Add reusable LLM parser profiles with multi-event extract.

Support kind=llm profiles (instruction/schema), optional multi-event posts via #eN URLs, and recover stale running/queued parse jobs after worker crashes.

Co-authored-by: Cursor <cursoragent@cursor.com>

.env.example

@@ -3,20 +3,33 @@ TELEGRAM_API_ID=12345678
 TELEGRAM_API_HASH=your_api_hash_here
 TELEGRAM_SESSION_PATH=/data/telegram.session
 
-# Optional Telegram proxy
+# Optional Telegram proxy fallback (when VPN tab is off)
+# Prefer Admin UI → VPN (subscription / socks5 / http).
 # TELEGRAM_PROXY_TYPE=socks5
 # TELEGRAM_PROXY_HOST=127.0.0.1
 # TELEGRAM_PROXY_PORT=1080
+# TELEGRAM_PROXY_USER=
+# TELEGRAM_PROXY_PASS=
 
 # Platform internals
 INTERNAL_TOKEN=dev-internal-token
 TEST_PI_API_KEY=test-pi-api-key-change-me
 
-# DeepSeek LLM for extract_mode=llm (Telegram unstructured + Crawl4AI LLM)
+# Admin UI login (public map stays open; /admin and object writes need JWT)
+ADMIN_USER=admin
+ADMIN_PASSWORD=change-me
+ADMIN_JWT_SECRET=change-me-jwt-secret
+
+# DeepSeek LLM:
+# - extract_mode=llm on CP workers (Telegram unstructured + Crawl4AI)
+# - Generate профиля парсера на ca-api (один раз; runtime — только правила)
 # DEEPSEEK_API_KEY=sk-...
 # DEEPSEEK_BASE_URL=https://api.deepseek.com
 # DEEPSEEK_MODEL=deepseek-chat
 
+# Recover parse jobs stuck in running/queued after worker crash (seconds, default 900)
+# STALE_JOB_SECONDS=900
+
 # CP adapter workers (set in docker-compose; override locally if needed)
 # ENABLED_ADAPTERS=telegram
 # WORKER_FAMILIES=telegram

AGENTS.md

@@ -36,10 +36,21 @@ docker compose up --build cp-workers-web   # single worker rebuild
 
 Health: `curl http://localhost:8080/api/health`
 
+Admin auth (map is public; `/admin/*` needs JWT from `ADMIN_USER` / `ADMIN_PASSWORD`):
+
+```bash
+TOKEN=$(curl -s -X POST http://localhost:8080/admin/auth/login \
+  -H 'Content-Type: application/json' \
+  -d '{"username":"admin","password":"change-me"}' | jq -r .access_token)
+```
+
+VPN (subscription / SOCKS): Admin UI → **VPN**, or `PUT /admin/vpn`. Workers and `cp-vpn` read `GET /internal/vpn`.
+
 Create Telegram parser:
 
 ```bash
 curl -X POST http://localhost:8080/admin/jobs \
+  -H "Authorization: Bearer $TOKEN" \
   -H 'Content-Type: application/json' \
   -d '{"source_type":"telegram","source_config":{"channel":"example","limit":50}}'
 ```
@@ -63,7 +74,7 @@ Adapters return dicts matching `IngestEventItem` (`contracts/ingest.py`). Requir
 
 ### LLM extract (runtime, not dev)
 
-`extract_mode: llm` uses DeepSeek via `workers/llm_extract.py`. Key: `DEEPSEEK_API_KEY` in `.env`. Do not confuse with Cursor dev agents.
+`extract_mode: llm` uses DeepSeek via `workers/llm_extract.py` (batch). Reusable profiles in admin UI (`kind=heuristic|llm`) flatten into `extract_mode=profile` or `llm`. Key: `DEEPSEEK_API_KEY` in `.env`. Do not confuse with Cursor dev agents.
 
 ## Common tasks
 
@@ -86,7 +97,7 @@ Adapters return dicts matching `IngestEventItem` (`contracts/ingest.py`). Requir
 
 - Rule: `.cursor/rules/platform.mdc` (always on)
 - Skill: `.cursor/skills/add-parser-adapter/` for end-to-end new sources
-- Deep docs: `README.md`, `centers/parsing/ARCHITECTURE.md`
+- Deep docs: `docs/` (overview + data-flow), `centers/analytics/ARCHITECTURE.md`, `centers/parsing/ARCHITECTURE.md`
 
 ## Commit style
 

README.md

@@ -2,6 +2,8 @@
 
 Единая платформа: ЦА (аналитика и карта) + ЦП (адаптеры парсинга: Telegram, Crawl4AI, VIINA).
 
+**Документация для разработчиков:** [`docs/`](docs/README.md) — [обзор архитектуры](docs/architecture-overview.md), [поток данных](docs/data-flow.md), [локальный запуск](docs/local-dev.md).
+
 ## Архитектура
 
 ```mermaid
@@ -15,13 +17,14 @@ flowchart LR
   CA -->|GET /api/v1/events| PI
 ```
 
-| Центр | Контейнеры | Назначение |
-|-------|------------|------------|
-| **ЦА** | `ca-db`, `ca-api`, `ca-frontend` | PostgreSQL, ingest API, карта, distribution API |
-| **ЦП** | `cp-workers`, `cp-workers-web`, `cp-workers-nlp` | Адаптеры: Telegram / Crawl4AI / VIINA |
-| **Общее** | `redis` | Очереди `cp:jobs:{telegram\|web\|nlp}` |
+| Центр | Контейнеры | Назначение | Документ |
+|-------|------------|------------|----------|
+| **ЦА** | `ca-db`, `ca-api`, `ca-frontend` | PostgreSQL, ingest API, карта, distribution API | [`centers/analytics/ARCHITECTURE.md`](centers/analytics/ARCHITECTURE.md) |
+| **ЦП** | `cp-workers`, `cp-workers-web`, `cp-workers-nlp` | Адаптеры: Telegram / Crawl4AI / VIINA | [`centers/parsing/ARCHITECTURE.md`](centers/parsing/ARCHITECTURE.md) |
+| **ПИ** | (маршрут в ЦА) | `GET /api/v1/events` | [`centers/analytics/DISTRIBUTION.md`](centers/analytics/DISTRIBUTION.md) |
+| **Общее** | `redis` | Очереди `cp:jobs:{telegram\|web\|nlp}` | [`docs/contracts.md`](docs/contracts.md) |
 
-Подробности ЦП: [`centers/parsing/ARCHITECTURE.md`](centers/parsing/ARCHITECTURE.md).
+Подробный обзор платформы: [`docs/architecture-overview.md`](docs/architecture-overview.md).
 
 ## Структура monorepo
 
@@ -29,13 +32,16 @@ flowchart LR
 MapMil/
 ├── centers/
 │   ├── analytics/
-│   │   ├── api/          # CA backend (FastAPI + PostgreSQL)
-│   │   └── frontend/     # CA admin UI (Vue + Leaflet)
+│   │   ├── api/              # CA backend (FastAPI + PostgreSQL)
+│   │   ├── frontend/         # CA admin UI (Vue + Leaflet)
+│   │   ├── ARCHITECTURE.md
+│   │   └── DISTRIBUTION.md   # ПИ
 │   └── parsing/
 │       ├── ARCHITECTURE.md
-│       └── workers/      # CP workers + adapters
-├── contracts/            # Shared schemas (ingest, jobs, sources, queues)
-├── data/                 # telegram.session (локально, не в git)
+│       └── workers/          # CP workers + adapters
+├── contracts/                # Shared schemas (ingest, jobs, sources, queues)
+├── docs/                     # Документация для разработчиков
+├── data/                     # telegram.session (локально, не в git)
 ├── docker-compose.yml
 └── .env
 ```
@@ -61,20 +67,23 @@ cp ../SocialParser/data/telegram.session data/
 docker compose up --build
 ```
 
-4. Откройте UI: [http://localhost:8080](http://localhost:8080)
+4. Откройте UI: [http://localhost:8080](http://localhost:8080) — карта публичная; разделы админки — после входа (`ADMIN_USER` / `ADMIN_PASSWORD` из `.env`).
 
 ## Admin UI
 
-Веб-интерфейс ЦА доступен по тем же адресу. Навигация в шапке:
+Веб-интерфейс ЦА доступен по тому же адресу. Карта (`/`) открыта всем. Остальные пункты навигации видны после логина (`/login`).
 
 | Раздел | Путь | Описание |
 |--------|------|----------|
-| **Карта** | `/` | Интерактивная карта событий: навигация по датам (flatpickr), пресеты периода, фильтры региона/темы/источника, подложки Яндекс/OSM/Topo/ESRI, линейка, полноэкранный режим, центрирование по координатам и городам, поиск населённых пунктов (Nominatim). CRUD для ручных объектов (ПКМ). Поддерживает `?eventId=` |
-| **Парсеры** | `/parsers` | Адаптеры `telegram` / `crawl4ai` / `viina`, интервал, CRUD; дедуп по `source_url` |
+| **Карта** | `/` | Интерактивная карта событий (публичный просмотр). CRUD ручных объектов — только для админа (ПКМ). Поддерживает `?eventId=` |
+| **Парсеры** | `/parsers` | Связка канал + профиль (`heuristic`→`extract_mode=profile`, `llm`→`extract_mode=llm`); дедуп по `source_url` |
+| **Профили** | `/parser-profiles` | Reusable heuristic (правила) или llm (instruction/schema); целевые поля = Event |
 | **События** | `/events` | Фильтрация, пагинация, просмотр деталей, ссылка «На карте» для событий с координатами |
 | **Аналитика** | `/analytics` | KPI-карточки, график динамики ingest за 30 дней, топ населённых пунктов и регионов |
 | **ПИ** | `/consumers` | CRUD подписчиков distribution API, ротация ключей, тест среза через `/api/v1/events` |
 
+Авторизация админки: `POST /admin/auth/login` → JWT Bearer. Все `/admin/*` (кроме login) и мутации `/api/objects` требуют токен. `GET /api/map/*` публичен.
+
 ## Сохранение Telegram-сессии
 
 **Важно:** существующий файл сессии **не удаляется и не пересоздаётся**.
@@ -126,10 +135,15 @@ docker compose up --build
 | PATCH | `/admin/consumers/{id}` | Обновить подписчика |
 | POST | `/admin/consumers/{id}/rotate-key` | Сменить API-ключ |
 
