибки</th><th class="table__head">Старт</th></tr></thead>
    <tbody>
      {% for run in runs %}
-      <tr class="table__row" data-row-href="/admin/runs/{{ run.id }}" role="link" tabindex="0"><td class="table__cell">{{ run.id }}</td><td class="table__cell">{{ run.status_display }}</td><td class="table__cell">{{ run.parsed_count }}</td><td class="table__cell">{{ run.error_count }}</td><td class="table__cell">{{ run.started_display }}</td></tr>
+      <tr class="table__row" onclick="window.location.href='/admin/runs/{{ run.id }}'" onkeydown="if (event.key === 'Enter' || event.key === ' ') { event.preventDefault(); window.location.href='/admin/runs/{{ run.id }}'; }" role="link" tabindex="0"><td class="table__cell">{{ run.id }}</td><td class="table__cell">{{ run.status_display }}</td><td class="table__cell">{{ run.parsed_count }}</td><td class="table__cell">{{ run.skipped_count }}</td><td class="table__cell">{{ run.error_count }}</td><td class="table__cell">{{ run.started_display }}</td></tr>
      {% endfor %}
    </tbody>
  </table>
--- a/app/templates/employee_detail.html
+++ b/app/templates/employee_detail.html
@@ -7,8 +7,18 @@
      <h2 class="employee-card__title">{{ employee_view.full_name or employee.profile_key }}</h2>
      <span class="badge {% if employee_view.status == "dismissed" %}badge--dismissed{% endif %}">{{ employee_view.status_display }}</span>
    </div>
-    <a class="admin__link" href="{{ employee_view.canonical_url }}">{{ employee_view.canonical_url }}</a>
+    <div class="employee-card__actions">
+      <form method="post" action="/admin/employees/{{ employee.id }}/refresh">
+        <button class="button button--compact" type="submit">Обновить данные</button>
+      </form>
+      <a class="admin__link" href="{{ employee_view.canonical_url }}">{{ employee_view.canonical_url }}</a>
+    </div>
  </div>
+  {% if refresh_status == "success" %}
+  <p class="employee-card__notice employee-card__notice--success">Данные сотрудника обновлены.</p>
+  {% elif refresh_status == "error" %}
+  <p class="employee-card__notice employee-card__notice--error">Не удалось обновить данные сотрудника.</p>
+  {% endif %}

  <section class="employee-card__section">
    <h3 class="employee-section__title">Основная информация</h3>
--- a/app/templates/run_detail.html
+++ b/app/templates/run_detail.html
@@ -12,6 +12,7 @@
  <div class="stats-strip">
    <div class="stats-strip__item"><span class="stats-strip__label">Найдено</span><span class="stats-strip__value">{{ run.found_count }}</span></div>
    <div class="stats-strip__item"><span class="stats-strip__label">Обработано</span><span class="stats-strip__value">{{ run.parsed_count }}</span></div>
+    <div class="stats-strip__item"><span class="stats-strip__label">Без изменений</span><span class="stats-strip__value">{{ run.skipped_count }}</span></div>
    <div class="stats-strip__item"><span class="stats-strip__label">Новые</span><span class="stats-strip__value">{{ run.new_count }}</span></div>
    <div class="stats-strip__item"><span class="stats-strip__label">Потеряшки</span><span class="stats-strip__value">{{ run.changes.missing_from_source | length }}</span></div>
    <div class="stats-strip__item"><span class="stats-strip__label">Уволены</span><span class="stats-strip__value">{{ run.dismissed_count }}</span></div>
--- a/app/templates/runs.html
+++ b/app/templates/runs.html
@@ -8,12 +8,13 @@
  </div>
  {% set run = runs[0] if runs else none %}
  {% if run %}
-  {% set processed = run.parsed_count + run.error_count %}
+  {% set processed = run.parsed_count + run.skipped_count + run.error_count %}
  {% set percent = ((processed / run.found_count) * 100) | round(1) if run.found_count else 0 %}
  <div class="progress-panel" data-progress-panel>
    <div class="progress-panel__meta">
      <span data-progress-status>{{ run.status_display }}</span>
      <span>обработано: <span data-progress-processed>{{ processed }}</span> / <span data-progress-found>{{ run.found_count }}</span></span>
+      <span>без изменений: <span data-progress-skipped>{{ run.skipped_count }}</span></span>
      <span>ошибок: <span data-progress-errors>{{ run.error_count }}</span></span>
    </div>
    <div class="progress-bar" aria-label="Parsing progress">
@@ -26,6 +27,7 @@
    <div class="progress-panel__meta">
      <span data-progress-status>Ожидание</span>
      <span>обработано: <span data-progress-processed>0</span> / <span data-progress-found>0</span></span>
+      <span>без изменений: <span data-progress-skipped>0</span></span>
      <span>ошибок: <span data-progress-errors>0</span></span>
    </div>
    <div class="progress-bar" aria-label="Parsing progress">
@@ -35,10 +37,10 @@
  </div>
  {% endif %}
  <table class="table">
-    <thead><tr><th class="table__head">ID</th><th class="table__head">Статус</th><th class="table__head">Найдено</th><th class="table__head">Обработано</th><th class="table__head">Новые</th><th class="table__head">Ошибки</th><th class="table__head">Уволены</th><th class="table__head">Старт</th></tr></thead>
+    <thead><tr><th class="table__head">ID</th><th class="table__head">Статус</th><th class="table__head">Найдено</th><th class="table__head">Обработано</th><th class="table__head">Без изменений</th><th class="table__head">Новые</th><th class="table__head">Ошибки</th><th class="table__head">Уволены</th><th class="table__head">Старт</th></tr></thead>
    <tbody>
      {% for run in runs %}
-      <tr class="table__row" data-row-href="/admin/runs/{{ run.id }}" role="link" tabindex="0"><td class="table__cell">{{ run.id }}</td><td class="table__cell">{{ run.status_display }}</td><td class="table__cell">{{ run.found_count }}</td><td class="table__cell">{{ run.parsed_count }}</td><td class="table__cell">{{ run.new_count }}</td><td class="table__cell">{{ run.error_count }}</td><td class="table__cell">{{ run.dismissed_count }}</td><td class="table__cell">{{ run.started_display }}</td></tr>
+      <tr class="table__row" onclick="window.location.href='/admin/runs/{{ run.id }}'" onkeydown="if (event.key === 'Enter' || event.key === ' ') { event.preventDefault(); window.location.href='/admin/runs/{{ run.id }}'; }" role="link" tabindex="0"><td class="table__cell">{{ run.id }}</td><td class="table__cell">{{ run.status_display }}</td><td class="table__cell">{{ run.found_count }}</td><td class="table__cell">{{ run.parsed_count }}</td><td class="table__cell">{{ run.skipped_count }}</td><td class="table__cell">{{ run.new_count }}</td><td class="table__cell">{{ run.error_count }}</td><td class="table__cell">{{ run.dismissed_count }}</td><td class="table__cell">{{ run.started_display }}</td></tr>
      {% endfor %}
    </tbody>
  </table>
--- a/app/version.py
+++ b/app/version.py
@@ -1,3 +1,3 @@
-APP_VERSION = "0.4.3"
-FRONTEND_VERSION = "0.4.3"
-BACKEND_VERSION = "0.4.3"
+APP_VERSION = "0.6.0"
+FRONTEND_VERSION = "0.6.0"
+BACKEND_VERSION = "0.6.0"
--- a/app/worker.py
+++ b/app/worker.py
@@ -17,7 +17,14 @@ def crawl_once() -> None:
    settings = get_settings()
    with SessionLocal() as db:
        run = run_crawl(db, settings)
-        logger.info("crawl finished: id=%s status=%s parsed=%s errors=%s", run.id, run.status, run.parsed_count, run.error_count)
+        logger.info(
+            "crawl finished: id=%s status=%s parsed=%s skipped=%s errors=%s",
+            run.id,
+            run.status,
+            run.parsed_count,
+            run.skipped_count,
+            run.error_count,
+        )


