feat: initial microservices project structure

Services:
- api: FastAPI gateway with JWT auth, async endpoints, WebSocket
- worker-gpu: CUDA sentence-transformers, FAISS IVFFlat, Ollama LLM
- worker-indexer: Winnowing+MinHash plagiarism detection, PDF/DOCX extraction
- worker-notifier: SMTP email notifications
- worker-gost: GOST 7.1-2003 and GOST R 7.0.5-2008 formatting

Infrastructure:
- docker-compose.yml (production) + docker-compose.dev.yml (hot reload)
- Nginx reverse proxy + WebSocket support
- PostgreSQL 16 with Alembic migrations
- Elasticsearch 8 with Russian/English analyzers
- MinIO, RabbitMQ, Redis, Ollama

Frontend:
- React 18 + Vite + TypeScript + TailwindCSS + Zustand + React Query v5
- 9 pages: Home, Search, Cabinet, Task, Check, Bibliography, Pricing, Login, Register

Scripts:
- Parser stubs: OpenAlex, КиберЛенинка, arXiv (Phase 0 - to be filled)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

services/worker-gpu/app/es_client.py

@@ -0,0 +1,151 @@
+"""Elasticsearch клиент для полнотекстового поиска (BM25)."""
+
+import logging
+from typing import Any
+
+from elasticsearch import Elasticsearch, exceptions as es_exceptions
+
+from app.config import settings
+
+logger = logging.getLogger(__name__)
+
+_es_client: Elasticsearch | None = None
+
+INDEX_NAME = "documents"
+
+
+def get_es_client() -> Elasticsearch:
+    """Получить или создать синглтон ES клиент."""
+    global _es_client
+    if _es_client is None:
+        _es_client = Elasticsearch(
+            settings.ELASTICSEARCH_URL,
+            retry_on_timeout=True,
+            max_retries=3,
+            request_timeout=30,
+        )
+    return _es_client
+
+
+def search_fulltext(
+    query: str,
+    lang: str | None = None,
+    year_from: int | None = None,
+    year_to: int | None = None,
+    category: str | None = None,
+    size: int = 50,
+) -> list[dict[str, Any]]:
+    """
+    Полнотекстовый BM25 поиск в Elasticsearch.
+
+    Args:
+        query: Поисковый запрос
+        lang: Фильтр языка (ru, en, None = все)
+        year_from: Фильтр года публикации (от)
+        year_to: Фильтр года публикации (до)
+        category: Фильтр тематической категории
+        size: Максимальное количество результатов
+
+    Returns:
+        Список словарей с полями doc_id и score
+    """
+    es = get_es_client()
+
+    # Составить bool запрос
+    must_clauses: list[dict] = [
+        {
+            "multi_match": {
+                "query": query,
+                "fields": ["title^3", "abstract^2", "authors"],
+                "type": "best_fields",
+                "operator": "or",
+                "minimum_should_match": "30%",
+            }
+        }
+    ]
+
+    filter_clauses: list[dict] = []
+
+    if lang:
+        filter_clauses.append({"term": {"lang": lang}})
+
+    if year_from or year_to:
+        range_filter: dict = {"range": {"year": {}}}
+        if year_from:
+            range_filter["range"]["year"]["gte"] = year_from
+        if year_to:
+            range_filter["range"]["year"]["lte"] = year_to
+        filter_clauses.append(range_filter)
+
+    body: dict[str, Any] = {
+        "query": {
+            "bool": {
+                "must": must_clauses,
+                "filter": filter_clauses,
+            }
+        },
+        "size": size,
+        "_source": ["doc_id"],
+    }
+
+    try:
+        response = es.search(index=INDEX_NAME, body=body)
+        hits = response["hits"]["hits"]
+        return [
+            {
+                "doc_id": hit["_source"]["doc_id"],
+                "es_score": hit["_score"],
+                "es_id": hit["_id"],
+            }
+            for hit in hits
+        ]
+    except es_exceptions.ConnectionError as e:
+        logger.error(f"Elasticsearch недоступен: {e}")
+        return []
+    except es_exceptions.NotFoundError:
+        logger.warning(f"Индекс {INDEX_NAME!r} не найден в Elasticsearch")
+        return []
+    except Exception as e:
+        logger.error(f"Ошибка поиска в Elasticsearch: {e}")
+        return []
+
+
+def index_document(doc_id: int, doc_data: dict[str, Any]) -> bool:
+    """
+    Индексировать документ в Elasticsearch.
+
+    Args:
+        doc_id: ID документа в PostgreSQL
+        doc_data: Словарь с метаданными документа
+
+    Returns:
+        True при успехе, False при ошибке
+    """
+    es = get_es_client()
+
+    doc = {
+        "doc_id": doc_id,
+        "source": doc_data.get("source"),
+        "title": doc_data.get("title", ""),
+        "abstract": doc_data.get("abstract", ""),
+        "authors": " ".join(
+            f"{a.get('last_name', '')} {a.get('first_name', '')}"
+            for a in doc_data.get("authors", [])
+        ),
+        "year": doc_data.get("year"),
+        "lang": doc_data.get("lang"),
+        "journal": doc_data.get("journal"),
+        "doi": doc_data.get("doi"),
+        "url": doc_data.get("url"),
+    }
+
+    try:
+        es.index(
+            index=INDEX_NAME,
+            id=str(doc_id),
+            document=doc,
+        )
+        return True
+    except Exception as e:
+        logger.error(f"Ошибка индексации документа {doc_id} в ES: {e}")
+        return False