-Пример задания Telegram:
+Пример задания Telegram (нужен JWT после `POST /admin/auth/login`):
 
 ```bash
+TOKEN=$(curl -s -X POST http://localhost:8080/admin/auth/login \
+  -H 'Content-Type: application/json' \
+  -d '{"username":"admin","password":"change-me"}' | jq -r .access_token)
+
 curl -X POST http://localhost:8080/admin/jobs \
+  -H "Authorization: Bearer $TOKEN" \
   -H 'Content-Type: application/json' \
   -d '{"source_type":"telegram","source_config":{"channel":"creamy_caprice","limit":50}}'
 ```
@@ -201,3 +215,8 @@ docker compose down
 ```
 
 Данные PostgreSQL сохраняются в volume `pgdata`.
+
+## Вход в админку (/login):
+
+Логин: admin
+Пароль: change-me

centers/analytics/ARCHITECTURE.md

@@ -0,0 +1,182 @@
+# Архитектура ЦА (Analytics Center)
+
+Центр аналитики — ядро MapMil: PostgreSQL, FastAPI, Vue admin/карта, distribution API для ПИ.
+
+Общий контекст: [docs/architecture-overview.md](../../docs/architecture-overview.md).  
+ПИ отдельно: [DISTRIBUTION.md](DISTRIBUTION.md).
+
+---
+
+## Общими чертами
+
+ЦА:
+
+- хранит события, jobs, объекты карты, consumers;
+- отдаёт единственный UI (`ca-frontend`);
+- ставит задания в Redis для ЦП;
+- принимает ingest от ЦП;
+- отдаёт срезы внешним системам через `/api/v1/events`.
+
+```mermaid
+flowchart TB
+  UI[ca-frontend Vue]
+  API[ca-api FastAPI]
+  DB[(PostgreSQL)]
+  Redis[(Redis)]
+  CP[CP workers]
+
+  UI --> API
+  API --> DB
+  API -->|enqueue| Redis
+  Redis --> CP
+  CP -->|/internal/*| API
+```
+
+Контейнеры: `ca-db`, `ca-api`, `ca-frontend` (+ общий `redis`).
+
+---
+
+## Подробнее
+
+### Структура
+
+```text
+centers/analytics/
+├── ARCHITECTURE.md
+├── DISTRIBUTION.md
+├── api/
+│   ├── Dockerfile
+│   ├── requirements.txt
+│   └── app/
+│       ├── main.py              # lifespan: migrations, scheduler, seed
+│       ├── database.py
+│       ├── models.py
+│       ├── schemas.py
+│       ├── deps.py              # X-Internal-Token
+│       ├── seed.py
+│       ├── storage.py           # uploads
+│       ├── routers/
+│       │   ├── objects.py       # /api/health, /api/objects, media
+│       │   ├── map.py           # /api/map/*
+│       │   ├── admin.py         # /admin/* (jobs, events, consumers)
+│       │   ├── parser_profiles.py # /admin/parser-profiles/*
+│       │   ├── parse_channels.py  # /admin/parse-channels/*
+│       │   ├── internal.py      # /internal/* (только ЦП)
+│       │   └── v1.py            # /api/v1/* (ПИ)
+│       └── services/
+│           ├── jobs.py          # Redis RPUSH + flatten Profile/Channel
+│           ├── scheduler.py     # периодический re-queue
+│           ├── ingest.py        # дедуп + map sync
+│           ├── parser_builder.py # DeepSeek → HeuristicProfile (один раз)
+│           ├── filtering.py
+│           ├── map_query.py
+│           ├── events_query.py
+│           ├── analytics.py
+│           └── migrations.py
+└── frontend/
+    ├── Dockerfile               # Vite build + nginx
+    ├── nginx.conf               # proxy /api /admin /internal → ca-api
+    └── src/
+        ├── views/               # Map, Channels, Profiles, Parsers, …
+        ├── components/          # карта, CRUD объектов
+        ├── api/                 # HTTP-клиенты
+        └── router/index.ts
+```
+
+### Модели данных
+
+| Модель | Таблица | Назначение |
+|--------|--------|------------|
+| `Event` | `events` | Нормализованное событие; UK `source_url` |
+| `ParserProfile` | `parser_profiles` | Статичный heuristic-профиль (JSON-правила) |
+| `ParseChannel` | `parse_channels` | Канал (Telegram handle) |
+| `ParseJob` | `parse_jobs` | Связка канал+профиль, интервал, статус; FK `profile_id` + `channel_id` |
+| `MapObject` | `map_objects` | Точка на карте (event или ручная) |
+| `ObjectMedia` | `object_media` | Файлы к объектам |
+| `Consumer` | `consumers` | Подписчик ПИ (hash ключа) |
+| `ConsumerFilter` | `consumer_filters` | Фильтры среза ПИ |
+
+### HTTP-поверхности
+
+| Prefix | Кто вызывает | Содержание |
+|--------|--------------|------------|
+| `/api/*` | UI, публичный health | Карта, объекты, медиа |
+| `/admin/*` | UI admin | Jobs, channels, profiles, events, analytics, consumers |
+| `/internal/*` | Только ЦП | ingest, job status, listener subscriptions |
+| `/api/v1/*` | Внешние клиенты | Events с Bearer-ключом |
+
+Internal защищён заголовком `X-Internal-Token` (`INTERNAL_TOKEN`).
+
+### Профили, каналы и пары
+
+1. **Канал** (`CRUD /admin/parse-channels`) — handle Telegram + метаданные.
+2. **Профиль** (`CRUD /admin/parser-profiles`) — Generate (DeepSeek один раз по образцу) → Preview → Save `heuristic_profile`.
+3. **Пара** (`POST /admin/jobs` с `channel_id` + `profile_id`) → `ParseJob`; при enqueue ЦА **разворачивает** пару в плоский `source_config`:
+
+```json
+{
+  "channel": "<ParseChannel.channel>",
+  "limit": 100,
+  "extract_mode": "profile",
+  "heuristic_profile": { "...из ParserProfile..." }
+}
+```
+
+ЦП по-прежнему получает только Redis payload — без доступа к таблицам Profile/Channel.
+
+Roadmap: кастомные пользовательские таблицы подменяют список `target-fields` при генерации; профиль будет ссылаться на `schema_id`.
+
+### Jobs и scheduler
+
+1. `POST /admin/jobs` / retry → запись `ParseJob` + `enqueue_parse_job` (flatten + RPUSH).
+2. Очередь: `cp:jobs:{family}` из `contracts/queues.py`.
+3. `services/scheduler.py` — тик ~30 с, повторная постановка активных jobs по `interval_seconds`.
+
+### Ingest
+
+`POST /internal/ingest` → `services/ingest.py`:
+
+- дедуп по `source_url`;
+- создание `Event`;
+- при координатах — sync `MapObject`;
+- batch обновляет статус job; `listener: true` — нет.
+
+### Frontend (маршруты)
+
+| Path | View |
+|------|------|
+| `/` | `MapViewPage.vue` |
+| `/channels` | `ChannelsView.vue` |
+| `/parser-profiles` | `ParserProfilesView.vue` |
+| `/parsers` | `ParsersView.vue` (связки канал+профиль) |
+| `/events` | `EventsView.vue` |
+| `/analytics` | `AnalyticsView.vue` |
+| `/consumers` | `ConsumersView.vue` |
+
+Карта: Leaflet, фильтры дат/региона/темы/источника, CRUD объектов (ПКМ), медиа, таймлайн появления (если включён в UI).
+
+### Nginx
+
+`ca-frontend` слушает `:80` (с хоста `:8080`), проксирует backend-пути на `http://ca-api:8000`. Лимит тела для медиа задаётся в `nginx.conf`.
+
+### Env (ЦА)
+
+| Переменная | Назначение |
+|------------|------------|
+| `DATABASE_URL` | PostgreSQL |
+| `REDIS_URL` | Очереди |
+| `INTERNAL_TOKEN` | Auth ЦП ↔ ЦА |
+| `TEST_PI_API_KEY` | Seed consumer `test-pi` |
+| `UPLOAD_DIR` | Медиа (по умолчанию `/data/uploads`) |
+| `DEEPSEEK_API_KEY` | Generate профиля на ca-api; не нужен для preview/runtime |
+| `DEEPSEEK_BASE_URL` / `DEEPSEEK_MODEL` | Опционально |
+
+### Типовые точки входа в код
+
+| Задача | Файл |
+|--------|------|
+| Новый admin endpoint | `routers/admin.py` |
+| Логика ingest | `services/ingest.py` |
+| Фильтры карты | `services/map_query.py` + `routers/map.py` |
+| Новый экран UI | `frontend/src/views/` + `router/index.ts` |
+| Поля парсера в форме | `ParsersView.vue` (+ contracts) |

centers/analytics/DISTRIBUTION.md

@@ -0,0 +1,91 @@
+# Distribution API (ПИ)
+
+**ПИ** — не отдельный Docker-сервис, а внешний HTTP-срез поверх данных ЦА.
+
+Общий контекст: [docs/architecture-overview.md](../../docs/architecture-overview.md).
+
+---
+
+## Общими чертами
+
+1. В UI «ПИ» (`/consumers`) создаётся подписчик.
+2. Ему выдаётся API-ключ (в БД хранится только SHA-256 hash).
+3. Клиент читает события:
+
+```http
+GET /api/v1/events
+Authorization: Bearer <api_key>
+```
+
+4. ЦА отдаёт события с учётом фильтров consumer (регионы, темы, дата).
+
+```mermaid
+flowchart LR
+  Client[External system]
+  V1[GET /api/v1/events]
+  Cons[Consumer + Filter]
+  Events[(events)]
+
+  Client -->|Bearer key| V1
+  V1 --> Cons
+  Cons --> Events
+```
+
+---
+
+## Подробнее
+
+### Код
+
+| Часть | Путь |
+|-------|------|
+| Роут | `centers/analytics/api/app/routers/v1.py` |
+| Модели | `Consumer`, `ConsumerFilter` в `models.py` |
+| Фильтрация | `services/filtering.py` |
+| Admin CRUD | `routers/admin.py` — `/admin/consumers`, `…/rotate-key` |
+| UI | `frontend/src/views/ConsumersView.vue` |
+| Seed | `seed.py` — consumer `test-pi` из `TEST_PI_API_KEY` |
+
+### Аутентификация
+
+- Заголовок: `Authorization: Bearer <plain_api_key>`.
+- Сравнение с `consumers.api_key_hash` (SHA-256).
+- Неактивный consumer → отказ.
+
+### Фильтры подписчика
+
+Типичные ограничения (через `ConsumerFilter`):
+
+- список регионов;
+- список тем;
+- `date_from` (нижняя граница даты события).
+
+Точный набор полей — в модели/схемах admin API; при изменении фильтров обновляйте UI consumers и `filtering.py`.
+
+### Admin операции
+
+| Метод | Путь | Действие |
+|-------|------|----------|
+| GET | `/admin/consumers` | Список |
+| POST | `/admin/consumers` | Создать (+ ключ в ответе один раз) |
+| PATCH | `/admin/consumers/{id}` | Обновить |
+| POST | `/admin/consumers/{id}/rotate-key` | Новый ключ |
+
+### Локальный тест
+
+После `docker compose up` (если не меняли seed):
+
+```bash
+curl -s -H "Authorization: Bearer test-pi-api-key-change-me" \
+  "http://localhost:8080/api/v1/events" | head
+```
+
+В проде обязательно смените `TEST_PI_API_KEY` и ротируйте ключи.
+
+### Что ПИ не делает
+
+- не пишет в БД;
+- не ставит parse jobs;
+- не ходит в Redis / ЦП.
+
+Только чтение уже ingest'нутых событий через контракт `/api/v1`.

centers/analytics/api/app/auth.py

@@ -0,0 +1,74 @@
+"""Admin auth: single env-based user + JWT bearer tokens."""
+
+from __future__ import annotations
+
+import hmac
+import os
+import time
+from typing import Any
+
+import jwt
+from fastapi import HTTPException
+
+ALGORITHM = "HS256"
+DEFAULT_TTL_SECONDS = 60 * 60 * 24  # 24h
+
+
+def _admin_user() -> str:
+    return os.getenv("ADMIN_USER", "admin").strip() or "admin"
+
+
+def _admin_password() -> str:
+    return os.getenv("ADMIN_PASSWORD", "").strip()
+
+
+def _jwt_secret() -> str:
+    secret = os.getenv("ADMIN_JWT_SECRET", "").strip()
+    if not secret:
+        # Dev fallback: derive from password so local stacks boot without extra secret.
+        password = _admin_password()
+        if not password:
+            raise HTTPException(
+                status_code=503,
+                detail="ADMIN_PASSWORD is not configured",
+            )
+        return f"mapmil-dev:{password}"
+    return secret
+
+
+def admin_credentials_configured() -> bool:
+    return bool(_admin_password())
+
+
+def verify_credentials(username: str, password: str) -> bool:
+    expected_user = _admin_user()
+    expected_password = _admin_password()
+    if not expected_password:
+        return False
+    user_ok = hmac.compare_digest(username.strip(), expected_user)
+    pass_ok = hmac.compare_digest(password, expected_password)
+    return user_ok and pass_ok
+
+
+def create_access_token(*, username: str, ttl_seconds: int = DEFAULT_TTL_SECONDS) -> str:
+    now = int(time.time())
+    payload: dict[str, Any] = {
+        "sub": username,
+        "role": "admin",
+        "iat": now,
+        "exp": now + ttl_seconds,
+    }
+    return jwt.encode(payload, _jwt_secret(), algorithm=ALGORITHM)
+
+
+def decode_access_token(token: str) -> dict[str, Any]:
+    try:
+        payload = jwt.decode(token, _jwt_secret(), algorithms=[ALGORITHM])
+    except jwt.ExpiredSignatureError as exc:
+        raise HTTPException(status_code=401, detail="Token expired") from exc
+    except jwt.InvalidTokenError as exc:
+        raise HTTPException(status_code=401, detail="Invalid token") from exc
+
+    if payload.get("role") != "admin" or not payload.get("sub"):
+        raise HTTPException(status_code=401, detail="Invalid token")
+    return payload

centers/analytics/api/app/deps.py

@@ -2,6 +2,8 @@ import os
 
 from fastapi import Header, HTTPException
 
+from .auth import decode_access_token
+
 
 def verify_internal_token(
     x_internal_token: str | None = Header(default=None, alias="X-Internal-Token"),
@@ -9,3 +11,16 @@ def verify_internal_token(
     expected = os.getenv("INTERNAL_TOKEN", "dev-internal-token")
     if not x_internal_token or x_internal_token != expected:
         raise HTTPException(status_code=401, detail="Invalid internal token")
+
+
+def verify_admin(
+    authorization: str | None = Header(default=None),
+) -> str:
+    """Require Authorization: Bearer <admin JWT>. Returns username (sub)."""
+    if not authorization or not authorization.startswith("Bearer "):
+        raise HTTPException(status_code=401, detail="Missing or invalid Authorization header")
+    token = authorization.removeprefix("Bearer ").strip()
+    if not token:
+        raise HTTPException(status_code=401, detail="Missing or invalid Authorization header")
+    payload = decode_access_token(token)
+    return str(payload["sub"])

centers/analytics/api/app/main.py

@@ -13,8 +13,8 @@ for _candidate in (_HERE.parent, *_HERE.parents):
         break
 
 from .database import Base, engine, get_db
-from .routers import admin, internal, map, objects, v1
-from .seed import seed_objects, seed_test_consumer
+from .routers import admin, auth, internal, map, objects, parse_channels, parser_profiles, v1, vpn
+from .seed import seed_objects, seed_test_consumer, seed_vpn_settings
 from .services.migrations import migrate_schema
 from .services.scheduler import start_scheduler
 from .storage import ensure_upload_dir
@@ -30,6 +30,7 @@ async def lifespan(_: FastAPI):
     try:
         seed_objects(db)
         seed_test_consumer(db)
+        seed_vpn_settings(db)
     finally:
         db.close()
     yield
@@ -49,5 +50,9 @@ app.add_middleware(
 app.include_router(objects.router)
 app.include_router(map.router)
 app.include_router(internal.router)
+app.include_router(auth.router)
 app.include_router(admin.router)
+app.include_router(parser_profiles.router)
+app.include_router(parse_channels.router)
+app.include_router(vpn.router)
 app.include_router(v1.router)

centers/analytics/api/app/models.py

@@ -93,12 +93,57 @@ class Event(Base):
     )
 
 
+class ParserProfile(Base):
+    __tablename__ = "parser_profiles"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, index=True)
+    name: Mapped[str] = mapped_column(String(255), nullable=False)
+    # heuristic = static rules; llm = instruction + extract_schema at CP runtime
+    kind: Mapped[str] = mapped_column(String(50), default="heuristic", nullable=False, index=True)
+    sample_post: Mapped[str] = mapped_column(Text, default="")
+    heuristic_profile: Mapped[dict | None] = mapped_column(JSON, nullable=True)
+    llm_profile: Mapped[dict | None] = mapped_column(JSON, nullable=True)
+    status: Mapped[str] = mapped_column(String(50), default="draft", index=True)
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True),
+        default=lambda: datetime.now(timezone.utc),
+    )
+
+    jobs: Mapped[list["ParseJob"]] = relationship(back_populates="profile")
+
+
+class ParseChannel(Base):
+    __tablename__ = "parse_channels"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, index=True)
+    name: Mapped[str] = mapped_column(String(255), nullable=False)
+    source_type: Mapped[str] = mapped_column(String(50), nullable=False, default="telegram")
+    channel: Mapped[str] = mapped_column(String(255), nullable=False)
+    is_active: Mapped[bool] = mapped_column(Boolean, default=True, nullable=False)
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True),
+        default=lambda: datetime.now(timezone.utc),
+    )
+
+    jobs: Mapped[list["ParseJob"]] = relationship(back_populates="channel")
+
+
 class ParseJob(Base):
     __tablename__ = "parse_jobs"
 
     id: Mapped[int] = mapped_column(Integer, primary_key=True, index=True)
     source_type: Mapped[str] = mapped_column(String(50), nullable=False)
     source_config: Mapped[dict] = mapped_column(JSON, nullable=False, default=dict)
+    profile_id: Mapped[int | None] = mapped_column(
+        ForeignKey("parser_profiles.id", ondelete="SET NULL"),
+        nullable=True,
+        index=True,
+    )
+    channel_id: Mapped[int | None] = mapped_column(
+        ForeignKey("parse_channels.id", ondelete="SET NULL"),
+        nullable=True,
+        index=True,
+    )
     schedule: Mapped[str | None] = mapped_column(String(100), nullable=True)
     interval_seconds: Mapped[int] = mapped_column(Integer, default=3600, nullable=False)
     is_active: Mapped[bool] = mapped_column(Boolean, default=True, nullable=False)
@@ -110,6 +155,9 @@ class ParseJob(Base):
         default=lambda: datetime.now(timezone.utc),
     )
 
+    profile: Mapped["ParserProfile | None"] = relationship(back_populates="jobs")
+    channel: Mapped["ParseChannel | None"] = relationship(back_populates="jobs")
+
 
 class Consumer(Base):
     __tablename__ = "consumers"
@@ -145,3 +193,22 @@ class ConsumerFilter(Base):
     date_from: Mapped[datetime | None] = mapped_column(DateTime(timezone=True), nullable=True)
 
     consumer: Mapped["Consumer"] = relationship(back_populates="filter")
+
+
+class VpnSettingsRow(Base):
+    """Singleton VPN / proxy settings (id=1)."""
+
+    __tablename__ = "vpn_settings"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True)
+    enabled: Mapped[bool] = mapped_column(Boolean, default=False, nullable=False)
+    mode: Mapped[str] = mapped_column(String(32), default="subscription", nullable=False)
+    subscription_url: Mapped[str | None] = mapped_column(Text, nullable=True)
+    subscription_interval_seconds: Mapped[int] = mapped_column(
+        Integer, default=3600, nullable=False
+    )
+    host: Mapped[str | None] = mapped_column(String(255), nullable=True)
+    port: Mapped[int | None] = mapped_column(Integer, nullable=True)
+    username: Mapped[str | None] = mapped_column(String(255), nullable=True)
+    password: Mapped[str | None] = mapped_column(String(255), nullable=True)
+    proxied_source_types: Mapped[list | None] = mapped_column(JSON, nullable=True)