 def main() -> None:
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -20,7 +20,7 @@ services:
    environment:
      DATABASE_URL: postgresql+psycopg://${POSTGRES_USER:-miem}:${POSTGRES_PASSWORD:-miem_password}@postgres:5432/${POSTGRES_DB:-miem_workers}
    ports:
-      - "127.0.0.1:8000:8000"
+      - "127.0.0.1:${API_PORT:-8000}:8000"
    depends_on:
      postgres:
        condition: service_healthy
@@ -42,33 +42,7 @@ services:
    environment:
      DATABASE_URL: postgresql+psycopg://${POSTGRES_USER:-miem}:${POSTGRES_PASSWORD:-miem_password}@postgres:5432/${POSTGRES_DB:-miem_workers}
    ports:
-      - "127.0.0.1:8001:8000"
-    depends_on:
-      postgres:
-        condition: service_healthy
-
-  keycloak:
-    image: quay.io/keycloak/keycloak:latest
-    container_name: keycloak
-    restart: unless-stopped
-    environment:
-      KC_DB: postgres
-      KC_DB_URL: jdbc:postgresql://postgres:5432/${KEYCLOAK_DB_NAME}
-      KC_DB_USERNAME: ${KEYCLOAK_DB_USER}
-      KC_DB_PASSWORD: ${KEYCLOAK_DB_PASSWORD}
-
-      KEYCLOAK_ADMIN: ${KEYCLOAK_ADMIN}
-      KEYCLOAK_ADMIN_PASSWORD: ${KEYCLOAK_ADMIN_PASSWORD}
-
-      KC_HTTP_ENABLED: true
-      KC_PROXY_HEADERS: xforwarded
-      KC_HOSTNAME: ${KEYCLOAK_HOSTNAME}
-
-      KC_HEALTH_ENABLED: true
-      KC_METRICS_ENABLED: true
-    command: start
-    ports:
-      - "127.0.0.1:8080:8080"
+      - "127.0.0.1:${MCP_PORT:-8001}:8000"
    depends_on:
      postgres:
        condition: service_healthy
--- a/migrations/001_init.sql
+++ b/migrations/001_init.sql
@@ -13,6 +13,7 @@ CREATE TABLE IF NOT EXISTS crawl_runs (
  finished_at TIMESTAMPTZ,
  found_count INTEGER NOT NULL DEFAULT 0,
  parsed_count INTEGER NOT NULL DEFAULT 0,
+  skipped_count INTEGER NOT NULL DEFAULT 0,
  new_count INTEGER NOT NULL DEFAULT 0,
  error_count INTEGER NOT NULL DEFAULT 0,
  dismissed_count INTEGER NOT NULL DEFAULT 0,
@@ -73,3 +74,22 @@ CREATE TABLE IF NOT EXISTS profile_tabs (
 );

 CREATE INDEX IF NOT EXISTS ix_profile_tabs_employee_id ON profile_tabs (employee_id);
+
+CREATE TABLE IF NOT EXISTS parse_resource_cache (
+  id SERIAL PRIMARY KEY,
+  profile_key VARCHAR(255) NOT NULL,
+  resource_key VARCHAR(255) NOT NULL,
+  method VARCHAR(16) NOT NULL,
+  url TEXT NOT NULL,
+  request_fingerprint VARCHAR(64) NOT NULL,
+  etag TEXT,
+  last_modified TEXT,
+  body_hash VARCHAR(64) NOT NULL,
+  body_snapshot BYTEA NOT NULL,
+  parser_version VARCHAR(32),
+  fetched_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  CONSTRAINT uq_parse_resource_cache_resource UNIQUE (profile_key, resource_key, request_fingerprint)
+);
+
+CREATE INDEX IF NOT EXISTS ix_parse_resource_cache_profile_key
+  ON parse_resource_cache (profile_key);
--- a/migrations/004_dataset_versions.sql
+++ b/migrations/004_dataset_versions.sql
@@ -0,0 +1,29 @@
+CREATE TABLE IF NOT EXISTS dataset_versions (
+  id SERIAL PRIMARY KEY,
+  hash VARCHAR(64) NOT NULL UNIQUE,
+  previous_hash VARCHAR(64),
+  crawl_run_id INTEGER REFERENCES crawl_runs(id),
+  employee_count INTEGER NOT NULL DEFAULT 0,
+  active_count INTEGER NOT NULL DEFAULT 0,
+  dismissed_count INTEGER NOT NULL DEFAULT 0,
+  created_at TIMESTAMPTZ NOT NULL DEFAULT now()
+);
+
+CREATE INDEX IF NOT EXISTS ix_dataset_versions_created_at
+  ON dataset_versions (created_at);
+
+CREATE TABLE IF NOT EXISTS dataset_version_items (
+  id SERIAL PRIMARY KEY,
+  dataset_version_id INTEGER NOT NULL REFERENCES dataset_versions(id),
+  profile_key VARCHAR(255) NOT NULL,
+  employee_id INTEGER REFERENCES employees(id),
+  status VARCHAR(32) NOT NULL,
+  checksum VARCHAR(64) NOT NULL,
+  CONSTRAINT uq_dataset_version_items_version_profile UNIQUE (dataset_version_id, profile_key)
+);
+
+CREATE INDEX IF NOT EXISTS ix_dataset_version_items_hash
+  ON dataset_version_items (dataset_version_id);
+
+CREATE INDEX IF NOT EXISTS ix_dataset_version_items_profile_key
+  ON dataset_version_items (profile_key);
--- a/migrations/005_parse_resource_cache.sql
+++ b/migrations/005_parse_resource_cache.sql
@@ -0,0 +1,21 @@
+ALTER TABLE crawl_runs
+ADD COLUMN IF NOT EXISTS skipped_count INTEGER NOT NULL DEFAULT 0;
+
+CREATE TABLE IF NOT EXISTS parse_resource_cache (
+  id SERIAL PRIMARY KEY,
+  profile_key VARCHAR(255) NOT NULL,
+  resource_key VARCHAR(255) NOT NULL,
+  method VARCHAR(16) NOT NULL,
+  url TEXT NOT NULL,
+  request_fingerprint VARCHAR(64) NOT NULL,
+  etag TEXT,
+  last_modified TEXT,
+  body_hash VARCHAR(64) NOT NULL,
+  body_snapshot BYTEA NOT NULL,
+  parser_version VARCHAR(32),
+  fetched_at TIMESTAMPTZ NOT NULL DEFAULT now(),
+  CONSTRAINT uq_parse_resource_cache_resource UNIQUE (profile_key, resource_key, request_fingerprint)
+);
+
+CREATE INDEX IF NOT EXISTS ix_parse_resource_cache_profile_key
+  ON parse_resource_cache (profile_key);
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -1,6 +1,6 @@
 [project]
 name = "miem-workers"
-version = "0.4.0"
+version = "0.6.0"
 description = "MIEM employees parser, admin API, and MCP server"
 requires-python = ">=3.11"
 dependencies = [
@@ -12,7 +12,6 @@ dependencies = [
  "lxml>=5.2.0",
  "psycopg[binary]>=3.2.0",
  "pydantic-settings>=2.4.0",
-  "PyJWT[crypto]>=2.9.0",
  "python-multipart>=0.0.9",
  "requests>=2.32.0",
  "sqlalchemy>=2.0.32",
--- a/requirements.txt
+++ b/requirements.txt
@@ -6,7 +6,6 @@ jinja2>=3.1.4
 lxml>=5.2.0
 psycopg[binary]>=3.2.0
 pydantic-settings>=2.4.0
-PyJWT[crypto]>=2.9.0
 python-multipart>=0.0.9
 requests>=2.32.0
 sqlalchemy>=2.0.32
--- a/tests/test_admin_data.py
+++ b/tests/test_admin_data.py
@@ -200,13 +200,14 @@ def test_run_payload_calculates_progress():
        status="running",
        found_count=10,
        parsed_count=4,
+        skipped_count=2,
        error_count=1,
    )

    payload = run_payload(run)

-    assert payload["processed_count"] == 5
-    assert payload["progress_percent"] == 50.0
+    assert payload["processed_count"] == 7
+    assert payload["progress_percent"] == 70.0
    assert payload["status_display"] == "Выполняется"


--- a/tests/test_admin_templates.py
+++ b/tests/test_admin_templates.py
@@ -45,9 +45,11 @@ def test_dashboard_limits_latest_runs_to_five():
 def test_runs_template_links_to_run_detail():
    template = Path("app/templates/runs.html").read_text(encoding="utf-8")

-    assert 'data-row-href="/admin/runs/{{ run.id }}"' in template
+    assert 'onclick="window.location.href=\'/admin/runs/{{ run.id }}\'"' in template
+    assert "onkeydown=\"if (event.key === 'Enter' || event.key === ' ')" in template
    assert 'role="link"' in template
    assert 'tabindex="0"' in template
+    assert 'data-row-href="/admin/runs/{{ run.id }}"' not in template
    assert '<a class="admin__link" href="/admin/runs/{{ run.id }}">' not in template


@@ -75,9 +77,11 @@ def test_dashboard_metric_cards_link_to_admin_targets():
 def test_dashboard_latest_run_rows_link_to_run_detail():
    template = Path("app/templates/dashboard.html").read_text(encoding="utf-8")

-    assert 'data-row-href="/admin/runs/{{ run.id }}"' in template
+    assert 'onclick="window.location.href=\'/admin/runs/{{ run.id }}\'"' in template
+    assert "onkeydown=\"if (event.key === 'Enter' || event.key === ' ')" in template
    assert 'role="link"' in template
    assert 'tabindex="0"' in template
+    assert 'data-row-href="/admin/runs/{{ run.id }}"' not in template
    assert '<a class="admin__link" href="/admin/runs/{{ run.id }}">' not in template


--- a/tests/test_api_mcp.py
+++ b/tests/test_api_mcp.py
@@ -1,15 +1,12 @@
-import time
+import json
 from datetime import datetime, timezone
 from types import SimpleNamespace

-import jwt
 from fastapi.testclient import TestClient
-from cryptography.hazmat.primitives.asymmetric import rsa
-from sqlalchemy import create_engine
+from sqlalchemy import create_engine, select
 from sqlalchemy.orm import sessionmaker
 from sqlalchemy.pool import StaticPool

-import app.security as security
 from app.config import Settings, get_settings
 from app.db import Base, get_db
 from app.main import app
@@ -23,10 +20,10 @@ def test_health_returns_versions():
    response = client.get("/api/health")

    assert response.status_code == 200
-    assert response.json()["backend_version"] == "0.4.3"
+    assert response.json()["backend_version"] == "0.6.0"