centers/analytics/api/app/routers/admin.py

@@ -4,14 +4,17 @@ from fastapi import APIRouter, Depends, HTTPException, Query
 from sqlalchemy.orm import Session
 
 from ..database import get_db
-from ..models import Consumer, Event, ParseJob
+from ..deps import verify_admin
+from ..models import Consumer, Event, MapObject, ParseChannel, ParseJob, ParserProfile
 from ..schemas import (
     AnalyticsSummary,
     ConsumerCreate,
     ConsumerRead,
     ConsumerUpdate,
+    EventCreate,
     EventListResponse,
     EventRead,
+    EventUpdate,
     ParseJobCreate,
     ParseJobRead,
     ParseJobUpdate,
@@ -30,11 +33,12 @@ from ..services.filtering import (
     create_consumer,
     list_consumers,
     rotate_consumer_key,
+    sync_event_to_map_object,
     update_consumer,
 )
-from ..services.jobs import enqueue_job
+from ..services.jobs import enqueue_parse_job, flatten_pair_config
 
-router = APIRouter(prefix="/admin", tags=["admin"])
+router = APIRouter(prefix="/admin", tags=["admin"], dependencies=[Depends(verify_admin)])
 
 
 def _validated_source_config(source_type: str, source_config: dict) -> dict:
@@ -48,28 +52,89 @@ def _validated_source_config(source_type: str, source_config: dict) -> dict:
         raise HTTPException(status_code=400, detail=str(exc)) from exc
 
 
+def _job_to_read(job: ParseJob) -> ParseJobRead:
+    return ParseJobRead(
+        id=job.id,
+        source_type=job.source_type,
+        source_config=job.source_config or {},
+        profile_id=job.profile_id,
+        channel_id=job.channel_id,
+        schedule=job.schedule,
+        interval_seconds=job.interval_seconds,
+        is_active=job.is_active,
+        status=job.status,
+        last_run_at=job.last_run_at,
+        last_error=job.last_error,
+        created_at=job.created_at,
+        profile_name=job.profile.name if job.profile else None,
+        channel_name=job.channel.name if job.channel else None,
+        channel_handle=job.channel.channel if job.channel else None,
+    )
+
+
 @router.post("/jobs", response_model=ParseJobRead, status_code=201)
 def create_parse_job(payload: ParseJobCreate, db: Session = Depends(get_db)):
-    source_config = _validated_source_config(payload.source_type, payload.source_config)
-    job = ParseJob(
-        source_type=payload.source_type,
-        source_config=source_config,
-        schedule=payload.schedule,
-        interval_seconds=payload.interval_seconds,
-        is_active=payload.is_active,
-        status="queued",
-    )
+    pair_mode = payload.profile_id is not None or payload.channel_id is not None
+    if pair_mode:
+        if payload.profile_id is None or payload.channel_id is None:
+            raise HTTPException(
+                status_code=400,
+                detail="Both profile_id and channel_id are required to create a pair",
+            )
+        profile = db.query(ParserProfile).filter(ParserProfile.id == payload.profile_id).first()
+        channel = db.query(ParseChannel).filter(ParseChannel.id == payload.channel_id).first()
+        if not profile:
+            raise HTTPException(status_code=404, detail="Profile not found")
+        if not channel:
+            raise HTTPException(status_code=404, detail="Channel not found")
+        kind = (profile.kind or "heuristic").strip().lower()
+        if kind == "llm":
+            if not profile.llm_profile:
+                raise HTTPException(status_code=400, detail="LLM profile has no llm_profile")
+        elif not profile.heuristic_profile:
+            raise HTTPException(status_code=400, detail="Profile has no heuristic_profile")
+        if not channel.is_active:
+            raise HTTPException(status_code=400, detail="Channel is inactive")
+
+        try:
+            source_config = flatten_pair_config(channel, profile, limit=payload.limit)
+        except Exception as exc:
+            raise HTTPException(status_code=400, detail=str(exc)) from exc
+
+        job = ParseJob(
+            source_type=channel.source_type or "telegram",
+            source_config=source_config,
+            profile_id=profile.id,
+            channel_id=channel.id,
+            schedule=payload.schedule,
+            interval_seconds=payload.interval_seconds,
+            is_active=payload.is_active,
+            status="queued",
+        )
+    else:
+        source_config = _validated_source_config(payload.source_type, payload.source_config)
+        job = ParseJob(
+            source_type=payload.source_type,
+            source_config=source_config,
+            schedule=payload.schedule,
+            interval_seconds=payload.interval_seconds,
+            is_active=payload.is_active,
+            status="queued",
+        )
+
     db.add(job)
     db.commit()
     db.refresh(job)
 
-    enqueue_job(job.id, job.source_type, job.source_config)
-    return job
+    enqueue_parse_job(db, job)
+    db.refresh(job)
+    return _job_to_read(job)
 
 
 @router.get("/jobs", response_model=list[ParseJobRead])
 def list_parse_jobs(db: Session = Depends(get_db)):
-    return db.query(ParseJob).order_by(ParseJob.id.desc()).all()
+    jobs = db.query(ParseJob).order_by(ParseJob.id.desc()).all()
+    return [_job_to_read(job) for job in jobs]
 
 
 @router.get("/jobs/{job_id}", response_model=ParseJobRead)
@@ -77,7 +142,7 @@ def get_parse_job(job_id: int, db: Session = Depends(get_db)):
     job = db.query(ParseJob).filter(ParseJob.id == job_id).first()
     if not job:
         raise HTTPException(status_code=404, detail="Job not found")
-    return job
+    return _job_to_read(job)
 
 
 @router.post("/jobs/{job_id}/retry", response_model=ParseJobRead)
@@ -85,7 +150,10 @@ def retry_parse_job(job_id: int, db: Session = Depends(get_db)):
     job = db.query(ParseJob).filter(ParseJob.id == job_id).first()
     if not job:
         raise HTTPException(status_code=404, detail="Job not found")
-    if job.status in ("queued", "running"):
+
+    from ..services.job_stale import is_stale_job
+
+    if job.status in ("queued", "running") and not is_stale_job(job):
         raise HTTPException(status_code=409, detail="Job is already running or queued")
 
     job.status = "queued"
@@ -93,8 +161,12 @@ def retry_parse_job(job_id: int, db: Session = Depends(get_db)):
     db.commit()
     db.refresh(job)
 
-    enqueue_job(job.id, job.source_type, job.source_config)
-    return job
+    try:
+        enqueue_parse_job(db, job)
+    except ValueError as exc:
+        raise HTTPException(status_code=400, detail=str(exc)) from exc
+    db.refresh(job)
+    return _job_to_read(job)
 
 
 @router.patch("/jobs/{job_id}", response_model=ParseJobRead)
@@ -107,8 +179,36 @@ def update_parse_job(
     if not job:
         raise HTTPException(status_code=404, detail="Job not found")
 
-    if payload.source_config is not None:
+    if payload.profile_id is not None:
+        profile = db.query(ParserProfile).filter(ParserProfile.id == payload.profile_id).first()
+        if not profile:
+            raise HTTPException(status_code=404, detail="Profile not found")
+        job.profile_id = profile.id
+    if payload.channel_id is not None:
+        channel = db.query(ParseChannel).filter(ParseChannel.id == payload.channel_id).first()
+        if not channel:
+            raise HTTPException(status_code=404, detail="Channel not found")
+        job.channel_id = channel.id
+
+    if job.profile_id and job.channel_id:
+        profile = db.query(ParserProfile).filter(ParserProfile.id == job.profile_id).first()
+        channel = db.query(ParseChannel).filter(ParseChannel.id == job.channel_id).first()
+        if profile and channel:
+            existing_limit = (job.source_config or {}).get("limit", 100)
+            limit = payload.limit if payload.limit is not None else existing_limit
+            if not isinstance(limit, int):
+                limit = 100
+            try:
+                job.source_config = flatten_pair_config(channel, profile, limit=limit)
+            except Exception as exc:
+                raise HTTPException(status_code=400, detail=str(exc)) from exc
+    elif payload.source_config is not None:
         job.source_config = _validated_source_config(job.source_type, payload.source_config)
+    elif payload.limit is not None and isinstance(job.source_config, dict):
+        cfg = dict(job.source_config)
+        cfg["limit"] = payload.limit
+        job.source_config = _validated_source_config(job.source_type, cfg)
+
     if payload.interval_seconds is not None:
         job.interval_seconds = payload.interval_seconds
     if payload.is_active is not None:
@@ -116,7 +216,7 @@ def update_parse_job(
 
     db.commit()
     db.refresh(job)
-    return job
+    return _job_to_read(job)
 
 
 @router.delete("/jobs/{job_id}", status_code=204)
@@ -125,7 +225,14 @@ def delete_parse_job(job_id: int, db: Session = Depends(get_db)):
     if not job:
         raise HTTPException(status_code=404, detail="Job not found")
     if job.status == "running":
-        raise HTTPException(status_code=409, detail="Cannot delete a running job")
+        from ..services.job_stale import is_stale_job
+
+        if not is_stale_job(job):
+            raise HTTPException(status_code=409, detail="Cannot delete a running job")
+        # Stale running — allow delete after marking failed for audit trail
+        job.status = "failed"
+        job.last_error = "Deleted while stale running"
+        db.commit()
 
     db.delete(job)
     db.commit()
@@ -159,6 +266,112 @@ def list_events(
     return EventListResponse(items=items, total=total)
 
 
+def _ensure_unique_source_url(db: Session, source_url: str, *, exclude_id: int | None = None) -> None:
+    query = db.query(Event).filter(Event.source_url == source_url)
+    if exclude_id is not None:
+        query = query.filter(Event.id != exclude_id)
+    if query.first():
+        raise HTTPException(status_code=409, detail="Event with this source_url already exists")
+
+
+def _sync_or_clear_map_object(db: Session, event: Event) -> None:
+    if event.latitude is not None and event.longitude is not None:
+        sync_event_to_map_object(db, event)
+        return
+    linked = db.query(MapObject).filter(MapObject.event_id == event.id).first()
+    if linked:
+        db.delete(linked)
+
+
+@router.post("/events", response_model=EventRead, status_code=201)
+def create_event(payload: EventCreate, db: Session = Depends(get_db)):
+    source_url = payload.source_url.strip()
+    if not source_url:
+        raise HTTPException(status_code=400, detail="source_url is required")
+    _ensure_unique_source_url(db, source_url)
+
+    if (payload.latitude is None) != (payload.longitude is None):
+        raise HTTPException(status_code=400, detail="latitude and longitude must be set together")
+
+    event = Event(
+        source_type=payload.source_type.strip() or "manual",
+        source_url=source_url,
+        raw_text=payload.raw_text or "",
+        title=payload.title or "",
+        description=payload.description or "",
+        locality=payload.locality or "",
+        latitude=payload.latitude,
+        longitude=payload.longitude,
+        event_date=payload.event_date,
+        region=payload.region,
+        topic=payload.topic,
+        tags=payload.tags,
+        metadata_=payload.metadata,
+    )
+    db.add(event)
+    db.flush()
+    _sync_or_clear_map_object(db, event)
+    db.commit()
+    db.refresh(event)
+    return event
+
+
+@router.get("/events/{event_id}", response_model=EventRead)
+def get_event(event_id: int, db: Session = Depends(get_db)):
+    event = db.query(Event).filter(Event.id == event_id).first()
+    if not event:
+        raise HTTPException(status_code=404, detail="Event not found")
+    return event
+
+
+@router.patch("/events/{event_id}", response_model=EventRead)
+def update_event(event_id: int, payload: EventUpdate, db: Session = Depends(get_db)):
+    event = db.query(Event).filter(Event.id == event_id).first()
+    if not event:
+        raise HTTPException(status_code=404, detail="Event not found")
+
+    data = payload.model_dump(exclude_unset=True)
+    if "source_url" in data and data["source_url"] is not None:
+        source_url = data["source_url"].strip()
+        if not source_url:
+            raise HTTPException(status_code=400, detail="source_url cannot be empty")
+        _ensure_unique_source_url(db, source_url, exclude_id=event.id)
+        event.source_url = source_url
+        data.pop("source_url")
+
+    if "metadata" in data:
+        event.metadata_ = data.pop("metadata")
+
+    if "source_type" in data and data["source_type"] is not None:
+        event.source_type = data.pop("source_type").strip() or event.source_type
+
+    for field, value in data.items():
+        setattr(event, field, value)
+
+    lat = event.latitude
+    lon = event.longitude
+    if (lat is None) != (lon is None):
+        raise HTTPException(status_code=400, detail="latitude and longitude must be set together")
+
+    _sync_or_clear_map_object(db, event)
+    db.commit()
+    db.refresh(event)
+    return event
+
+
+@router.delete("/events/{event_id}", status_code=204)
+def delete_event(event_id: int, db: Session = Depends(get_db)):
+    event = db.query(Event).filter(Event.id == event_id).first()
+    if not event:
+        raise HTTPException(status_code=404, detail="Event not found")
+
+    linked = db.query(MapObject).filter(MapObject.event_id == event.id).all()
+    for obj in linked:
+        db.delete(obj)
+    db.delete(event)
+    db.commit()
+
+
 @router.get("/analytics/summary", response_model=AnalyticsSummary)
 def analytics_summary(db: Session = Depends(get_db)):
     return get_analytics_summary(db)

centers/analytics/api/app/routers/auth.py

@@ -0,0 +1,41 @@
+from fastapi import APIRouter, Depends, HTTPException
+from pydantic import BaseModel, Field
+
+from ..auth import (
+    admin_credentials_configured,
+    create_access_token,
+    verify_credentials,
+)
+from ..deps import verify_admin
+
+router = APIRouter(prefix="/admin/auth", tags=["auth"])
+
+
+class LoginRequest(BaseModel):
+    username: str = Field(min_length=1)
+    password: str = Field(min_length=1)
+
+
+class TokenResponse(BaseModel):
+    access_token: str
+    token_type: str = "bearer"
+
+
+class MeResponse(BaseModel):
+    username: str
+    role: str = "admin"
+
+
+@router.post("/login", response_model=TokenResponse)
+def login(payload: LoginRequest):
+    if not admin_credentials_configured():
+        raise HTTPException(status_code=503, detail="Admin auth is not configured")
+    if not verify_credentials(payload.username, payload.password):
+        raise HTTPException(status_code=401, detail="Invalid username or password")
+    token = create_access_token(username=payload.username.strip())
+    return TokenResponse(access_token=token)
+
+
+@router.get("/me", response_model=MeResponse)
+def me(username: str = Depends(verify_admin)):
+    return MeResponse(username=username)

centers/analytics/api/app/routers/internal.py

@@ -6,8 +6,15 @@ from sqlalchemy.orm import Session
 from ..database import get_db
 from ..deps import verify_internal_token
 from ..models import ParseJob
-from ..schemas import IngestRequest, IngestResponse, ListenerSubscription
+from ..schemas import (
+    IngestRequest,
+    IngestResponse,
+    ListenerSubscription,
+    VpnSettingsInternalRead,
+)
 from ..services.ingest import ingest_events
+from ..services.jobs import resolve_job_source_config
+from ..services.vpn import ensure_vpn_settings, to_internal_read
 
 router = APIRouter(prefix="/internal", tags=["internal"])
 
@@ -45,7 +52,8 @@ def update_job_status(
         raise HTTPException(status_code=404, detail="Job not found")
 
     job.status = status
-    if status in ("completed", "failed"):
+    # Anchor staleness detection: running/queued start, and terminal finish
+    if status in ("running", "queued", "completed", "failed"):
         job.last_run_at = datetime.now(timezone.utc)
     job.last_error = error
     db.commit()
@@ -68,8 +76,28 @@ def listener_subscriptions(
     )
     result: list[ListenerSubscription] = []
     for job in jobs:
-        channel = job.source_config.get("channel") if job.source_config else None
+        try:
+            source_config = resolve_job_source_config(db, job)
+        except ValueError:
+            continue
+        channel = source_config.get("channel")
         if not channel:
             continue
-        result.append(ListenerSubscription(job_id=job.id, channel=str(channel)))
+        result.append(
+            ListenerSubscription(
+                job_id=job.id,
+                channel=str(channel),
+                source_config=dict(source_config),
+            )
+        )
+    db.commit()
     return result
+
+
+@router.get("/vpn", response_model=VpnSettingsInternalRead)
+def internal_vpn_settings(
+    _: None = Depends(verify_internal_token),
+    db: Session = Depends(get_db),
+):
+    row = ensure_vpn_settings(db)
+    return to_internal_read(row)

centers/analytics/api/app/routers/objects.py

@@ -3,6 +3,7 @@ from fastapi.responses import FileResponse
 from sqlalchemy.orm import Session
 
 from ..database import get_db
+from ..deps import verify_admin
 from ..models import MapObject, ObjectMedia
 from ..schemas import MapObjectCreate, MapObjectRead, MapObjectUpdate, ObjectMediaRead
 from ..storage import (
@@ -48,7 +49,11 @@ def get_object(object_id: int, db: Session = Depends(get_db)):
 
 
 @router.post("/api/objects", response_model=MapObjectRead, status_code=201)
-def create_object(payload: MapObjectCreate, db: Session = Depends(get_db)):
+def create_object(
+    payload: MapObjectCreate,
+    db: Session = Depends(get_db),
+    _: str = Depends(verify_admin),
+):
     data = payload.model_dump()
     created_at = data.pop("created_at", None)
     obj = MapObject(**data)
@@ -65,6 +70,7 @@ def update_object(