-def test_mcp_requires_token_and_lists_tools():
+def test_mcp_lists_tools_without_auth_and_ignores_auth_header():
    engine = create_engine(
        "sqlite:///:memory:",
        connect_args={"check_same_thread": False},
@@ -43,22 +40,23 @@ def test_mcp_requires_token_and_lists_tools():
            session.close()

    app.dependency_overrides[get_db] = override_db
-    app.dependency_overrides[get_settings] = lambda: Settings(
-        mcp_auth_mode="token", mcp_token="secret", session_secret="session-secret"
-    )
    client = TestClient(app)

-    unauthorized = client.post("/mcp", json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}})
-    authorized = client.post(
+    without_auth = client.post("/mcp", json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}})
+    with_auth = client.post(
        "/mcp",
-        headers={"Authorization": "Bearer secret"},
+        headers={"Authorization": "Bearer anything"},
        json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}},
    )

-    assert unauthorized.status_code == 401
-    assert authorized.status_code == 200
-    assert authorized.json()["result"]["tools"][0]["name"] == "search_employees"
-    assert any(tool["name"] == "get_crawl_run_details" for tool in authorized.json()["result"]["tools"])
+    assert without_auth.status_code == 200
+    assert with_auth.status_code == 200
+    tool_names = {tool["name"] for tool in without_auth.json()["result"]["tools"]}
+    assert "search_employees" in tool_names
+    assert "get_service_info" in tool_names
+    assert "sync_employees" in tool_names
+    assert any(tool["name"] == "get_crawl_run_details" for tool in without_auth.json()["result"]["tools"])
+    assert with_auth.json()["result"]["tools"] == without_auth.json()["result"]["tools"]

    app.dependency_overrides.clear()

@@ -96,14 +94,10 @@ def test_mcp_search_employees_returns_matching_employee():
            db.close()

    app.dependency_overrides[get_db] = override_db
-    app.dependency_overrides[get_settings] = lambda: Settings(
-        mcp_auth_mode="token", mcp_token="secret", session_secret="session-secret"
-    )
    client = TestClient(app)

    response = client.post(
        "/mcp",
-        headers={"Authorization": "Bearer secret"},
        json={
            "jsonrpc": "2.0",
            "id": 1,
@@ -118,6 +112,128 @@ def test_mcp_search_employees_returns_matching_employee():
    app.dependency_overrides.clear()


+def test_mcp_service_info_returns_tools_and_dataset_hash():
+    engine = create_engine(
+        "sqlite:///:memory:",
+        connect_args={"check_same_thread": False},
+        poolclass=StaticPool,
+    )
+    Base.metadata.create_all(engine)
+    Session = sessionmaker(bind=engine)
+    session = Session()
+    session.add(
+        Employee(
+            profile_key="staff:alpha",
+            profile_type="staff",
+            profile_id="alpha",
+            canonical_url="https://www.hse.ru/staff/alpha",
+            full_name="Alpha Person",
+            status="active",
+            current_checksum="a" * 64,
+            current_data={"sections": []},
+        )
+    )
+    session.commit()
+    session.close()
+
+    def override_db():
+        db = Session()
+        try:
+            yield db
+        finally:
+            db.close()
+
+    app.dependency_overrides[get_db] = override_db
+    client = TestClient(app)
+
+    response = client.post(
+        "/mcp",
+        json={"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "get_service_info", "arguments": {}}},
+    )
+
+    assert response.status_code == 200
+    payload = json.loads(response.json()["result"]["content"][0]["text"])
+    assert payload["service_name"] == "miem-employees"
+    assert payload["backend_version"] == "0.6.0"
+    assert payload["dataset"]["hash"]
+    assert any(tool["name"] == "sync_employees" for tool in payload["tools"])
+
+    app.dependency_overrides.clear()
+
+
+def test_mcp_sync_employees_full_empty_and_unknown_hash_modes():
+    engine = create_engine(
+        "sqlite:///:memory:",
+        connect_args={"check_same_thread": False},
+        poolclass=StaticPool,
+    )
+    Base.metadata.create_all(engine)
+    Session = sessionmaker(bind=engine)
+    session = Session()
+    session.add(
+        Employee(
+            profile_key="staff:alpha",
+            profile_type="staff",
+            profile_id="alpha",
+            canonical_url="https://www.hse.ru/staff/alpha",
+            full_name="Alpha Person",
+            status="active",
+            current_checksum="a" * 64,
+            current_data={"sections": [{"type": "paragraphs"}]},
+        )
+    )
+    session.commit()
+    session.close()
+
+    def override_db():
+        db = Session()
+        try:
+            yield db
+        finally:
+            db.close()
+
+    app.dependency_overrides[get_db] = override_db
+    client = TestClient(app)
+
+    full_response = client.post(
+        "/mcp",
+        json={"jsonrpc": "2.0", "id": 1, "method": "tools/call", "params": {"name": "sync_employees", "arguments": {}}},
+    )
+    full_payload = json.loads(full_response.json()["result"]["content"][0]["text"])
+    current_hash = full_payload["to_hash"]
+
+    empty_response = client.post(
+        "/mcp",
+        json={
+            "jsonrpc": "2.0",
+            "id": 2,
+            "method": "tools/call",
+            "params": {"name": "sync_employees", "arguments": {"client_hash": current_hash}},
+        },
+    )
+    empty_payload = json.loads(empty_response.json()["result"]["content"][0]["text"])
+
+    unknown_response = client.post(
+        "/mcp",
+        json={
+            "jsonrpc": "2.0",
+            "id": 3,
+            "method": "tools/call",
+            "params": {"name": "sync_employees", "arguments": {"client_hash": "missing"}},
+        },
+    )
+    unknown_payload = json.loads(unknown_response.json()["result"]["content"][0]["text"])
+
+    assert full_payload["mode"] == "full"
+    assert full_payload["items"][0]["data"] == {"sections": [{"type": "paragraphs"}]}
+    assert empty_payload["mode"] == "delta"
+    assert empty_payload["changes"] == {"added": [], "updated": [], "dismissed": [], "removed": []}
+    assert unknown_payload["mode"] == "full"
+    assert unknown_payload["reason"] == "unknown_client_hash"
+
+    app.dependency_overrides.clear()
+
+
 def test_mcp_get_crawl_run_details_returns_changes():
    engine = create_engine(
        "sqlite:///:memory:",
@@ -164,14 +280,10 @@ def test_mcp_get_crawl_run_details_returns_changes():
            db.close()

    app.dependency_overrides[get_db] = override_db
-    app.dependency_overrides[get_settings] = lambda: Settings(
-        mcp_auth_mode="token", mcp_token="secret", session_secret="session-secret"
-    )
    client = TestClient(app)

    response = client.post(
        "/mcp",
-        headers={"Authorization": "Bearer secret"},
        json={
            "jsonrpc": "2.0",
            "id": 1,
@@ -188,146 +300,12 @@ def test_mcp_get_crawl_run_details_returns_changes():
    app.dependency_overrides.clear()


-def test_mcp_oauth_rejects_static_token():
-    engine = create_engine(
-        "sqlite:///:memory:",
-        connect_args={"check_same_thread": False},
-        poolclass=StaticPool,
-    )
-    Base.metadata.create_all(engine)
-    Session = sessionmaker(bind=engine)
-
-    def override_db():
-        session = Session()
-        try:
-            yield session
-        finally:
-            session.close()
-
-    settings = Settings(
-        mcp_auth_mode="oauth",
-        mcp_token="secret",
-        session_secret="session-secret",
-        mcp_oauth_issuer="https://auth.example.com",
-        mcp_oauth_audience="miem-mcp",
-        mcp_oauth_jwks_url="https://auth.example.com/.well-known/jwks.json",
-    )
-    app.dependency_overrides[get_db] = override_db
-    app.dependency_overrides[get_settings] = lambda: settings
-    client = TestClient(app)
-
-    response = client.post(
-        "/mcp",
-        headers={"Authorization": "Bearer secret"},
-        json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}},
-    )
-
-    assert response.status_code == 401
-    assert response.headers["www-authenticate"] == (
-        'Bearer resource_metadata="http://localhost:8001/.well-known/oauth-protected-resource"'
-    )
-
-    app.dependency_overrides.clear()
-
-
-def test_mcp_oauth_missing_auth_returns_metadata_challenge():
-    settings = Settings(
-        mcp_auth_mode="oauth",
-        mcp_resource_url="https://api.example.com/mcp",
-        mcp_oauth_issuer="https://auth.example.com",
-        mcp_oauth_audience="miem-mcp",
-        mcp_oauth_jwks_url="https://auth.example.com/.well-known/jwks.json",
-    )
-    app.dependency_overrides[get_settings] = lambda: settings
-    client = TestClient(app)
-
-    response = client.post("/mcp", json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}})
-
-    assert response.status_code == 401
-    assert response.headers["www-authenticate"] == (
-        'Bearer resource_metadata="https://api.example.com/.well-known/oauth-protected-resource"'
-    )
-
-    app.dependency_overrides.clear()
-
-
-def test_mcp_accepts_valid_oauth_jwt(monkeypatch):
-    public_key, token = _oauth_key_and_token()
-    monkeypatch.setattr(security, "_get_mcp_oauth_signing_key", lambda _token, _settings: SimpleNamespace(key=public_key))
-    app.dependency_overrides[get_settings] = lambda: _oauth_settings()
-    client = TestClient(app)
-
-    response = client.post(
-        "/mcp",
-        headers={"Authorization": f"Bearer {token}"},
-        json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}},
-    )
-
-    assert response.status_code == 200
-    assert response.json()["result"]["tools"][0]["name"] == "search_employees"
-
-    app.dependency_overrides.clear()
-
-
-def test_mcp_rejects_invalid_oauth_jwts(monkeypatch):
-    public_key, expired_token = _oauth_key_and_token(exp=int(time.time()) - 60)
-    _, wrong_issuer_token = _oauth_key_and_token(issuer="https://other.example.com")
-    _, wrong_audience_token = _oauth_key_and_token(audience="other-audience")
-    _, bad_signature_token = _oauth_key_and_token(public_key=public_key)
-    monkeypatch.setattr(security, "_get_mcp_oauth_signing_key", lambda _token, _settings: SimpleNamespace(key=public_key))
-    app.dependency_overrides[get_settings] = lambda: _oauth_settings()
-    client = TestClient(app)
-
-    for token in [expired_token, wrong_issuer_token, wrong_audience_token, bad_signature_token]:
-        response = client.post(
-            "/mcp",
-            headers={"Authorization": f"Bearer {token}"},
-            json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}},
-        )
-
-        assert response.status_code == 401
-
-    app.dependency_overrides.clear()
-
-
-def test_mcp_rejects_oauth_jwt_without_required_scope(monkeypatch):
-    public_key, token = _oauth_key_and_token(scope="profile")
-    monkeypatch.setattr(security, "_get_mcp_oauth_signing_key", lambda _token, _settings: SimpleNamespace(key=public_key))
-    app.dependency_overrides[get_settings] = lambda: _oauth_settings()
-    client = TestClient(app)
-
-    response = client.post(
-        "/mcp",
-        headers={"Authorization": f"Bearer {token}"},
-        json={"jsonrpc": "2.0", "id": 1, "method": "tools/list", "params": {}},
-    )
-
-    assert response.status_code == 403
-
-    app.dependency_overrides.clear()
-
-
-def test_mcp_protected_resource_metadata_uses_settings():
-    settings = Settings(
-        mcp_resource_url="https://api.example.com/mcp",
-        mcp_oauth_issuer="https://auth.example.com/",
-        mcp_oauth_required_scope="mcp:tools",
-    )
-    app.dependency_overrides[get_settings] = lambda: settings
+def test_mcp_protected_resource_metadata_route_is_removed():
    client = TestClient(app)

    response = client.get("/.well-known/oauth-protected-resource")

-    assert response.status_code == 200
-    assert response.json() == {
-        "resource": "https://api.example.com/mcp",
-        "authorization_servers": ["https://auth.example.com"],
-        "bearer_methods_supported": ["header"],
-        "scopes_supported": ["mcp:tools"],
-        "resource_documentation": "https://api.example.com/mcp",
-    }
-
-    app.dependency_overrides.clear()
+    assert response.status_code == 404


 def test_api_employees_and_stats_require_admin_session():
@@ -399,33 +377,54 @@ def test_api_employees_and_stats_require_admin_session():
    app.dependency_overrides.clear()