     object_id: int,
     payload: MapObjectUpdate,
     db: Session = Depends(get_db),
+    _: str = Depends(verify_admin),
 ):
     obj = db.query(MapObject).filter(MapObject.id == object_id).first()
     if not obj:
@@ -79,7 +85,11 @@ def update_object(
 
 
 @router.delete("/api/objects/{object_id}", status_code=204)
-def delete_object(object_id: int, db: Session = Depends(get_db)):
+def delete_object(
+    object_id: int,
+    db: Session = Depends(get_db),
+    _: str = Depends(verify_admin),
+):
     obj = db.query(MapObject).filter(MapObject.id == object_id).first()
     if not obj:
         raise HTTPException(status_code=404, detail="Объект не найден")
@@ -111,6 +121,7 @@ async def upload_object_media(
     object_id: int,
     file: UploadFile = File(...),
     db: Session = Depends(get_db),
+    _: str = Depends(verify_admin),
 ):
     obj = db.query(MapObject).filter(MapObject.id == object_id).first()
     if not obj:
@@ -159,7 +170,11 @@ def get_media_file(media_id: int, db: Session = Depends(get_db)):
 
 
 @router.delete("/api/media/{media_id}", status_code=204)
-def delete_media(media_id: int, db: Session = Depends(get_db)):
+def delete_media(
+    media_id: int,
+    db: Session = Depends(get_db),
+    _: str = Depends(verify_admin),
+):
     media = db.query(ObjectMedia).filter(ObjectMedia.id == media_id).first()
     if not media:
         raise HTTPException(status_code=404, detail="Медиафайл не найден")

centers/analytics/api/app/routers/parse_channels.py

@@ -0,0 +1,103 @@
+"""Admin CRUD for parse channels (Telegram handles in MVP)."""
+
+from __future__ import annotations
+
+from fastapi import APIRouter, Depends, HTTPException
+from sqlalchemy.orm import Session
+
+from ..database import get_db
+from ..deps import verify_admin
+from ..models import ParseChannel, ParseJob
+from ..schemas import ParseChannelCreate, ParseChannelRead, ParseChannelUpdate
+
+router = APIRouter(
+    prefix="/admin/parse-channels",
+    tags=["parse-channels"],
+    dependencies=[Depends(verify_admin)],
+)
+
+ALLOWED_SOURCE_TYPES = {"telegram"}
+
+
+@router.get("", response_model=list[ParseChannelRead])
+def list_channels(db: Session = Depends(get_db)):
+    return db.query(ParseChannel).order_by(ParseChannel.id.desc()).all()
+
+
+@router.post("", response_model=ParseChannelRead, status_code=201)
+def create_channel(payload: ParseChannelCreate, db: Session = Depends(get_db)):
+    source_type = (payload.source_type or "telegram").strip()
+    if source_type not in ALLOWED_SOURCE_TYPES:
+        raise HTTPException(
+            status_code=400,
+            detail=f"source_type must be one of: {', '.join(sorted(ALLOWED_SOURCE_TYPES))}",
+        )
+    channel = ParseChannel(
+        name=payload.name.strip(),
+        source_type=source_type,
+        channel=payload.channel.strip().lstrip("@"),
+        is_active=payload.is_active,
+    )
+    db.add(channel)
+    db.commit()
+    db.refresh(channel)
+    return channel
+
+
+@router.get("/{channel_id}", response_model=ParseChannelRead)
+def get_channel(channel_id: int, db: Session = Depends(get_db)):
+    channel = db.query(ParseChannel).filter(ParseChannel.id == channel_id).first()
+    if not channel:
+        raise HTTPException(status_code=404, detail="Channel not found")
+    return channel
+
+
+@router.patch("/{channel_id}", response_model=ParseChannelRead)
+def update_channel(
+    channel_id: int,
+    payload: ParseChannelUpdate,
+    db: Session = Depends(get_db),
+):
+    channel = db.query(ParseChannel).filter(ParseChannel.id == channel_id).first()
+    if not channel:
+        raise HTTPException(status_code=404, detail="Channel not found")
+
+    if payload.name is not None:
+        channel.name = payload.name.strip()
+    if payload.source_type is not None:
+        source_type = payload.source_type.strip()
+        if source_type not in ALLOWED_SOURCE_TYPES:
+            raise HTTPException(
+                status_code=400,
+                detail=f"source_type must be one of: {', '.join(sorted(ALLOWED_SOURCE_TYPES))}",
+            )
+        channel.source_type = source_type
+    if payload.channel is not None:
+        channel.channel = payload.channel.strip().lstrip("@")
+    if payload.is_active is not None:
+        channel.is_active = payload.is_active
+
+    db.commit()
+    db.refresh(channel)
+    return channel
+
+
+@router.delete("/{channel_id}", status_code=204)
+def delete_channel(channel_id: int, db: Session = Depends(get_db)):
+    channel = db.query(ParseChannel).filter(ParseChannel.id == channel_id).first()
+    if not channel:
+        raise HTTPException(status_code=404, detail="Channel not found")
+
+    linked = (
+        db.query(ParseJob)
+        .filter(ParseJob.channel_id == channel_id, ParseJob.status.in_(("queued", "running")))
+        .count()
+    )
+    if linked:
+        raise HTTPException(
+            status_code=409,
+            detail="Cannot delete channel used by queued/running jobs",
+        )
+
+    db.delete(channel)
+    db.commit()

centers/analytics/api/app/routers/parser_profiles.py

@@ -0,0 +1,304 @@
+"""Admin CRUD for reusable parser profiles + generate/preview."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from fastapi import APIRouter, Depends, HTTPException
+from pydantic import BaseModel, Field
+from sqlalchemy.orm import Session
+
+from contracts.heuristic_profile import HeuristicProfile
+from contracts.llm_profile import LlmProfile
+
+from ..database import get_db
+from ..deps import verify_admin
+from ..models import ParseJob, ParserProfile
+from ..schemas import ParserProfileCreate, ParserProfileRead, ParserProfileUpdate
+from ..services import parser_builder as builder
+
+router = APIRouter(
+    prefix="/admin/parser-profiles",
+    tags=["parser-profiles"],
+    dependencies=[Depends(verify_admin)],
+)
+
+
+class GenerateRequest(BaseModel):
+    sample_post: str = Field(min_length=1)
+    hint: str | None = None
+    current_profile: dict[str, Any] | None = None
+
+
+class GenerateResponse(BaseModel):
+    profile: dict[str, Any]
+    preview: dict[str, str]
+    empty_fields: list[str] = Field(default_factory=list)
+
+
+class PreviewRequest(BaseModel):
+    sample_post: str = Field(min_length=1)
+    heuristic_profile: dict[str, Any] | None = None
+    # Accept alias used by older builder UI
+    profile: dict[str, Any] | None = None
+
+
+class PreviewResponse(BaseModel):
+    fields: dict[str, str]
+    matched: bool = True
+    missing_required: list[str] = Field(default_factory=list)
+
+
+class PreviewLlmRequest(BaseModel):
+    sample_post: str = Field(min_length=1)
+    llm_profile: dict[str, Any]
+
+
+class PreviewLlmResponse(BaseModel):
+    fields: dict[str, str]
+    events: list[dict[str, str]] = Field(default_factory=list)
+    matched: bool = True
+    missing_required: list[str] = Field(default_factory=list)
+    is_event: bool = True
+    matched_count: int = 0
+
+
+def _validate_heuristic_profile(raw: dict[str, Any] | None) -> dict[str, Any] | None:
+    if raw is None:
+        return None
+    try:
+        return HeuristicProfile.model_validate(raw).model_dump()
+    except Exception as exc:
+        raise HTTPException(status_code=400, detail=f"Invalid heuristic_profile: {exc}") from exc
+
+
+def _validate_llm_profile(raw: dict[str, Any] | None) -> dict[str, Any] | None:
+    if raw is None:
+        return None
+    try:
+        return LlmProfile.model_validate(raw).model_dump()
+    except Exception as exc:
+        raise HTTPException(status_code=400, detail=f"Invalid llm_profile: {exc}") from exc
+
+
+def _normalize_kind(kind: str | None) -> str:
+    value = (kind or "heuristic").strip().lower()
+    if value not in ("heuristic", "llm"):
+        raise HTTPException(status_code=400, detail="kind must be heuristic or llm")
+    return value
+
+
+def _profile_status(
+    *,
+    kind: str,
+    heuristic_profile: dict | None,
+    llm_profile: dict | None,
+    explicit: str | None = None,
+) -> str:
+    if explicit:
+        return explicit
+    if kind == "llm":
+        return "ready" if llm_profile else "draft"
+    return "ready" if heuristic_profile else "draft"
+
+
+@router.get("/target-fields")
+def target_fields():
+    return {"fields": builder.get_target_fields()}
+
+
+@router.get("/llm-defaults")
+def llm_defaults():
+    from contracts.llm_profile import DEFAULT_EXTRACT_SCHEMA, DEFAULT_INSTRUCTION
+
+    return {
+        "instruction": DEFAULT_INSTRUCTION,
+        "extract_schema": DEFAULT_EXTRACT_SCHEMA,
+        "required_fields": [],
+        "multi_event": False,
+    }
+
+
+@router.post("/generate", response_model=GenerateResponse)
+async def generate_profile(payload: GenerateRequest):
+    if not builder.deepseek_enabled():
+        raise HTTPException(
+            status_code=503,
+            detail="DEEPSEEK_API_KEY is not set on ca-api. Add it to .env for parser generation.",
+        )
+    try:
+        profile = await builder.generate_profile(
+            payload.sample_post,
+            hint=payload.hint,
+            current_profile=payload.current_profile,
+        )
+    except ValueError as exc:
+        raise HTTPException(status_code=400, detail=str(exc)) from exc
+    except Exception as exc:
+        raise HTTPException(
+            status_code=502,
+            detail=f"DeepSeek generate failed: {exc}",
+        ) from exc
+    dumped = profile.model_dump()
+    if payload.current_profile and isinstance(payload.current_profile.get("required_fields"), list):
+        dumped["required_fields"] = payload.current_profile["required_fields"]
+        dumped = HeuristicProfile.model_validate(dumped).model_dump()
+    preview = builder.preview_with_profile(payload.sample_post, dumped)
+    return GenerateResponse(
+        profile=dumped,
+        preview=preview,
+        empty_fields=builder.empty_preview_fields(preview),
+    )
+
+
+@router.post("/preview", response_model=PreviewResponse)
+def preview_profile(payload: PreviewRequest):
+    raw = payload.heuristic_profile if payload.heuristic_profile is not None else payload.profile
+    if raw is None:
+        raise HTTPException(status_code=400, detail="heuristic_profile is required")
+    try:
+        HeuristicProfile.model_validate(raw)
+        fields, matched, missing = builder.match_preview(payload.sample_post, raw)
+    except Exception as exc:
+        raise HTTPException(status_code=400, detail=str(exc)) from exc
+    return PreviewResponse(fields=fields, matched=matched, missing_required=missing)
+
+
+@router.post("/preview-llm", response_model=PreviewLlmResponse)
+async def preview_llm_profile(payload: PreviewLlmRequest):
+    if not builder.deepseek_enabled():
+        raise HTTPException(
+            status_code=503,
+            detail="DEEPSEEK_API_KEY is not set on ca-api. Add it to .env for LLM preview.",
+        )
+    try:
+        LlmProfile.model_validate(payload.llm_profile)
+        events, fields, matched, missing, is_event, matched_count = await builder.preview_llm_extract(
+            payload.sample_post,
+            payload.llm_profile,
+        )
+    except ValueError as exc:
+        raise HTTPException(status_code=400, detail=str(exc)) from exc
+    except Exception as exc:
+        raise HTTPException(
+            status_code=502,
+            detail=f"DeepSeek LLM preview failed: {exc}",
+        ) from exc
+    return PreviewLlmResponse(
+        fields=fields,
+        events=events,
+        matched=matched,
+        missing_required=missing,
+        is_event=is_event,
+        matched_count=matched_count,
+    )
+
+
+@router.get("", response_model=list[ParserProfileRead])
+def list_profiles(db: Session = Depends(get_db)):
+    return db.query(ParserProfile).order_by(ParserProfile.id.desc()).all()
+
+
+@router.post("", response_model=ParserProfileRead, status_code=201)
+def create_profile(payload: ParserProfileCreate, db: Session = Depends(get_db)):
+    kind = _normalize_kind(payload.kind)
+    heuristic = _validate_heuristic_profile(payload.heuristic_profile)
+    llm = _validate_llm_profile(payload.llm_profile)
+    if kind == "heuristic" and llm and not heuristic:
+        # ignore stray llm blob when creating heuristic
+        llm = None
+    if kind == "llm" and heuristic and not llm:
+        heuristic = None
+    profile = ParserProfile(
+        name=payload.name.strip(),
+        kind=kind,
+        sample_post=payload.sample_post or "",
+        heuristic_profile=heuristic if kind == "heuristic" else None,
+        llm_profile=llm if kind == "llm" else None,
+        status=_profile_status(
+            kind=kind,
+            heuristic_profile=heuristic if kind == "heuristic" else None,
+            llm_profile=llm if kind == "llm" else None,
+            explicit=payload.status,
+        ),
+    )
+    db.add(profile)
+    db.commit()
+    db.refresh(profile)
+    return profile
+
+
+@router.get("/{profile_id}", response_model=ParserProfileRead)
+def get_profile(profile_id: int, db: Session = Depends(get_db)):
+    profile = db.query(ParserProfile).filter(ParserProfile.id == profile_id).first()
+    if not profile:
+        raise HTTPException(status_code=404, detail="Profile not found")
+    return profile
+
+
+@router.patch("/{profile_id}", response_model=ParserProfileRead)
+def update_profile(
+    profile_id: int,
+    payload: ParserProfileUpdate,
+    db: Session = Depends(get_db),
+):
+    profile = db.query(ParserProfile).filter(ParserProfile.id == profile_id).first()
+    if not profile:
+        raise HTTPException(status_code=404, detail="Profile not found")
+
+    if payload.name is not None:
+        profile.name = payload.name.strip()
+    if payload.sample_post is not None:
+        profile.sample_post = payload.sample_post
+    if payload.kind is not None:
+        profile.kind = _normalize_kind(payload.kind)
+    kind = _normalize_kind(profile.kind)
+
+    if "heuristic_profile" in payload.model_fields_set:
+        profile.heuristic_profile = _validate_heuristic_profile(payload.heuristic_profile)
+    if "llm_profile" in payload.model_fields_set:
+        profile.llm_profile = _validate_llm_profile(payload.llm_profile)
+
+    # Keep only the blob matching kind
+    if kind == "heuristic":
+        profile.llm_profile = None
+    else:
+        profile.heuristic_profile = None
+
+    if payload.status is not None:
+        profile.status = payload.status
+    elif (
+        "heuristic_profile" in payload.model_fields_set
+        or "llm_profile" in payload.model_fields_set
+        or payload.kind is not None
+    ):
+        profile.status = _profile_status(
+            kind=kind,
+            heuristic_profile=profile.heuristic_profile,
+            llm_profile=profile.llm_profile,
+        )
+
+    db.commit()
+    db.refresh(profile)
+    return profile
+
+
+@router.delete("/{profile_id}", status_code=204)
+def delete_profile(profile_id: int, db: Session = Depends(get_db)):
+    profile = db.query(ParserProfile).filter(ParserProfile.id == profile_id).first()
+    if not profile:
+        raise HTTPException(status_code=404, detail="Profile not found")
+
+    linked = (
+        db.query(ParseJob)
+        .filter(ParseJob.profile_id == profile_id, ParseJob.status.in_(("queued", "running")))
+        .count()
+    )
+    if linked:
+        raise HTTPException(
+            status_code=409,
+            detail="Cannot delete profile used by queued/running jobs",
+        )
+
+    db.delete(profile)
+    db.commit()

centers/analytics/api/app/routers/vpn.py

@@ -0,0 +1,32 @@
+"""Admin VPN / proxy settings."""
+
+from __future__ import annotations
+
+from fastapi import APIRouter, Depends, HTTPException
+from sqlalchemy.orm import Session
+
+from ..database import get_db
+from ..deps import verify_admin
+from ..schemas import VpnSettingsAdminRead, VpnSettingsUpdate
+from ..services.vpn import apply_vpn_update, ensure_vpn_settings, to_admin_read
+
+router = APIRouter(
+    prefix="/admin/vpn",
+    tags=["vpn"],
+    dependencies=[Depends(verify_admin)],
+)
+
+
+@router.get("", response_model=VpnSettingsAdminRead)
+def get_vpn_settings(db: Session = Depends(get_db)):
+    row = ensure_vpn_settings(db)
+    return to_admin_read(row)
+
+
+@router.put("", response_model=VpnSettingsAdminRead)
+def put_vpn_settings(payload: VpnSettingsUpdate, db: Session = Depends(get_db)):
+    try:
+        row = apply_vpn_update(db, payload)
+    except ValueError as exc:
+        raise HTTPException(status_code=400, detail=str(exc)) from exc
+    return to_admin_read(row)

centers/analytics/api/app/schemas.py

@@ -92,6 +92,38 @@ class EventRead(BaseModel):
     metadata: dict[str, Any] | None = Field(validation_alias="metadata_")
 
 
+class EventCreate(BaseModel):
+    source_type: str = Field(default="manual", min_length=1, max_length=50)
+    source_url: str = Field(min_length=1, max_length=512)
+    raw_text: str = ""
+    title: str = ""
+    description: str = ""
+    locality: str = ""
+    latitude: float | None = Field(default=None, ge=-90, le=90)
+    longitude: float | None = Field(default=None, ge=-180, le=180)
+    event_date: datetime | None = None
+    region: str | None = None
+    topic: str | None = None
+    tags: list[str] | None = None
+    metadata: dict[str, Any] | None = None
+
+
+class EventUpdate(BaseModel):
+    source_type: str | None = Field(default=None, min_length=1, max_length=50)
+    source_url: str | None = Field(default=None, min_length=1, max_length=512)
+    raw_text: str | None = None
+    title: str | None = None
+    description: str | None = None
+    locality: str | None = None
+    latitude: float | None = Field(default=None, ge=-90, le=90)
+    longitude: float | None = Field(default=None, ge=-180, le=180)
+    event_date: datetime | None = None
+    region: str | None = None
+    topic: str | None = None