-def _oauth_settings() -> Settings:
-    return Settings(
-        mcp_auth_mode="oauth",
-        mcp_resource_url="https://api.example.com/mcp",
-        mcp_oauth_issuer="https://auth.example.com",
-        mcp_oauth_audience="miem-mcp",
-        mcp_oauth_jwks_url="https://auth.example.com/.well-known/jwks.json",
-        session_secret="session-secret",
+def test_admin_refresh_employee_route_updates_only_requested_employee(monkeypatch):
+    engine = create_engine(
+        "sqlite:///:memory:",
+        connect_args={"check_same_thread": False},
+        poolclass=StaticPool,
    )
+    Base.metadata.create_all(engine)
+    Session = sessionmaker(bind=engine)
+    db = Session()
+    db.add(
+        Employee(
+            profile_key="org_person:133709486",
+            profile_type="org_person",
+            profile_id="133709486",
+            canonical_url="https://www.hse.ru/org/persons/133709486",
+            full_name="Будков Юрий Алексеевич",
+            status="active",
+        )
+    )
+    db.commit()
+    employee_id = db.scalar(select(Employee.id))
+    db.close()

+    settings = Settings(admin_username="admin", admin_password="password", session_secret="session-secret")

-def _oauth_key_and_token(
-    *,
-    issuer: str = "https://auth.example.com",
-    audience: str = "miem-mcp",
-    scope: str = "mcp:tools",
-    exp: int | None = None,
-    public_key=None,
-):
-    private_key = rsa.generate_private_key(public_exponent=65537, key_size=2048)
-    claims = {
-        "iss": issuer,
-        "aud": audience,
-        "scope": scope,
-        "sub": "mcp-client",
-        "iat": int(time.time()),
-        "exp": exp or int(time.time()) + 300,
-    }
-    token = jwt.encode(claims, private_key, algorithm="RS256", headers={"kid": "test-key"})
-    return public_key or private_key.public_key(), token
+    def override_db():
+        session = Session()
+        try:
+            yield session
+        finally:
+            session.close()
+
+    calls = []
+
+    def fake_refresh_employee(db, refreshed_employee, route_settings):
+        calls.append((refreshed_employee.id, route_settings))
+        return SimpleNamespace(status="completed")
+
+    app.dependency_overrides[get_db] = override_db
+    app.dependency_overrides[get_settings] = lambda: settings
+    monkeypatch.setattr("app.admin.refresh_employee", fake_refresh_employee)
+    client = TestClient(app)
+    client.cookies.set(SESSION_COOKIE, sign_session("admin", settings))
+
+    response = client.post(f"/admin/employees/{employee_id}/refresh", follow_redirects=False)
+
+    assert response.status_code == 303
+    assert response.headers["location"] == f"/admin/employees/{employee_id}?refresh_status=success"
+    assert calls == [(employee_id, settings)]
+
+    app.dependency_overrides.clear()
--- a/tests/test_config.py
+++ b/tests/test_config.py
@@ -1,6 +1,3 @@
-import pytest
-from pydantic import ValidationError
-
 from app.config import Settings


@@ -14,8 +11,3 @@ def test_numeric_crawl_limit_is_parsed():
    settings = Settings(crawl_limit="25")

    assert settings.crawl_limit == 25
-
-
-def test_mcp_auth_mode_rejects_oauth_or_token_fallback():
-    with pytest.raises(ValidationError):
-        Settings(mcp_auth_mode="oauth_or_token")
--- a/tests/test_crawler.py
+++ b/tests/test_crawler.py
@@ -1,7 +1,9 @@
+import gzip
 from datetime import datetime, timezone

-from app.models import CrawlRun, CrawlRunEmployeeChange, Employee
-from app.services.crawler import _mark_dismissed, _upsert_employee
+from app.models import CrawlRun, CrawlRunEmployeeChange, Employee, EmployeeSnapshot, ParseResourceCache
+from app.services.crawler import _checksum, _mark_dismissed, _upsert_employee
+from app.services.resource_cache import ResourceCache


 class FakeResponse:
@@ -17,6 +19,34 @@ class FakeSession:
        return FakeResponse(self.statuses[url])