+    tags: list[str] | None = None
+    metadata: dict[str, Any] | None = None
+
+
 class IngestEventItem(BaseModel):
     source_type: str = "telegram"
     source_url: str
@@ -117,6 +149,7 @@ class IngestRequest(BaseModel):
 class ListenerSubscription(BaseModel):
     job_id: int
     channel: str
+    source_config: dict[str, Any] = Field(default_factory=dict)
 
 
 class IngestResponse(BaseModel):
@@ -126,9 +159,68 @@ class IngestResponse(BaseModel):
     map_objects_synced: int
 
 
+class ParserProfileCreate(BaseModel):
+    name: str = Field(min_length=1, max_length=255)
+    kind: str = Field(default="heuristic", pattern="^(heuristic|llm)$")
+    sample_post: str = ""
+    heuristic_profile: dict[str, Any] | None = None
+    llm_profile: dict[str, Any] | None = None
+    status: str | None = None
+
+
+class ParserProfileUpdate(BaseModel):
+    name: str | None = Field(default=None, min_length=1, max_length=255)
+    kind: str | None = Field(default=None, pattern="^(heuristic|llm)$")
+    sample_post: str | None = None
+    heuristic_profile: dict[str, Any] | None = None
+    llm_profile: dict[str, Any] | None = None
+    status: str | None = None
+
+
+class ParserProfileRead(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: int
+    name: str
+    kind: str = "heuristic"
+    sample_post: str
+    heuristic_profile: dict[str, Any] | None
+    llm_profile: dict[str, Any] | None = None
+    status: str
+    created_at: datetime
+
+
+class ParseChannelCreate(BaseModel):
+    name: str = Field(min_length=1, max_length=255)
+    source_type: str = "telegram"
+    channel: str = Field(min_length=1, max_length=255)
+    is_active: bool = True
+
+
+class ParseChannelUpdate(BaseModel):
+    name: str | None = Field(default=None, min_length=1, max_length=255)
+    source_type: str | None = None
+    channel: str | None = Field(default=None, min_length=1, max_length=255)
+    is_active: bool | None = None
+
+
+class ParseChannelRead(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: int
+    name: str
+    source_type: str
+    channel: str
+    is_active: bool
+    created_at: datetime
+
+
 class ParseJobCreate(BaseModel):
     source_type: str = "telegram"
     source_config: dict[str, Any] = Field(default_factory=dict)
+    profile_id: int | None = None
+    channel_id: int | None = None
+    limit: int = Field(default=100, ge=1, le=1000)
     schedule: str | None = None
     interval_seconds: int = Field(default=3600, ge=60, le=604800)
     is_active: bool = True
@@ -136,6 +228,9 @@ class ParseJobCreate(BaseModel):
 
 class ParseJobUpdate(BaseModel):
     source_config: dict[str, Any] | None = None
+    profile_id: int | None = None
+    channel_id: int | None = None
+    limit: int | None = Field(default=None, ge=1, le=1000)
     interval_seconds: int | None = Field(default=None, ge=60, le=604800)
     is_active: bool | None = None
 
@@ -146,6 +241,8 @@ class ParseJobRead(BaseModel):
     id: int
     source_type: str
     source_config: dict[str, Any]
+    profile_id: int | None = None
+    channel_id: int | None = None
     schedule: str | None
     interval_seconds: int
     is_active: bool
@@ -153,6 +250,9 @@ class ParseJobRead(BaseModel):
     last_run_at: datetime | None
     last_error: str | None
     created_at: datetime
+    profile_name: str | None = None
+    channel_name: str | None = None
+    channel_handle: str | None = None
 
 
 class ConsumerCreate(BaseModel):
@@ -206,3 +306,43 @@ class TimelinePoint(BaseModel):
 class TopItem(BaseModel):
     name: str
     count: int
+
+
+class VpnSettingsUpdate(BaseModel):
+    enabled: bool | None = None
+    mode: Literal["subscription", "socks5", "http"] | None = None
+    subscription_url: str | None = None
+    clear_subscription_url: bool | None = None
+    subscription_interval_seconds: int | None = Field(default=None, ge=60, le=86400)
+    host: str | None = None
+    port: int | None = Field(default=None, ge=1, le=65535)
+    username: str | None = None
+    password: str | None = None
+    clear_password: bool | None = None
+    proxied_source_types: list[str] | None = None
+
+
+class VpnSettingsAdminRead(BaseModel):
+    enabled: bool
+    mode: Literal["subscription", "socks5", "http"]
+    subscription_url_set: bool
+    subscription_url_hint: str | None = None
+    subscription_interval_seconds: int
+    host: str | None = None
+    port: int | None = None
+    username: str | None = None
+    password_set: bool
+    proxied_source_types: list[str]
+    available_source_types: list[str]
+
+
+class VpnSettingsInternalRead(BaseModel):
+    enabled: bool
+    mode: Literal["subscription", "socks5", "http"]
+    subscription_url: str | None = None
+    subscription_interval_seconds: int
+    host: str | None = None
+    port: int | None = None
+    username: str | None = None
+    password: str | None = None
+    proxied_source_types: list[str]

centers/analytics/api/app/seed.py

@@ -4,6 +4,7 @@ from sqlalchemy.orm import Session
 
 from .models import Consumer, ConsumerFilter
 from .services.filtering import hash_api_key
+from .services.vpn import ensure_vpn_settings
 
 
 def seed_objects(db: Session) -> None:
@@ -24,3 +25,7 @@ def seed_test_consumer(db: Session) -> None:
     db.flush()
     db.add(ConsumerFilter(consumer_id=consumer.id, regions=None, topics=None))
     db.commit()
+
+
+def seed_vpn_settings(db: Session) -> None:
+    ensure_vpn_settings(db)

centers/analytics/api/app/services/job_stale.py

@@ -0,0 +1,36 @@
+"""Stale parse-job recovery helpers (running/queued left behind after worker crash)."""
+
+from __future__ import annotations
+
+import os
+from datetime import datetime, timezone
+
+from ..models import ParseJob
+
+# Jobs stuck in running/queued longer than this are considered abandoned.
+STALE_JOB_SECONDS = int(os.getenv("STALE_JOB_SECONDS", "900"))
+
+
+def _aware(dt: datetime | None) -> datetime | None:
+    if dt is None:
+        return None
+    if dt.tzinfo is None:
+        return dt.replace(tzinfo=timezone.utc)
+    return dt
+
+
+def job_anchor_time(job: ParseJob) -> datetime | None:
+    """Best available timestamp for staleness (prefer last_run_at)."""
+    return _aware(job.last_run_at) or _aware(getattr(job, "created_at", None))
+
+
+def is_stale_job(job: ParseJob, now: datetime | None = None, *, ttl: int | None = None) -> bool:
+    if job.status not in ("running", "queued"):
+        return False
+    now = now or datetime.now(timezone.utc)
+    anchor = job_anchor_time(job)
+    if anchor is None:
+        # No timestamp — treat long-lived running as stale immediately for recovery
+        return job.status == "running"
+    limit = ttl if ttl is not None else STALE_JOB_SECONDS
+    return (now - anchor).total_seconds() >= limit

centers/analytics/api/app/services/jobs.py

@@ -4,6 +4,7 @@ import sys
 from pathlib import Path
 
 import redis
+from sqlalchemy.orm import Session
 
 # Allow importing shared contracts from monorepo root in local runs
 _HERE = Path(__file__).resolve()
@@ -14,7 +15,9 @@ for _candidate in (_HERE.parent, *_HERE.parents):
         break
 
 from contracts.queues import queue_key_for_source  # noqa: E402
-from contracts.sources import parse_source_config  # noqa: E402
+from contracts.sources import TelegramSourceConfig, parse_source_config  # noqa: E402
+
+from ..models import ParseChannel, ParseJob, ParserProfile  # noqa: E402
 
 REDIS_URL = os.getenv("REDIS_URL", "redis://redis:6379/0")
 
@@ -23,6 +26,66 @@ def get_redis() -> redis.Redis:
     return redis.from_url(REDIS_URL, decode_responses=True)
 
 
+def flatten_pair_config(
+    channel: ParseChannel,
+    profile: ParserProfile,
+    *,
+    limit: int = 100,
+) -> dict:
+    """Expand Profile + Channel into Redis/CP source_config."""
+    kind = (profile.kind or "heuristic").strip().lower()
+    if kind == "llm":
+        from contracts.llm_profile import LlmProfile
+
+        if not profile.llm_profile:
+            raise ValueError("ParserProfile.llm_profile is empty")
+        llm = LlmProfile.model_validate(profile.llm_profile)
+        cfg = TelegramSourceConfig(
+            channel=channel.channel.strip(),
+            limit=limit,
+            extract_mode="llm",
+            extract_schema=llm.extract_schema,
+            instruction=llm.instruction,
+            required_fields=list(llm.required_fields),
+            multi_event=bool(llm.multi_event),
+            sample_post=profile.sample_post or None,
+        )
+        return cfg.model_dump()
+
+    if not profile.heuristic_profile:
+        raise ValueError("ParserProfile.heuristic_profile is empty")
+    cfg = TelegramSourceConfig(
+        channel=channel.channel.strip(),
+        limit=limit,
+        extract_mode="profile",
+        heuristic_profile=profile.heuristic_profile,
+        sample_post=profile.sample_post or None,
+    )
+    return cfg.model_dump()
+
+
+def resolve_job_source_config(db: Session, job: ParseJob) -> dict:
+    """
+    For pair jobs, rebuild flat source_config from current Profile/Channel.
+    Legacy jobs keep stored source_config.
+    """
+    if job.profile_id and job.channel_id:
+        profile = db.query(ParserProfile).filter(ParserProfile.id == job.profile_id).first()
+        channel = db.query(ParseChannel).filter(ParseChannel.id == job.channel_id).first()
+        if not profile:
+            raise ValueError(f"ParserProfile {job.profile_id} not found")
+        if not channel:
+            raise ValueError(f"ParseChannel {job.channel_id} not found")
+        existing = job.source_config or {}
+        limit = existing.get("limit", 100)
+        if not isinstance(limit, int) or limit < 1:
+            limit = 100
+        flattened = flatten_pair_config(channel, profile, limit=limit)
+        job.source_config = flattened
+        return flattened
+    return dict(job.source_config or {})
+
+
 def enqueue_job(job_id: int, source_type: str, source_config: dict) -> None:
     # Validate known configs early; unknown types still raise from queue_key_for_source
     try:
@@ -38,3 +101,10 @@ def enqueue_job(job_id: int, source_type: str, source_config: dict) -> None:
     }
     key = queue_key_for_source(source_type)
     get_redis().rpush(key, json.dumps(payload))
+
+
+def enqueue_parse_job(db: Session, job: ParseJob) -> None:
+    """Resolve (flatten if pair) then push to Redis."""
+    source_config = resolve_job_source_config(db, job)
+    db.commit()
+    enqueue_job(job.id, job.source_type, source_config)

centers/analytics/api/app/services/migrations.py

@@ -5,20 +5,32 @@ from sqlalchemy.engine import Engine
 def migrate_schema(engine: Engine) -> None:
     """Apply lightweight schema updates for existing deployments."""
     inspector = inspect(engine)
-    if "parse_jobs" not in inspector.get_table_names():
-        return
-
-    columns = {col["name"] for col in inspector.get_columns("parse_jobs")}
+    tables = set(inspector.get_table_names())
     statements: list[str] = []
 
-    if "interval_seconds" not in columns:
-        statements.append(
-            "ALTER TABLE parse_jobs ADD COLUMN interval_seconds INTEGER NOT NULL DEFAULT 3600"
-        )
-    if "is_active" not in columns:
-        statements.append(
-            "ALTER TABLE parse_jobs ADD COLUMN is_active BOOLEAN NOT NULL DEFAULT TRUE"
-        )
+    if "parse_jobs" in tables:
+        columns = {col["name"] for col in inspector.get_columns("parse_jobs")}
+        if "interval_seconds" not in columns:
+            statements.append(
+                "ALTER TABLE parse_jobs ADD COLUMN interval_seconds INTEGER NOT NULL DEFAULT 3600"
+            )
+        if "is_active" not in columns:
+            statements.append(
+                "ALTER TABLE parse_jobs ADD COLUMN is_active BOOLEAN NOT NULL DEFAULT TRUE"
+            )
+        if "profile_id" not in columns:
+            statements.append("ALTER TABLE parse_jobs ADD COLUMN profile_id INTEGER")
+        if "channel_id" not in columns:
+            statements.append("ALTER TABLE parse_jobs ADD COLUMN channel_id INTEGER")
+
+    if "parser_profiles" in tables:
+        columns = {col["name"] for col in inspector.get_columns("parser_profiles")}
+        if "kind" not in columns:
+            statements.append(
+                "ALTER TABLE parser_profiles ADD COLUMN kind VARCHAR(50) NOT NULL DEFAULT 'heuristic'"
+            )
+        if "llm_profile" not in columns:
+            statements.append("ALTER TABLE parser_profiles ADD COLUMN llm_profile JSON")
 
     if not statements:
         return

centers/analytics/api/app/services/parser_builder.py

@@ -0,0 +1,310 @@
+"""Parser builder: one-shot DeepSeek profile generation + static preview."""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import re
+from typing import Any
+
+import httpx
+
+from contracts.heuristic_profile import (
+    HeuristicProfile,
+    TARGET_FIELDS,
+    apply_profile,
+    match_profile,
+    target_field_specs,
+)
+
+logger = logging.getLogger("ca.parser_builder")
+
+GENERATE_SYSTEM = (
+    "You design a static heuristic parser profile for structured Telegram posts. "
+    "Output valid JSON only, no markdown, no Python code. "
+    "The profile is applied with regex/line/marker rules at runtime — never with an LLM. "
+    "Every rule must actually extract a non-empty value from the given sample when applied."
+)
+
+
+def deepseek_enabled() -> bool:
+    return bool(os.getenv("DEEPSEEK_API_KEY", "").strip())
+
+
+def deepseek_settings() -> dict[str, str]:
+    return {
+        "api_key": os.getenv("DEEPSEEK_API_KEY", "").strip(),
+        "base_url": os.getenv("DEEPSEEK_BASE_URL", "https://api.deepseek.com").rstrip("/"),
+        "model": os.getenv("DEEPSEEK_MODEL", "deepseek-chat"),
+    }
+
+
+def get_target_fields() -> list[dict[str, str]]:
+    return target_field_specs()
+
+
+def preview_with_profile(sample_post: str, profile: dict[str, Any] | HeuristicProfile) -> dict[str, str]:
+    return apply_profile(sample_post, profile)
+
+
+def match_preview(
+    sample_post: str,
+    profile: dict[str, Any] | HeuristicProfile,
+) -> tuple[dict[str, str], bool, list[str]]:
+    matched, fields, missing = match_profile(sample_post, profile)
+    return fields, matched, missing
+
+
+def empty_preview_fields(preview: dict[str, str]) -> list[str]:
+    return [name for name, value in preview.items() if not (value or "").strip()]
+
+
+def _extract_json_object(content: str) -> dict[str, Any]:
+    content = (content or "").strip()
+    if content.startswith("```"):
+        content = re.sub(r"^```(?:json)?\s*", "", content)
+        content = re.sub(r"\s*```$", "", content)
+    try:
+        parsed = json.loads(content)
+    except json.JSONDecodeError:
+        match = re.search(r"\{[\s\S]*\}", content)
+        if not match:
+            raise
+        parsed = json.loads(match.group(0))
+    if not isinstance(parsed, dict):
+        raise ValueError("LLM response must be a JSON object")
+    return parsed
+
+
+def _base_user_prompt(sample: str) -> str:
+    fields_help = "\n".join(
+        f"- {spec['name']}: {spec['description']}" for spec in target_field_specs()
+    )
+    return (
+        "Build a HeuristicProfile JSON for this sample Telegram post.\n\n"
+        "Schema:\n"
+        '{"version": 1, "notes": "...", "fields": {'
+        '"<field>": {"strategy": "regex|line|after_marker|between|full_text|literal", '
+        '"pattern": "...", "group": 1, "line_index": 0, "marker": "...", '
+        '"end_marker": "...", "value": "...", "flags": "im", "strip": true}'
+        "}}\n\n"
+        "Rules:\n"
+        "- Only include fields you can extract reliably from the sample.\n"
+        f"- Allowed field names: {', '.join(TARGET_FIELDS)}.\n"
+        "- Prefer regex / line / after_marker / between over literal.\n"
+        "- Use literal only for constant topic/region tags.\n"
+        "- title: usually the first non-empty line (strategy line, line_index 0).\n"
+        "- description: the multi-line body AFTER the title and BEFORE coords/hashtags. "
+        "Prefer strategy between or regex with flags including \"s\" (DOTALL). "
+        "Do NOT leave description empty if the sample has a body paragraph.\n"
+        "- locality: settlement name from the title line when present.\n"
+        "- coords should capture 'lat, lon' when present.\n"
+        "- event_date should capture DD.MM.YYYY / DD.MM.YY or similar.\n"
+        "- topic/region: hashtags like #ru → topic \"ru\"; avoid stuffing unrelated tags into region.\n"
+        "- Do not invent Python code; only declarative rules.\n"
+        "- Mentally apply each rule to the sample: every included field must yield a non-empty string.\n\n"
+        f"Target fields:\n{fields_help}\n\n"
+        f"Sample post:\n{sample[:12000]}"
+    )
+
+
+def _parse_profile_response(content: str) -> HeuristicProfile:
+    raw = _extract_json_object(content)
+    if "fields" not in raw and isinstance(raw.get("profile"), dict):
+        raw = raw["profile"]
+    if "version" not in raw:
+        raw["version"] = 1
+    return HeuristicProfile.model_validate(raw)
+
+
+async def generate_profile(
+    sample_post: str,
+    *,
+    hint: str | None = None,
+    current_profile: dict[str, Any] | HeuristicProfile | None = None,
+) -> HeuristicProfile:
+    settings = deepseek_settings()
+    if not settings["api_key"]:
+        raise RuntimeError(
+            "DEEPSEEK_API_KEY is not set on ca-api. Add it to .env for parser generation."
+        )
+
+    sample = (sample_post or "").strip()
+    if not sample:
+        raise ValueError("sample_post is required")
+
+    hint_text = (hint or "").strip()
+    current: HeuristicProfile | None = None
+    if current_profile is not None:
+        current = (
+            current_profile
+            if isinstance(current_profile, HeuristicProfile)
+            else HeuristicProfile.model_validate(current_profile)
+        )
+
+    if current is not None or hint_text:
+        preview = preview_with_profile(sample, current) if current is not None else {}
+        empty = empty_preview_fields(preview) if preview else []
+        parts = [
+            "Revise the HeuristicProfile JSON for this sample Telegram post.",
+            "Keep strategies that already work; fix only what the manager asks for "
+            "and any fields that preview as empty.",
+            "Output the full updated profile JSON (same schema), not a patch.",
+            "",
+            _base_user_prompt(sample),
+        ]
+        if current is not None:
+            parts.extend(
+                [
+                    "",
+                    "Current profile JSON:",
+                    json.dumps(current.model_dump(), ensure_ascii=False, indent=2)[:12000],
+                ]
+            )
+            if empty:
+                parts.append(
+                    "Fields currently empty in preview (must become non-empty if present in sample): "
+                    + ", ".join(empty)
+                )
+            else:
+                parts.append("Current preview fields: " + json.dumps(preview, ensure_ascii=False))
+        if hint_text:
+            parts.extend(["", "Manager hint (follow this):", hint_text[:4000]])
+        user_prompt = "\n".join(parts)
+    else:
+        user_prompt = _base_user_prompt(sample)
+
+    payload = {
+        "model": settings["model"],
+        "messages": [
+            {"role": "system", "content": GENERATE_SYSTEM},
+            {"role": "user", "content": user_prompt},
+        ],
+        "temperature": 0.1,
+        "response_format": {"type": "json_object"},
+    }
+
+    url = f"{settings['base_url']}/chat/completions"
+    async with httpx.AsyncClient(timeout=90.0) as client:
+        response = await client.post(
+            url,
+            headers={
+                "Authorization": f"Bearer {settings['api_key']}",
+                "Content-Type": "application/json",
+            },
+            json=payload,
+        )
+        response.raise_for_status()
+        data = response.json()
+
+    content = data["choices"][0]["message"]["content"]
+    return _parse_profile_response(content)
+
+
+async def preview_llm_extract(
+    sample_post: str,
+    llm_profile: dict[str, Any],
+) -> tuple[list[dict[str, str]], dict[str, str], bool, list[str], bool, int]:
+    """One-shot DeepSeek extract for CA preview (does not persist).
+
+    Returns (events, fields, matched, missing_required, is_event, matched_count).
+    ``fields`` is the first event (or empty) for legacy UI compatibility.
+    """
+    from contracts.llm_profile import DEFAULT_INSTRUCTION, LlmProfile, match_llm_required
+
+    settings = deepseek_settings()
+    if not settings["api_key"]:
+        raise RuntimeError(
+            "DEEPSEEK_API_KEY is not set on ca-api. Add it to .env for LLM preview."
+        )
+
+    sample = (sample_post or "").strip()
+    if not sample:
+        raise ValueError("sample_post is required")
+
+    profile = LlmProfile.model_validate(llm_profile)
+    schema = profile.extract_schema
+    instr = profile.instruction or DEFAULT_INSTRUCTION
+    schema_lines = "\n".join(f"- {k}: {v}" for k, v in schema.items())
+    multi = bool(profile.multi_event)
+
+    if multi:
+        user_prompt = (
+            f"{instr}\n\n"
+            "If the text describes multiple distinct events (different places, "
+            "coords, or dates), return one object per event in \"events\".\n"
+            f"Fields per event:\n{schema_lines}\n\n"
+            'Return JSON: {"is_event": true|false, "events": [{<field>: <string>}, ...]}\n'
+            "If there is no event, return is_event=false and events=[].\n\n"
+            f"Text:\n{sample[:12000]}"
+        )
+    else:
+        user_prompt = (
+            f"{instr}\n\n"
+            f"Fields to extract:\n{schema_lines}\n\n"
+            'Return JSON: {"is_event": true|false, "fields": {<field>: <string>}}\n\n'
+            f"Text:\n{sample[:12000]}"
+        )
+
+    payload = {
+        "model": settings["model"],
+        "messages": [
+            {
+                "role": "system",
+                "content": (
+                    "You extract structured event data for a geoint map. "
+                    "Output valid JSON only, no markdown."
+                ),
+            },
+            {"role": "user", "content": user_prompt},
+        ],
+        "temperature": 0.1,
+        "response_format": {"type": "json_object"},
+    }
+
+    url = f"{settings['base_url']}/chat/completions"
+    async with httpx.AsyncClient(timeout=90.0) as client:
+        response = await client.post(
+            url,
+            headers={
+                "Authorization": f"Bearer {settings['api_key']}",
+                "Content-Type": "application/json",
+            },
+            json=payload,
+        )
+        response.raise_for_status()
+        data = response.json()
+
+    content = data["choices"][0]["message"]["content"]
+    parsed = _extract_json_object(content)
+    is_event = bool(parsed.get("is_event", True))
+    empty_fields = {key: "" for key in schema}
+
+    if not is_event:
+        return [], empty_fields, False, [], False, 0
+
+    events: list[dict[str, str]] = []
+    if multi and isinstance(parsed.get("events"), list):
+        for item in parsed["events"]:
+            if isinstance(item, dict):
+                events.append({key: str(item.get(key) or "").strip() for key in schema})
+    else:
+        fields_raw = parsed.get("fields") if isinstance(parsed.get("fields"), dict) else parsed
+        if not isinstance(fields_raw, dict):
+            fields_raw = {}
+        events.append({key: str(fields_raw.get(key) or "").strip() for key in schema})
+
+    if not events:
+        return [], empty_fields, False, [], False, 0
+
+    matched_events = [ev for ev in events if match_llm_required(ev, list(profile.required_fields))]
+    fields = events[0]
+    missing = [
+        name
+        for name in profile.required_fields
+        if not str(fields.get(name) or "").strip()
+    ]
+    matched = match_llm_required(fields, list(profile.required_fields))
+    return events, fields, matched, missing, True, len(matched_events)
+

centers/analytics/api/app/services/scheduler.py

@@ -5,7 +5,8 @@ from datetime import datetime, timezone
 
 from ..database import SessionLocal
 from ..models import ParseJob
-from .jobs import enqueue_job
+from .job_stale import STALE_JOB_SECONDS, is_stale_job
+from .jobs import enqueue_parse_job
 
 logger = logging.getLogger(__name__)
 
@@ -13,10 +14,47 @@ TICK_SECONDS = 30
 RECURRING_STATUSES = ("completed", "failed")
 
 
+def recover_stale_jobs(db, now: datetime) -> None:
+    """Mark abandoned running/queued jobs as failed; re-queue if still active."""
+    stuck = (
+        db.query(ParseJob)
+        .filter(ParseJob.status.in_(("running", "queued")))
+        .all()
+    )
+    for job in stuck:
+        if not is_stale_job(job, now):
+            continue
+        prev = job.status
+        job.status = "failed"
+        job.last_error = (
+            f"Stale {prev} recovered after {STALE_JOB_SECONDS}s "
+            "(worker likely restarted)"
+        )
+        job.last_run_at = now
+        db.commit()
+        logger.warning("Recovered stale job %s (was %s)", job.id, prev)
+
+        if not job.is_active or job.interval_seconds <= 0:
+            continue
+        job.status = "queued"
+        job.last_error = None
+        db.commit()
+        try:
+            enqueue_parse_job(db, job)
+            logger.info("Re-queued recovered job %s", job.id)
+        except ValueError as exc:
+            job.status = "failed"
+            job.last_error = str(exc)
+            db.commit()
+            logger.warning("Skip re-queue recovered job %s: %s", job.id, exc)
+
+
 def run_scheduler_tick() -> None:
     db = SessionLocal()
     try:
         now = datetime.now(timezone.utc)
+        recover_stale_jobs(db, now)
+
         jobs = (
             db.query(ParseJob)
             .filter(
@@ -36,7 +74,14 @@ def run_scheduler_tick() -> None:
             job.status = "queued"
             job.last_error = None
             db.commit()
-            enqueue_job(job.id, job.source_type, job.source_config)
+            try:
+                enqueue_parse_job(db, job)
+            except ValueError as exc:
+                job.status = "failed"
+                job.last_error = str(exc)
+                db.commit()
+                logger.warning("Skip re-queue job %s: %s", job.id, exc)
+                continue
             logger.info("Re-queued recurring job %s (interval %ss)", job.id, job.interval_seconds)
     except Exception:
         logger.exception("Scheduler tick failed")
@@ -54,5 +99,9 @@ def start_scheduler() -> threading.Event:
     stop_event = threading.Event()
     thread = threading.Thread(target=_scheduler_loop, args=(stop_event,), daemon=True)
     thread.start()
-    logger.info("Parse job scheduler started (tick every %ss)", TICK_SECONDS)
+    logger.info(
+        "Parse job scheduler started (tick every %ss, stale after %ss)",
+        TICK_SECONDS,
+        STALE_JOB_SECONDS,
+    )
     return stop_event

centers/analytics/api/app/services/vpn.py

@@ -0,0 +1,137 @@
+"""VPN settings persistence helpers."""
+
+from __future__ import annotations
+
+from contracts.queues import known_source_types
+from contracts.vpn import VpnSettings
+from sqlalchemy.orm import Session
+
+from ..models import VpnSettingsRow
+from ..schemas import VpnSettingsAdminRead, VpnSettingsInternalRead, VpnSettingsUpdate
+
+SINGLETON_ID = 1
+
+
+def _hint_secret(value: str | None) -> str | None:
+    if not value:
+        return None
+    if len(value) <= 4:
+        return "****"
+    return f"…{value[-4:]}"
+
+
+def ensure_vpn_settings(db: Session) -> VpnSettingsRow:
+    row = db.query(VpnSettingsRow).filter(VpnSettingsRow.id == SINGLETON_ID).first()
+    if row is not None:
+        return row
+    row = VpnSettingsRow(
+        id=SINGLETON_ID,
+        enabled=False,
+        mode="subscription",
+        subscription_interval_seconds=3600,
+        proxied_source_types=[],
+    )
+    db.add(row)
+    db.commit()
+    db.refresh(row)
+    return row
+
+
+def row_to_contract(row: VpnSettingsRow) -> VpnSettings:
+    return VpnSettings.model_validate(
+        {
+            "enabled": row.enabled,
+            "mode": row.mode,
+            "subscription_url": row.subscription_url,
+            "subscription_interval_seconds": row.subscription_interval_seconds,
+            "host": row.host,
+            "port": row.port,
+            "username": row.username,
+            "password": row.password,
+            "proxied_source_types": row.proxied_source_types or [],
+        }
+    )
+
+
+def to_admin_read(row: VpnSettingsRow) -> VpnSettingsAdminRead:
+    return VpnSettingsAdminRead(
+        enabled=row.enabled,
+        mode=row.mode,  # type: ignore[arg-type]
+        subscription_url_set=bool(row.subscription_url),
+        subscription_url_hint=_hint_secret(row.subscription_url),
+        subscription_interval_seconds=row.subscription_interval_seconds,
+        host=row.host,
+        port=row.port,
+        username=row.username,
+        password_set=bool(row.password),
+        proxied_source_types=list(row.proxied_source_types or []),
+        available_source_types=known_source_types(),
+    )
+
+
+def to_internal_read(row: VpnSettingsRow) -> VpnSettingsInternalRead:
+    return VpnSettingsInternalRead(
+        enabled=row.enabled,
+        mode=row.mode,  # type: ignore[arg-type]
+        subscription_url=row.subscription_url,
+        subscription_interval_seconds=row.subscription_interval_seconds,
+        host=row.host,
+        port=row.port,
+        username=row.username,
+        password=row.password,
+        proxied_source_types=list(row.proxied_source_types or []),
+    )
+
+
+def apply_vpn_update(db: Session, payload: VpnSettingsUpdate) -> VpnSettingsRow:
+    row = ensure_vpn_settings(db)
+    data = {
+        "enabled": row.enabled if payload.enabled is None else payload.enabled,
+        "mode": row.mode if payload.mode is None else payload.mode,
+        "subscription_url": row.subscription_url,
+        "subscription_interval_seconds": (
+            row.subscription_interval_seconds
+            if payload.subscription_interval_seconds is None
+            else payload.subscription_interval_seconds
+        ),
+        "host": row.host if payload.host is None else (payload.host.strip() or None),
+        "port": row.port if payload.port is None else payload.port,
+        "username": (
+            row.username
+            if payload.username is None
+            else (payload.username.strip() or None)
+        ),
+        "password": row.password,
+        "proxied_source_types": (
+            list(row.proxied_source_types or [])
+            if payload.proxied_source_types is None
+            else payload.proxied_source_types
+        ),
+    }
+
+    if payload.clear_subscription_url:
+        data["subscription_url"] = None
+    elif payload.subscription_url is not None:
+        stripped = payload.subscription_url.strip()
+        data["subscription_url"] = stripped or None
+
+    if payload.clear_password:
+        data["password"] = None
+    elif payload.password is not None:
+        stripped = payload.password.strip()
+        data["password"] = stripped or None
+
+    validated = VpnSettings.model_validate(data)
+
+    row.enabled = validated.enabled
+    row.mode = validated.mode
+    row.subscription_url = validated.subscription_url
+    row.subscription_interval_seconds = validated.subscription_interval_seconds
+    row.host = validated.host
+    row.port = validated.port
+    row.username = validated.username
+    row.password = validated.password
+    row.proxied_source_types = validated.proxied_source_types
+    db.commit()
+    db.refresh(row)
+    return row

centers/analytics/api/requirements.txt

@@ -5,3 +5,5 @@ pydantic==2.10.3
 python-multipart==0.0.20
 psycopg2-binary==2.9.10
 redis==5.2.1
+httpx==0.28.1
+PyJWT==2.10.1

centers/analytics/frontend/src/api/admin.ts

@@ -4,14 +4,18 @@ import type {
   Consumer,
   ConsumerCreate,
   ConsumerUpdate,
+  EventCreate,
   EventFilters,
   EventListResponse,
   EventRecord,
+  EventUpdate,
   ParseJob,
   ParseJobCreate,
   ParseJobUpdate,
   TimelinePoint,
   TopItem,
+  VpnSettings,
+  VpnSettingsUpdate,
 } from "../types/admin";
 
 function buildQuery(params: Record<string, string | number | undefined>): string {
@@ -69,6 +73,26 @@ export function fetchEvents(filters: EventFilters = {}): Promise<EventListRespon
   );
 }
 
+export function createEvent(payload: EventCreate): Promise<EventRecord> {
+  return request<EventRecord>(
+    "/events",
+    { method: "POST", body: JSON.stringify(payload) },
+    ADMIN_BASE,
+  );
+}
+
+export function updateEvent(id: number, payload: EventUpdate): Promise<EventRecord> {
+  return request<EventRecord>(
+    `/events/${id}`,
+    { method: "PATCH", body: JSON.stringify(payload) },
+    ADMIN_BASE,
+  );
+}
+
+export function deleteEvent(id: number): Promise<void> {
+  return request<void>(`/events/${id}`, { method: "DELETE" }, ADMIN_BASE);
+}
+
 export function fetchConsumers(): Promise<Consumer[]> {
   return request<Consumer[]>("/consumers", undefined, ADMIN_BASE);
 }
@@ -135,3 +159,15 @@ export async function testDistribution(apiKey: string, limit = 5): Promise<Event
   }
   return response.json() as Promise<EventRecord[]>;
 }
+
+export function fetchVpnSettings(): Promise<VpnSettings> {
+  return request<VpnSettings>("/vpn", undefined, ADMIN_BASE);
+}
+
+export function updateVpnSettings(payload: VpnSettingsUpdate): Promise<VpnSettings> {
+  return request<VpnSettings>(
+    "/vpn",
+    { method: "PUT", body: JSON.stringify(payload) },
+    ADMIN_BASE,
+  );
+}

centers/analytics/frontend/src/api/client.ts

@@ -1,3 +1,5 @@
+import { clearToken, getToken } from "../auth";
+
 const API_BASE = "/api";
 const ADMIN_BASE = "/admin";
 
@@ -13,11 +15,24 @@ export async function request<T>(
     headers.set("Content-Type", "application/json");
   }
 
+  const token = getToken();
+  if (token && !headers.has("Authorization")) {
+    headers.set("Authorization", `Bearer ${token}`);
+  }
+
   const response = await fetch(`${base}${url}`, {
     ...options,
     headers,
   });
 
+  if (response.status === 401 && token) {
+    clearToken();
+    if (typeof window !== "undefined" && !window.location.pathname.startsWith("/login")) {
+      const next = `${window.location.pathname}${window.location.search}`;
+      window.location.assign(`/login?next=${encodeURIComponent(next)}`);
+    }
+  }
+
   if (!response.ok) {
     const message = await response.text();
     throw new Error(message || `Ошибка запроса: ${response.status}`);

centers/analytics/frontend/src/api/entities.ts

@@ -0,0 +1,133 @@
+import { ADMIN_BASE, request } from "./client";
+import type {
+  HeuristicProfile,
+  LlmProfile,
+  ParseChannel,
+  ParseChannelCreate,
+  ParseChannelUpdate,
+  ParserProfile,
+  ParserProfileCreate,
+  ParserProfileUpdate,
+  TargetField,
+} from "../types/admin";
+
+export type { HeuristicProfile, LlmProfile, TargetField } from "../types/admin";
+
+export function fetchChannels(): Promise<ParseChannel[]> {
+  return request<ParseChannel[]>("/parse-channels", undefined, ADMIN_BASE);
+}
+
+export function createChannel(payload: ParseChannelCreate): Promise<ParseChannel> {
+  return request<ParseChannel>(
+    "/parse-channels",
+    { method: "POST", body: JSON.stringify(payload) },
+    ADMIN_BASE,
+  );
+}
+
+export function updateChannel(id: number, payload: ParseChannelUpdate): Promise<ParseChannel> {