+class ConditionalResponse:
+    def __init__(self, status_code, text="", headers=None):
+        self.status_code = status_code
+        self._text = text
+        self.headers = headers or {}
+        self.text_read = False
+
+    @property
+    def text(self):
+        self.text_read = True
+        return self._text
+
+    def raise_for_status(self):
+        return None
+
+
+class ConditionalSession:
+    def __init__(self):
+        self.requests = []
+        self.not_modified_response = ConditionalResponse(304)
+
+    def get(self, url, **kwargs):
+        self.requests.append((url, kwargs))
+        if kwargs["headers"].get("If-None-Match") == '"cached"':
+            return self.not_modified_response
+        return ConditionalResponse(200, "fresh", {"ETag": '"fresh"'})
+
+
 def test_mark_dismissed_records_missing_source_when_profile_is_available(db_session):
    run = CrawlRun(source_url="https://miem.hse.ru/persons", status="running")
    db_session.add(run)
@@ -111,3 +141,86 @@ def test_upsert_employee_increments_new_count_and_records_change_for_new_employe
    change = db_session.query(CrawlRunEmployeeChange).one()
    assert change.change_type == "new"
    assert change.full_name == "New Person"
+
+
+def test_resource_cache_uses_etag_and_reuses_cached_body_on_304(db_session):
+    db_session.add(
+        ParseResourceCache(
+            profile_key="staff:cached",
+            resource_key="main-html",
+            method="GET",
+            url="https://www.hse.ru/staff/cached",
+            request_fingerprint="020d59db7b358d9023d0f185bcbf5a9c085d3cf2bf91d92d48eee9147e8d0f01",
+            etag='"cached"',
+            body_hash="cached-hash",
+            body_snapshot=gzip.compress("cached body".encode("utf-8")),
+            parser_version="0.6.0",
+        )
+    )
+    db_session.commit()
+    session = ConditionalSession()
+
+    result = ResourceCache(db_session).fetch_text(
+        session,
+        profile_key="staff:cached",
+        resource_key="main-html",
+        method="GET",
+        url="https://www.hse.ru/staff/cached",
+        headers={"User-Agent": "test"},
+        timeout=10,
+    )
+
+    assert session.requests[0][1]["headers"]["If-None-Match"] == '"cached"'
+    assert result.text == "cached body"
+    assert result.from_cache is True
+    assert session.not_modified_response.text_read is False
+
+
+def test_upsert_employee_skips_snapshot_when_checksum_is_unchanged(db_session):
+    first_run = CrawlRun(source_url="https://miem.hse.ru/persons", status="running")
+    second_run = CrawlRun(source_url="https://miem.hse.ru/persons", status="running")
+    db_session.add_all([first_run, second_run])
+    db_session.commit()
+
+    _, first_changed = _upsert_employee(db_session, first_run, _parsed_employee("same"))
+    _, second_changed = _upsert_employee(db_session, second_run, _parsed_employee("same"))
+    db_session.commit()
+
+    assert first_changed is True
+    assert second_changed is False
+    assert db_session.query(EmployeeSnapshot).count() == 1
+
+
+def test_checksum_changes_when_widget_data_changes():
+    base = _parsed_employee("widgets")
+    changed = _parsed_employee("widgets")
+    changed["sections"] = [
+        {
+            "type": "publications",
+            "publications": [{"id": "1", "title": "New publication"}],
+        }
+    ]
+
+    assert _checksum(base) != _checksum(changed)
+
+
+def test_checksum_ignores_date_dependent_experience_text():
+    first = _parsed_employee("experience")
+    second = _parsed_employee("experience")
+    first["sections"] = [{"raw_text": "Стаж работы в НИУ ВШЭ: 5 лет"}]
+    second["sections"] = [{"raw_text": "Стаж работы в НИУ ВШЭ: 6 лет"}]
+
+    assert _checksum(first) == _checksum(second)
+
+
+def _parsed_employee(profile_id: str) -> dict:
+    return {
+        "source_url": f"https://www.hse.ru/staff/{profile_id}",
+        "profile_type": "staff",
+        "profile_id": profile_id,
+        "full_name": "Same Person",
+        "tabs": [],
+        "sections": [],
+        "parser_version": "0.6.0",
+        "_html": "<html></html>",
+    }
--- a/tests/test_dataset_versions.py
+++ b/tests/test_dataset_versions.py
@@ -0,0 +1,88 @@
+from datetime import datetime, timezone
+
+from app.models import Employee
+from app.services.dataset_versions import get_or_create_current_version, sync_employees_payload
+
+
+def _employee(profile_key: str, checksum: str, *, status: str = "active") -> Employee:
+    return Employee(
+        profile_key=profile_key,
+        profile_type=profile_key.split(":", 1)[0],
+        profile_id=profile_key.split(":", 1)[1],
+        canonical_url=f"https://www.hse.ru/{profile_key}",
+        full_name=profile_key,
+        status=status,
+        first_seen_at=datetime.now(timezone.utc),
+        last_seen_at=datetime.now(timezone.utc),
+        current_data={"profile_key": profile_key},
+        current_checksum=checksum,
+    )
+
+
+def test_dataset_version_hash_is_stable_for_same_employee_state(db_session):
+    db_session.add(_employee("staff:alpha", "a" * 64))
+    db_session.commit()
+
+    first = get_or_create_current_version(db_session)
+    db_session.commit()
+    second = get_or_create_current_version(db_session)
+
+    assert second.id == first.id
+    assert second.hash == first.hash
+    assert second.employee_count == 1
+
+
+def test_dataset_version_hash_changes_when_employee_checksum_changes(db_session):
+    employee = _employee("staff:alpha", "a" * 64)
+    db_session.add(employee)
+    db_session.commit()
+    first = get_or_create_current_version(db_session)
+    db_session.commit()
+
+    employee.current_checksum = "b" * 64
+    db_session.commit()
+    second = get_or_create_current_version(db_session)
+
+    assert second.hash != first.hash
+    assert second.previous_hash == first.hash
+
+
+def test_sync_employees_diff_spans_multiple_intermediate_versions(db_session):
+    alpha = _employee("staff:alpha", "a" * 64)
+    db_session.add(alpha)
+    db_session.commit()
+    first = get_or_create_current_version(db_session)
+    db_session.commit()
+
+    beta = _employee("staff:beta", "b" * 64)
+    db_session.add(beta)
+    db_session.commit()
+    get_or_create_current_version(db_session)
+    db_session.commit()
+
+    alpha.current_checksum = "c" * 64
+    alpha.current_data = {"profile_key": "staff:alpha", "changed": True}
+    db_session.commit()
+
+    payload = sync_employees_payload(db_session, client_hash=first.hash, include_data=False)
+
+    assert payload["mode"] == "delta"
+    assert [item["profile_key"] for item in payload["changes"]["added"]] == ["staff:beta"]
+    assert [item["profile_key"] for item in payload["changes"]["updated"]] == ["staff:alpha"]
+    assert payload["changes"]["dismissed"] == []
+    assert payload["changes"]["removed"] == []
+
+
+def test_sync_employees_reports_dismissed_as_tombstone(db_session):
+    alpha = _employee("staff:alpha", "a" * 64)
+    db_session.add(alpha)
+    db_session.commit()
+    first = get_or_create_current_version(db_session)
+    db_session.commit()
+
+    alpha.status = "dismissed"
+    db_session.commit()
+    payload = sync_employees_payload(db_session, client_hash=first.hash, include_data=False)
+
+    assert payload["changes"]["dismissed"][0]["profile_key"] == "staff:alpha"
+    assert payload["changes"]["dismissed"][0]["status"] == "dismissed"
--- a/tests/test_employee_detail_template.py
+++ b/tests/test_employee_detail_template.py
@@ -27,4 +27,6 @@ def test_employee_detail_template_is_human_readable():
    assert "Дата увольнения" in template
    assert "Тип профиля" in template
    assert "ID профиля" in template
+    assert "Обновить данные" in template
+    assert 'action="/admin/employees/{{ employee.id }}/refresh"' in template
    assert "Снапшоты" in template
--- a/tests/test_parser.py
+++ b/tests/test_parser.py
@@ -1,6 +1,6 @@
 from bs4 import BeautifulSoup

-from app.parser.profile import enrich_sections_from_hse_widgets, extract_person_tabs
+from app.parser.profile import enrich_sections_from_hse_widgets, extract_person_tabs, extract_sections
 from app.parser.profile_url import normalize_profile_url, parse_profile_identity