+  return request<ParseChannel>(
+    `/parse-channels/${id}`,
+    { method: "PATCH", body: JSON.stringify(payload) },
+    ADMIN_BASE,
+  );
+}
+
+export function deleteChannel(id: number): Promise<void> {
+  return request<void>(`/parse-channels/${id}`, { method: "DELETE" }, ADMIN_BASE);
+}
+
+export function fetchProfiles(): Promise<ParserProfile[]> {
+  return request<ParserProfile[]>("/parser-profiles", undefined, ADMIN_BASE);
+}
+
+export function createProfile(payload: ParserProfileCreate): Promise<ParserProfile> {
+  return request<ParserProfile>(
+    "/parser-profiles",
+    { method: "POST", body: JSON.stringify(payload) },
+    ADMIN_BASE,
+  );
+}
+
+export function updateProfile(id: number, payload: ParserProfileUpdate): Promise<ParserProfile> {
+  return request<ParserProfile>(
+    `/parser-profiles/${id}`,
+    { method: "PATCH", body: JSON.stringify(payload) },
+    ADMIN_BASE,
+  );
+}
+
+export function deleteProfile(id: number): Promise<void> {
+  return request<void>(`/parser-profiles/${id}`, { method: "DELETE" }, ADMIN_BASE);
+}
+
+export function fetchTargetFields(): Promise<{ fields: TargetField[] }> {
+  return request<{ fields: TargetField[] }>("/parser-profiles/target-fields", undefined, ADMIN_BASE);
+}
+
+export function fetchLlmDefaults(): Promise<LlmProfile> {
+  return request<LlmProfile>("/parser-profiles/llm-defaults", undefined, ADMIN_BASE);
+}
+
+export function generateParserProfile(payload: {
+  sample_post: string;
+  hint?: string;
+  current_profile?: HeuristicProfile | Record<string, unknown> | null;
+}): Promise<{
+  profile: HeuristicProfile;
+  preview: Record<string, string>;
+  empty_fields: string[];
+}> {
+  return request<{
+    profile: HeuristicProfile;
+    preview: Record<string, string>;
+    empty_fields: string[];
+  }>(
+    "/parser-profiles/generate",
+    {
+      method: "POST",
+      body: JSON.stringify({
+        sample_post: payload.sample_post,
+        hint: payload.hint || undefined,
+        current_profile: payload.current_profile || undefined,
+      }),
+    },
+    ADMIN_BASE,
+  );
+}
+
+export function previewParserProfile(
+  sample_post: string,
+  heuristic_profile: HeuristicProfile | Record<string, unknown>,
+): Promise<{ fields: Record<string, string>; matched: boolean; missing_required: string[] }> {
+  return request<{ fields: Record<string, string>; matched: boolean; missing_required: string[] }>(
+    "/parser-profiles/preview",
+    { method: "POST", body: JSON.stringify({ sample_post, heuristic_profile }) },
+    ADMIN_BASE,
+  );
+}
+
+export function previewLlmProfile(
+  sample_post: string,
+  llm_profile: LlmProfile | Record<string, unknown>,
+): Promise<{
+  fields: Record<string, string>;
+  events: Record<string, string>[];
+  matched: boolean;
+  missing_required: string[];
+  is_event: boolean;
+  matched_count: number;
+}> {
+  return request<{
+    fields: Record<string, string>;
+    events: Record<string, string>[];
+    matched: boolean;
+    missing_required: string[];
+    is_event: boolean;
+    matched_count: number;
+  }>(
+    "/parser-profiles/preview-llm",
+    { method: "POST", body: JSON.stringify({ sample_post, llm_profile }) },
+    ADMIN_BASE,
+  );
+}

centers/analytics/frontend/src/auth.ts

@@ -0,0 +1,40 @@
+import { computed, ref } from "vue";
+
+const TOKEN_KEY = "mapmil_admin_token";
+const ADMIN_BASE = "/admin";
+
+const token = ref<string | null>(localStorage.getItem(TOKEN_KEY));
+
+export function getToken(): string | null {
+  return token.value;
+}
+
+export function setToken(value: string | null): void {
+  token.value = value;
+  if (value) localStorage.setItem(TOKEN_KEY, value);
+  else localStorage.removeItem(TOKEN_KEY);
+}
+
+export function clearToken(): void {
+  setToken(null);
+}
+
+export const isAuthenticated = computed(() => Boolean(token.value));
+
+export async function login(username: string, password: string): Promise<void> {
+  const response = await fetch(`${ADMIN_BASE}/auth/login`, {
+    method: "POST",
+    headers: { "Content-Type": "application/json" },
+    body: JSON.stringify({ username, password }),
+  });
+  if (!response.ok) {
+    const message = await response.text();
+    throw new Error(message || `Ошибка входа: ${response.status}`);
+  }
+  const data = (await response.json()) as { access_token: string };
+  setToken(data.access_token);
+}
+
+export function logout(): void {
+  clearToken();
+}

centers/analytics/frontend/src/components/MapView.vue

@@ -4,11 +4,15 @@ import { onMounted, onUnmounted, watch } from "vue";
 import { useLeafletMap } from "../composables/useLeafletMap";
 import type { MapObjectWithEvent } from "../types/map";
 
-const props = defineProps<{
-  objects: MapObjectWithEvent[];
-  selectedId: number | null;
-  openPopupId: number | null;
-}>();
+const props = withDefaults(
+  defineProps<{
+    objects: MapObjectWithEvent[];
+    selectedId: number | null;
+    openPopupId: number | null;
+    editable?: boolean;
+  }>(),
+  { editable: false },
+);
 
 const emit = defineEmits<{
   select: [object: MapObjectWithEvent];
@@ -53,7 +57,7 @@ function buildPopupHtml(obj: MapObjectWithEvent): string {
 }
 
 function isDraggable(obj: MapObjectWithEvent): boolean {
-  return obj.id === props.selectedId && !obj.event_id;
+  return props.editable && obj.id === props.selectedId && !obj.event_id;
 }
 
 function syncMarkers() {
@@ -170,7 +174,7 @@ onUnmounted(() => {
 });
 
 watch(
-  () => [props.objects, props.selectedId] as const,
+  () => [props.objects, props.selectedId, props.editable] as const,
   () => {
     syncMarkers();
     panToSelected(false);

centers/analytics/frontend/src/components/ObjectPanel.vue

@@ -9,9 +9,13 @@ import {
 import { MEDIA_ACCEPT, OBJECT_TYPE_LABELS } from "../types/object";
 import type { MapObjectWithEvent } from "../types/map";
 
-const props = defineProps<{
-  object: MapObjectWithEvent | null;
-}>();
+const props = withDefaults(
+  defineProps<{
+    object: MapObjectWithEvent | null;
+    canEdit?: boolean;
+  }>(),
+  { canEdit: false },
+);
 
 const emit = defineEmits<{
   openEvents: [];
@@ -151,7 +155,7 @@ watch(() => props.object?.id, () => loadMedia(), { immediate: true });
       <section v-if="!isIngested" class="media-section">
         <div class="media-header">
           <h4>Медиа</h4>
-          <label class="upload-btn">
+          <label v-if="canEdit" class="upload-btn">
             {{ uploading ? "Загрузка..." : "Добавить" }}
             <input
               type="file"
@@ -185,7 +189,14 @@ watch(() => props.object?.id, () => loadMedia(), { immediate: true });
               <span class="name" :title="item.original_name">{{ item.original_name }}</span>
               <span class="size">{{ formatFileSize(item.size) }}</span>
             </div>
-            <button type="button" class="delete-btn" @click="handleDelete(item.id)">Удалить</button>
+            <button
+              v-if="canEdit"
+              type="button"
+              class="delete-btn"
+              @click="handleDelete(item.id)"
+            >
+              Удалить
+            </button>
           </li>
         </ul>
       </section>

centers/analytics/frontend/src/composables/useAuth.ts

@@ -0,0 +1,41 @@
+import { computed, ref } from "vue";
+import { useRouter } from "vue-router";
+
+import { clearToken, getToken, isAuthenticated, login as doLogin, logout as doLogout } from "../auth";
+
+export function useAuth() {
+  const router = useRouter();
+  const busy = ref(false);
+  const error = ref("");
+
+  const authenticated = computed(() => isAuthenticated.value);
+
+  async function login(username: string, password: string): Promise<boolean> {
+    busy.value = true;
+    error.value = "";
+    try {
+      await doLogin(username, password);
+      return true;
+    } catch (err) {
+      clearToken();
+      error.value = err instanceof Error ? err.message : "Не удалось войти";
+      return false;
+    } finally {
+      busy.value = false;
+    }
+  }
+
+  function logout(): void {
+    doLogout();
+    void router.push({ name: "map" });
+  }
+
+  return {
+    authenticated,
+    busy,
+    error,
+    getToken,
+    login,
+    logout,
+  };
+}

centers/analytics/frontend/src/layouts/AppLayout.vue

@@ -2,23 +2,34 @@
 import { computed } from "vue";
 import { useRoute } from "vue-router";
 
-const route = useRoute();
+import { useAuth } from "../composables/useAuth";
 
-const navItems = [
-  { to: "/", label: "Карта", exact: true },
+const route = useRoute();
+const { authenticated, logout } = useAuth();
+
+const publicNavItems = [{ to: "/", label: "Карта", exact: true }];
+
+const adminNavItems = [
+  { to: "/channels", label: "Каналы" },
+  { to: "/parser-profiles", label: "Профили" },
   { to: "/parsers", label: "Парсеры" },
+  { to: "/vpn", label: "VPN" },
   { to: "/events", label: "События" },
   { to: "/analytics", label: "Аналитика" },
   { to: "/consumers", label: "ПИ" },
 ];
 
+const navItems = computed(() =>
+  authenticated.value ? [...publicNavItems, ...adminNavItems] : publicNavItems,
+);
+
 function isActive(path: string, exact = false): boolean {
   if (exact) return route.path === path;
   return route.path.startsWith(path);
 }
 
 const pageTitle = computed(() => {
-  const item = navItems.find((n) => isActive(n.to, n.exact));
+  const item = navItems.value.find((n) => isActive(n.to, n.exact));
   return item?.label ?? "MapMil";
 });
 </script>
@@ -40,6 +51,12 @@ const pageTitle = computed(() => {
         >
           {{ item.label }}
         </router-link>
+        <router-link v-if="!authenticated" to="/login" class="nav-link nav-auth">
+          Вход
+        </router-link>
+        <button v-else type="button" class="nav-link nav-auth nav-button" @click="logout">
+          Выход
+        </button>
       </nav>
     </header>
     <main class="admin-main">
@@ -85,6 +102,7 @@ const pageTitle = computed(() => {
 
 .nav {
   display: flex;
+  align-items: center;
   gap: 0.25rem;
 }
 
@@ -107,6 +125,19 @@ const pageTitle = computed(() => {
   font-weight: 500;
 }
 
+.nav-auth {
+  margin-left: 0.5rem;
+  color: #1565c0;
+  font-weight: 500;
+}
+
+.nav-button {
+  border: none;
+  background: transparent;
+  cursor: pointer;
+  font: inherit;
+}
+
 .admin-main {
   flex: 1;
   min-height: 0;

centers/analytics/frontend/src/router/index.ts

@@ -1,27 +1,85 @@
 import { createRouter, createWebHistory } from "vue-router";
 
+import { getToken } from "../auth";
 import AppLayout from "../layouts/AppLayout.vue";
 import AnalyticsView from "../views/AnalyticsView.vue";
+import ChannelsView from "../views/ChannelsView.vue";
 import ConsumersView from "../views/ConsumersView.vue";
 import EventsView from "../views/EventsView.vue";
+import LoginView from "../views/LoginView.vue";
 import MapViewPage from "../views/MapViewPage.vue";
+import ParserProfilesView from "../views/ParserProfilesView.vue";
 import ParsersView from "../views/ParsersView.vue";
+import VpnView from "../views/VpnView.vue";
 
 const router = createRouter({
   history: createWebHistory(),
   routes: [
+    {
+      path: "/login",
+      name: "login",
+      component: LoginView,
+      meta: { public: true },
+    },
     {
       path: "/",
       component: AppLayout,
       children: [
-        { path: "", name: "map", component: MapViewPage },
-        { path: "parsers", name: "parsers", component: ParsersView },
-        { path: "events", name: "events", component: EventsView },
-        { path: "analytics", name: "analytics", component: AnalyticsView },
-        { path: "consumers", name: "consumers", component: ConsumersView },
+        { path: "", name: "map", component: MapViewPage, meta: { public: true } },
+        {
+          path: "channels",
+          name: "channels",
+          component: ChannelsView,
+          meta: { requiresAuth: true },
+        },
+        {
+          path: "parser-profiles",
+          name: "parser-profiles",
+          component: ParserProfilesView,
+          meta: { requiresAuth: true },
+        },
+        {
+          path: "parsers",
+          name: "parsers",
+          component: ParsersView,
+          meta: { requiresAuth: true },
+        },
+        {
+          path: "vpn",
+          name: "vpn",
+          component: VpnView,
+          meta: { requiresAuth: true },
+        },
+        {
+          path: "events",
+          name: "events",
+          component: EventsView,
+          meta: { requiresAuth: true },
+        },
+        {
+          path: "analytics",
+          name: "analytics",
+          component: AnalyticsView,
+          meta: { requiresAuth: true },
+        },
+        {
+          path: "consumers",
+          name: "consumers",
+          component: ConsumersView,
+          meta: { requiresAuth: true },
+        },
       ],
     },
   ],
 });
 
+router.beforeEach((to) => {
+  if (!to.meta.requiresAuth) return true;
+  if (getToken()) return true;
+  return {
+    name: "login",
+    query: { next: to.fullPath },
+  };
+});
+
 export default router;

centers/analytics/frontend/src/style.css

@@ -125,6 +125,15 @@ textarea {
   font-size: 0.8rem;
 }
 
+.btn-danger {
+  color: #b71c1c;
+  border-color: #ef9a9a;
+}
+
+.btn-danger:hover {
+  background: #ffebee;
+}
+
 .link-btn {
   background: none;
   border: none;

centers/analytics/frontend/src/types/admin.ts

@@ -2,6 +2,8 @@ export interface ParseJob {
   id: number;
   source_type: string;
   source_config: Record<string, unknown>;
+  profile_id?: number | null;
+  channel_id?: number | null;
   schedule: string | null;
   interval_seconds: number;
   is_active: boolean;
@@ -9,11 +11,17 @@ export interface ParseJob {
   last_run_at: string | null;
   last_error: string | null;
   created_at: string;
+  profile_name?: string | null;
+  channel_name?: string | null;
+  channel_handle?: string | null;
 }
 
 export interface ParseJobCreate {
-  source_type: string;
-  source_config: Record<string, unknown>;
+  source_type?: string;
+  source_config?: Record<string, unknown>;
+  profile_id?: number;
+  channel_id?: number;
+  limit?: number;
   schedule?: string | null;
   interval_seconds?: number;
   is_active?: boolean;
@@ -21,10 +29,85 @@ export interface ParseJobCreate {
 
 export interface ParseJobUpdate {
   source_config?: Record<string, unknown>;
+  profile_id?: number;
+  channel_id?: number;
+  limit?: number;
   interval_seconds?: number;
   is_active?: boolean;
 }
 
+export type TargetField = {
+  name: string;
+  type: string;
+  description: string;
+};
+
+export type HeuristicProfile = {
+  version: number;
+  fields: Record<string, Record<string, unknown>>;
+  required_fields?: string[];
+  notes?: string;
+};
+
+export type LlmProfile = {
+  instruction?: string | null;
+  extract_schema: Record<string, string>;
+  required_fields?: string[];
+  multi_event?: boolean;
+};
+
+export interface ParserProfile {
+  id: number;
+  name: string;
+  kind: "heuristic" | "llm";
+  sample_post: string;
+  heuristic_profile: Record<string, unknown> | null;
+  llm_profile: LlmProfile | Record<string, unknown> | null;
+  status: string;
+  created_at: string;
+}
+
+export interface ParserProfileCreate {
+  name: string;
+  kind?: "heuristic" | "llm";
+  sample_post?: string;
+  heuristic_profile?: Record<string, unknown> | null;
+  llm_profile?: LlmProfile | Record<string, unknown> | null;
+  status?: string;
+}
+
+export interface ParserProfileUpdate {
+  name?: string;
+  kind?: "heuristic" | "llm";
+  sample_post?: string;
+  heuristic_profile?: Record<string, unknown> | null;
+  llm_profile?: LlmProfile | Record<string, unknown> | null;
+  status?: string;
+}
+
+export interface ParseChannel {
+  id: number;
+  name: string;
+  source_type: string;
+  channel: string;
+  is_active: boolean;
+  created_at: string;
+}
+
+export interface ParseChannelCreate {
+  name: string;
+  source_type?: string;
+  channel: string;
+  is_active?: boolean;
+}
+
+export interface ParseChannelUpdate {
+  name?: string;
+  source_type?: string;
+  channel?: string;
+  is_active?: boolean;
+}
+
 export interface EventRecord {
   id: number;
   source_type: string;
@@ -43,6 +126,38 @@ export interface EventRecord {
   metadata: Record<string, unknown> | null;
 }
 
+export interface EventCreate {
+  source_type?: string;
+  source_url: string;
+  raw_text?: string;
+  title?: string;
+  description?: string;
+  locality?: string;
+  latitude?: number | null;
+  longitude?: number | null;
+  event_date?: string | null;
+  region?: string | null;
+  topic?: string | null;
+  tags?: string[] | null;
+  metadata?: Record<string, unknown> | null;
+}
+
+export interface EventUpdate {
+  source_type?: string;
+  source_url?: string;
+  raw_text?: string;
+  title?: string;
+  description?: string;
+  locality?: string;
+  latitude?: number | null;
+  longitude?: number | null;
+  event_date?: string | null;
+  region?: string | null;
+  topic?: string | null;
+  tags?: string[] | null;
+  metadata?: Record<string, unknown> | null;
+}
+
 export interface EventListResponse {
   items: EventRecord[];
   total: number;
@@ -105,3 +220,33 @@ export interface TopItem {
   name: string;
   count: number;
 }
+
+export type VpnMode = "subscription" | "socks5" | "http";
+
+export interface VpnSettings {
+  enabled: boolean;
+  mode: VpnMode;
+  subscription_url_set: boolean;
+  subscription_url_hint: string | null;
+  subscription_interval_seconds: number;
+  host: string | null;
+  port: number | null;
+  username: string | null;
+  password_set: boolean;
+  proxied_source_types: string[];
+  available_source_types: string[];
+}
+
+export interface VpnSettingsUpdate {
+  enabled?: boolean;
+  mode?: VpnMode;
+  subscription_url?: string | null;
+  clear_subscription_url?: boolean;
+  subscription_interval_seconds?: number;
+  host?: string | null;
+  port?: number | null;