@@ -64,6 +64,47 @@ class FakeSession:
        )


+class GroupedPublicationsSession(FakeSession):
+    def post(self, url, **kwargs):
+        self.posts.append((url, kwargs))
+        return FakeResponse(
+            {
+                "status": "ok",
+                "result": {
+                    "more": False,
+                    "total": 1,
+                    "groupType": 2,
+                    "items": {
+                        "year": {
+                            "header": {"ru": "по году", "en": "by year"},
+                            "criteria": {"year": []},
+                            "items": {
+                                "2011": [
+                                    {
+                                        "id": "146366790",
+                                        "type": "ARTICLE",
+                                        "title": "Развитие теории самосогласованного поля",
+                                        "year": 2011,
+                                        "description": {"short": {"ru": "Журнал физической химии 2011."}},
+                                    }
+                                ],
+                                "2012": [
+                                    {
+                                        "id": "146367323",
+                                        "type": "ARTICLE",
+                                        "title": "Self-consistent field theory investigation",
+                                        "year": 2012,
+                                        "description": {"short": {"en": "Russian Journal of Physical Chemistry A 2012."}},
+                                    }
+                                ],
+                            },
+                        }
+                    },
+                },
+            }
+        )
+
+
 def test_normalize_profile_url_supports_staff_and_org_persons():
    assert normalize_profile_url("/staff/avsergeev#sci") == "https://www.hse.ru/staff/avsergeev"
    assert normalize_profile_url("https://www.hse.ru/org/persons/123/") == "https://www.hse.ru/org/persons/123"
@@ -117,3 +158,60 @@ def test_enrich_sections_from_hse_widgets_loads_publications_and_vkr():
    assert theses["theses"][0]["project_url"] == "https://www.hse.ru/edu/vkr/1045750164"
    assert session.posts[0][0] == "https://publications.hse.ru/api/searchPubs"
    assert session.gets[0][1]["params"] == {"supervisorId": "803294906"}
+
+
+def test_enrich_sections_from_hse_widgets_loads_grouped_publications():
+    soup = BeautifulSoup(
+        """
+        <script src="/n/stat/publications/dist-w/publs.js" data-author="133709486" data-widget-name="AuthorSearch"></script>
+        """,
+        "html.parser",
+    )
+    session = GroupedPublicationsSession()
+
+    sections = enrich_sections_from_hse_widgets(
+        session,
+        soup,
+        "https://www.hse.ru/org/persons/133709486",
+        {"User-Agent": "test"},
+        10,
+        [],
+    )
+
+    publications = next(section for section in sections if section["type"] == "publications")
+
+    assert publications["publications_count"] == 2
+    assert [item["id"] for item in publications["publications"]] == ["146366790", "146367323"]
+    assert publications["publications"][0]["url"] == "https://publications.hse.ru/view/146366790"
+    assert publications["publications"][1]["url"] == "https://publications.hse.ru/view/146367323"
+
+
+def test_news_heading_with_publications_word_does_not_absorb_widget_publications():
+    soup = BeautifulSoup(
+        """
+        <h2>Статья профессора МИЭМ вошла в число самых популярных публикаций на портале SpringerLink</h2>
+        <div class="post__text">
+          <p>Первоначально статья профессора вышла в российском журнале.</p>
+        </div>
+        <script src="/n/stat/publications/dist-w/publs.js" data-author="133709486" data-widget-name="AuthorSearch"></script>
+        """,
+        "html.parser",
+    )
+    session = FakeSession()
+
+    sections = extract_sections(soup, "https://www.hse.ru/org/persons/133709486")
+    sections = enrich_sections_from_hse_widgets(
+        session,
+        soup,
+        "https://www.hse.ru/org/persons/133709486",
+        {"User-Agent": "test"},
+        10,
+        sections,
+    )
+
+    assert sections[0]["type"] == "paragraphs"
+    assert sections[0]["title"].startswith("Статья профессора")
+    publications = [section for section in sections if section["type"] == "publications"]
+    assert len(publications) == 1
+    assert publications[0]["title"] == "Публикации и исследования"
+    assert publications[0]["publications_count"] == 1
Author	SHA1	Message	Date
Anton	6724b3f369	feat: adds crawl resource cache	2026-05-14 12:21:44 +03:00
admin	5180b89b81	Merge pull request 'feat: add dataset checkpoint sync for MCP' (#23 ) from feature/dataset-version-sync into main Reviewed-on: #23	2026-05-14 08:01:26 +00:00
Anton	29451ccee1	feat: add dataset checkpoint sync for MCP	2026-05-14 11:00:46 +03:00
admin	a3ff9c6e9c	Merge pull request 'fix: separate news from publications and add employee refresh' (#22 ) from fix/publications-news-refresh into main Reviewed-on: #22	2026-05-13 13:12:06 +00:00
Anton	8e19dc9f35	fix: separate news from publications and add employee refresh	2026-05-13 16:11:13 +03:00
admin	5b9d71426d	Merge pull request 'fix: support grouped HSE publication API responses' (#21 ) from fix/grouped-publications-parser into main Reviewed-on: #21	2026-05-13 09:46:48 +00:00
Anton	efa7192e45	fix: support grouped HSE publication API responses	2026-05-13 12:46:07 +03:00
admin	b27d613143	Merge pull request 'fix: remove mcp-auth from yml-file' (#20 ) from fix/remove-mcp-auth-compose into main Reviewed-on: #20	2026-05-08 09:33:17 +00:00
Anton	a1ab1c0319	fix: remove mcp-auth from yml-file	2026-05-08 12:32:40 +03:00
admin	0b4e04544d	Merge pull request 'fix: remove MCP application-level authorization' (#19 ) from fix/remove-mcp-auth into main Reviewed-on: #19	2026-05-08 09:15:18 +00:00
Anton	7593a460c7	fix: remove MCP application-level authorization	2026-05-08 12:14:19 +03:00
admin	a4e7388bcf	Merge pull request 'fix: use direct onclick handlers for run rows' (#18 ) from fix/direct-run-row-click-handler into main Reviewed-on: #18	2026-05-07 15:25:26 +00:00
Anton	ac319b3ee5	fix: use direct onclick handlers for run rows	2026-05-07 18:23:14 +03:00