+  username?: string | null;
+  password?: string | null;
+  clear_password?: boolean;
+  proxied_source_types?: string[];
+}

centers/analytics/frontend/src/views/ChannelsView.vue

@@ -0,0 +1,273 @@
+<script setup lang="ts">
+import { onMounted, ref } from "vue";
+import {
+  createChannel,
+  deleteChannel,
+  fetchChannels,
+  updateChannel,
+} from "../api/entities";
+import type { ParseChannel } from "../types/admin";
+
+const channels = ref<ParseChannel[]>([]);
+const loading = ref(true);
+const error = ref("");
+const submitting = ref(false);
+
+const form = ref({
+  name: "",
+  channel: "",
+  is_active: true,
+});
+
+const editModal = ref<{
+  open: boolean;
+  channel: ParseChannel | null;
+  name: string;
+  handle: string;
+  is_active: boolean;
+}>({
+  open: false,
+  channel: null,
+  name: "",
+  handle: "",
+  is_active: true,
+});
+
+async function loadChannels() {
+  loading.value = true;
+  error.value = "";
+  try {
+    channels.value = await fetchChannels();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось загрузить каналы";
+  } finally {
+    loading.value = false;
+  }
+}
+
+async function handleCreate() {
+  if (!form.value.name.trim() || !form.value.channel.trim()) {
+    error.value = "Укажите название и handle канала";
+    return;
+  }
+  submitting.value = true;
+  error.value = "";
+  try {
+    await createChannel({
+      name: form.value.name.trim(),
+      channel: form.value.channel.trim(),
+      source_type: "telegram",
+      is_active: form.value.is_active,
+    });
+    form.value = { name: "", channel: "", is_active: true };
+    await loadChannels();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось создать канал";
+  } finally {
+    submitting.value = false;
+  }
+}
+
+function openEdit(ch: ParseChannel) {
+  editModal.value = {
+    open: true,
+    channel: ch,
+    name: ch.name,
+    handle: ch.channel,
+    is_active: ch.is_active,
+  };
+}
+
+function closeEdit() {
+  editModal.value.open = false;
+  editModal.value.channel = null;
+}
+
+async function handleSaveEdit() {
+  if (!editModal.value.channel) return;
+  if (!editModal.value.name.trim() || !editModal.value.handle.trim()) {
+    error.value = "Укажите название и handle";
+    return;
+  }
+  submitting.value = true;
+  error.value = "";
+  try {
+    await updateChannel(editModal.value.channel.id, {
+      name: editModal.value.name.trim(),
+      channel: editModal.value.handle.trim(),
+      is_active: editModal.value.is_active,
+    });
+    closeEdit();
+    await loadChannels();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось сохранить";
+  } finally {
+    submitting.value = false;
+  }
+}
+
+async function handleDelete(ch: ParseChannel) {
+  if (!window.confirm(`Удалить канал «${ch.name}» (@${ch.channel})?`)) return;
+  error.value = "";
+  try {
+    await deleteChannel(ch.id);
+    await loadChannels();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось удалить";
+  }
+}
+
+function formatDate(value: string): string {
+  return new Date(value).toLocaleString("ru-RU");
+}
+
+onMounted(loadChannels);
+</script>
+
+<template>
+  <div class="page">
+    <h2 class="page-heading">Каналы</h2>
+    <p class="intro">
+      Каналы для парсинга (Telegram). Связка канал + профиль создаётся на странице
+      <router-link to="/parsers">Парсеры</router-link>.
+    </p>
+
+    <section class="card">
+      <h3>Новый канал</h3>
+      <form class="admin-form" @submit.prevent="handleCreate">
+        <div class="form-row">
+          <label>
+            Название
+            <input v-model="form.name" type="text" placeholder="Сводки фронта" required />
+          </label>
+          <label>
+            Handle / id
+            <input v-model="form.channel" type="text" placeholder="example_channel" required />
+          </label>
+          <label class="checkbox-label">
+            <input v-model="form.is_active" type="checkbox" />
+            Активен
+          </label>
+        </div>
+        <div class="form-actions">
+          <button type="submit" class="btn btn-primary" :disabled="submitting">
+            {{ submitting ? "Создание..." : "Создать канал" }}
+          </button>
+        </div>
+      </form>
+      <p v-if="error" class="form-error">{{ error }}</p>
+    </section>
+
+    <section class="card">
+      <h3>Список <span v-if="loading" class="muted">(загрузка...)</span></h3>
+      <div class="table-wrap">
+        <table class="admin-table">
+          <thead>
+            <tr>
+              <th>ID</th>
+              <th>Название</th>
+              <th>Канал</th>
+              <th>Тип</th>
+              <th>Активен</th>
+              <th>Создан</th>
+              <th></th>
+            </tr>
+          </thead>
+          <tbody>
+            <tr v-for="ch in channels" :key="ch.id">
+              <td>{{ ch.id }}</td>
+              <td>{{ ch.name }}</td>
+              <td><code>@{{ ch.channel }}</code></td>
+              <td>{{ ch.source_type }}</td>
+              <td>{{ ch.is_active ? "да" : "нет" }}</td>
+              <td>{{ formatDate(ch.created_at) }}</td>
+              <td class="actions">
+                <button type="button" class="btn btn-sm" @click="openEdit(ch)">Изменить</button>
+                <button type="button" class="btn btn-sm btn-danger" @click="handleDelete(ch)">
+                  Удалить
+                </button>
+              </td>
+            </tr>
+            <tr v-if="!loading && !channels.length">
+              <td colspan="7" class="muted">Пока нет каналов</td>
+            </tr>
+          </tbody>
+        </table>
+      </div>
+    </section>
+
+    <div v-if="editModal.open" class="modal-backdrop" @click.self="closeEdit">
+      <div class="modal">
+        <h3>Редактировать канал #{{ editModal.channel?.id }}</h3>
+        <form class="admin-form" @submit.prevent="handleSaveEdit">
+          <label>
+            Название
+            <input v-model="editModal.name" type="text" required />
+          </label>
+          <label>
+            Handle / id
+            <input v-model="editModal.handle" type="text" required />
+          </label>
+          <label class="checkbox-label">
+            <input v-model="editModal.is_active" type="checkbox" />
+            Активен
+          </label>
+          <div class="form-actions">
+            <button type="button" class="btn" @click="closeEdit">Отмена</button>
+            <button type="submit" class="btn btn-primary" :disabled="submitting">Сохранить</button>
+          </div>
+        </form>
+      </div>
+    </div>
+  </div>
+</template>
+
+<style scoped>
+.intro {
+  margin: 0 0 1rem;
+  font-size: 0.9rem;
+  color: #555;
+  line-height: 1.5;
+}
+
+.checkbox-label {
+  display: flex;
+  align-items: center;
+  gap: 0.5rem;
+  margin-top: 1.5rem;
+}
+
+.actions {
+  display: flex;
+  gap: 0.35rem;
+  flex-wrap: wrap;
+}
+
+.modal-backdrop {
+  position: fixed;
+  inset: 0;
+  background: rgba(0, 0, 0, 0.35);
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  z-index: 50;
+}
+
+.modal {
+  background: #fff;
+  border-radius: 8px;
+  padding: 1.25rem 1.5rem;
+  width: min(420px, 92vw);
+  box-shadow: 0 8px 24px rgba(0, 0, 0, 0.15);
+}
+
+.modal h3 {
+  margin: 0 0 1rem;
+  font-size: 1.05rem;
+}
+
+.modal .admin-form label {
+  display: block;
+  margin-bottom: 0.75rem;
+}
+</style>

centers/analytics/frontend/src/views/EventsView.vue

@@ -1,7 +1,7 @@
 <script setup lang="ts">
 import { computed, onMounted, ref, watch } from "vue";
 import { useRouter } from "vue-router";
-import { fetchEvents } from "../api/admin";
+import { createEvent, deleteEvent, fetchEvents, updateEvent } from "../api/admin";
 import type { EventFilters, EventRecord } from "../types/admin";
 
 const router = useRouter();
@@ -10,6 +10,7 @@ const items = ref<EventRecord[]>([]);
 const total = ref(0);
 const loading = ref(true);
 const error = ref("");
+const submitting = ref(false);
 
 const filters = ref({
   search: "",
@@ -24,8 +25,44 @@ const filters = ref({
 const page = ref(1);
 const pageSize = 25;
 
-const selectedEvent = ref<EventRecord | null>(null);
-const modalOpen = ref(false);
+type EventForm = {
+  source_type: string;
+  source_url: string;
+  title: string;
+  description: string;
+  locality: string;
+  region: string;
+  topic: string;
+  raw_text: string;
+  latitude: string;
+  longitude: string;
+  event_date: string;
+  tags: string;
+};
+
+function emptyForm(): EventForm {
+  return {
+    source_type: "manual",
+    source_url: "",
+    title: "",
+    description: "",
+    locality: "",
+    region: "",
+    topic: "",
+    raw_text: "",
+    latitude: "",
+    longitude: "",
+    event_date: "",
+    tags: "",
+  };
+}
+
+const createForm = ref(emptyForm());
+const editModal = ref<{ open: boolean; event: EventRecord | null; form: EventForm }>({
+  open: false,
+  event: null,
+  form: emptyForm(),
+});
 
 const totalPages = computed(() => Math.max(1, Math.ceil(total.value / pageSize)));
 
@@ -76,14 +113,128 @@ function resetFilters() {
   loadEvents();
 }
 
-function openModal(event: EventRecord) {
-  selectedEvent.value = event;
-  modalOpen.value = true;
+function parseOptionalFloat(value: string): number | null {
+  const trimmed = value.trim();
+  if (!trimmed) return null;
+  const num = Number(trimmed);
+  if (Number.isNaN(num)) throw new Error("Координаты должны быть числами");
+  return num;
 }
 
-function closeModal() {
-  modalOpen.value = false;
-  selectedEvent.value = null;
+function parseTags(value: string): string[] | null {
+  const tags = value
+    .split(",")
+    .map((t) => t.trim())
+    .filter(Boolean);
+  return tags.length ? tags : null;
+}
+
+function toDateTimeLocal(value: string | null): string {
+  if (!value) return "";
+  const d = new Date(value);
+  if (Number.isNaN(d.getTime())) return "";
+  const pad = (n: number) => String(n).padStart(2, "0");
+  return `${d.getFullYear()}-${pad(d.getMonth() + 1)}-${pad(d.getDate())}T${pad(d.getHours())}:${pad(d.getMinutes())}`;
+}
+
+function formToPayload(form: EventForm) {
+  const latitude = parseOptionalFloat(form.latitude);
+  const longitude = parseOptionalFloat(form.longitude);
+  if ((latitude == null) !== (longitude == null)) {
+    throw new Error("Укажите обе координаты или ни одной");
+  }
+  const source_url = form.source_url.trim();
+  if (!source_url) throw new Error("Укажите source_url");
+
+  return {
+    source_type: form.source_type.trim() || "manual",
+    source_url,
+    title: form.title.trim(),
+    description: form.description.trim(),
+    locality: form.locality.trim(),
+    region: form.region.trim() || null,
+    topic: form.topic.trim() || null,
+    raw_text: form.raw_text,
+    latitude,
+    longitude,
+    event_date: form.event_date ? new Date(form.event_date).toISOString() : null,
+    tags: parseTags(form.tags),
+  };
+}
+
+function eventToForm(event: EventRecord): EventForm {
+  return {
+    source_type: event.source_type || "manual",
+    source_url: event.source_url,
+    title: event.title || "",
+    description: event.description || "",
+    locality: event.locality || "",
+    region: event.region || "",
+    topic: event.topic || "",
+    raw_text: event.raw_text || "",
+    latitude: event.latitude != null ? String(event.latitude) : "",
+    longitude: event.longitude != null ? String(event.longitude) : "",
+    event_date: toDateTimeLocal(event.event_date),
+    tags: event.tags?.join(", ") ?? "",
+  };
+}
+
+async function handleCreate() {
+  submitting.value = true;
+  error.value = "";
+  try {
+    await createEvent(formToPayload(createForm.value));
+    createForm.value = emptyForm();
+    createForm.value.source_url = `manual://${crypto.randomUUID()}`;
+    await loadEvents();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось создать событие";
+  } finally {
+    submitting.value = false;
+  }
+}
+
+function openEdit(event: EventRecord) {
+  editModal.value = {
+    open: true,
+    event,
+    form: eventToForm(event),
+  };
+}
+
+function closeEdit() {
+  editModal.value.open = false;
+  editModal.value.event = null;
+}
+
+async function handleSaveEdit() {
+  if (!editModal.value.event) return;
+  submitting.value = true;
+  error.value = "";
+  try {
+    await updateEvent(editModal.value.event.id, formToPayload(editModal.value.form));
+    closeEdit();
+    await loadEvents();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось сохранить событие";
+  } finally {
+    submitting.value = false;
+  }
+}
+
+async function handleDelete(event: EventRecord) {
+  const label = event.title || event.source_url || `#${event.id}`;
+  if (!window.confirm(`Удалить событие #${event.id} («${label}»)? Связанная точка на карте тоже будет удалена.`)) {
+    return;
+  }
+  error.value = "";
+  try {
+    await deleteEvent(event.id);
+    if (editModal.value.event?.id === event.id) closeEdit();
+    await loadEvents();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось удалить событие";
+  }
 }
 
 function goToMap(event: EventRecord) {
@@ -103,13 +254,77 @@ function truncate(text: string, max = 80): string {
 
 watch(page, loadEvents);
 
-onMounted(loadEvents);
+onMounted(() => {
+  createForm.value.source_url = `manual://${crypto.randomUUID()}`;
+  loadEvents();
+});
 </script>
 
 <template>
   <div class="page">
     <h2 class="page-heading">События</h2>
 
+    <section class="card">
+      <h3>Новое событие</h3>
+      <form class="admin-form" @submit.prevent="handleCreate">
+        <div class="form-row">
+          <label>
+            Заголовок
+            <input v-model="createForm.title" type="text" placeholder="Краткий заголовок" />
+          </label>
+          <label>
+            Источник (type)
+            <input v-model="createForm.source_type" type="text" placeholder="manual" />
+          </label>
+          <label class="span-2">
+            source_url
+            <input v-model="createForm.source_url" type="text" required />
+          </label>
+          <label class="span-2">
+            Описание
+            <textarea v-model="createForm.description" rows="2" />
+          </label>
+          <label>
+            Регион
+            <input v-model="createForm.region" type="text" />
+          </label>
+          <label>
+            Тема
+            <input v-model="createForm.topic" type="text" />
+          </label>
+          <label>
+            Населённый пункт
+            <input v-model="createForm.locality" type="text" />
+          </label>
+          <label>
+            Дата события
+            <input v-model="createForm.event_date" type="datetime-local" />
+          </label>
+          <label>
+            Широта
+            <input v-model="createForm.latitude" type="text" placeholder="48.1234" />
+          </label>
+          <label>
+            Долгота
+            <input v-model="createForm.longitude" type="text" placeholder="37.6543" />
+          </label>
+          <label>
+            Теги (через запятую)
+            <input v-model="createForm.tags" type="text" placeholder="обстрел, нп" />
+          </label>
+          <label class="span-2">
+            Сырой текст
+            <textarea v-model="createForm.raw_text" rows="3" />
+          </label>
+        </div>
+        <div class="form-actions">
+          <button type="submit" class="btn btn-primary" :disabled="submitting">
+            {{ submitting ? "Создание..." : "Создать событие" }}
+          </button>
+        </div>
+      </form>
+    </section>
+
     <section class="card">
       <form class="admin-form filters-form" @submit.prevent="applyFilters">
         <div class="form-row">
@@ -177,7 +392,7 @@ onMounted(loadEvents);
             <tr v-for="event in items" :key="event.id">
               <td>{{ event.id }}</td>
               <td>
-                <button class="link-btn" @click="openModal(event)">
+                <button class="link-btn" type="button" @click="openEdit(event)">
                   {{ truncate(event.title || event.description || "—") }}
                 </button>
               </td>
@@ -191,14 +406,19 @@ onMounted(loadEvents);
                 </template>
                 <template v-else>—</template>
               </td>
-              <td>
+              <td class="actions">
+                <button class="btn btn-sm" type="button" @click="openEdit(event)">Изменить</button>
                 <button
                   v-if="event.latitude != null && event.longitude != null"
                   class="btn btn-sm"
+                  type="button"
                   @click="goToMap(event)"
                 >
                   На карте
                 </button>
+                <button class="btn btn-sm btn-danger" type="button" @click="handleDelete(event)">
+                  Удалить
+                </button>
               </td>
             </tr>
           </tbody>
@@ -206,56 +426,129 @@ onMounted(loadEvents);
       </div>
 
       <div class="pagination">
-        <button class="btn btn-sm" :disabled="page <= 1" @click="page--">← Назад</button>
+        <button class="btn btn-sm" type="button" :disabled="page <= 1" @click="page--">← Назад</button>
         <span>Стр. {{ page }} / {{ totalPages }}</span>
-        <button class="btn btn-sm" :disabled="page >= totalPages" @click="page++">Вперёд →</button>
+        <button class="btn btn-sm" type="button" :disabled="page >= totalPages" @click="page++">
+          Вперёд →
+        </button>
       </div>
     </section>
 
-    <div v-if="modalOpen && selectedEvent" class="modal-overlay" @click.self="closeModal">
-      <div class="modal">
+    <div v-if="editModal.open && editModal.event" class="modal-overlay" @click.self="closeEdit">
+      <div class="modal modal-wide">
         <header class="modal-header">
-          <h3>Событие #{{ selectedEvent.id }}</h3>
-          <button class="btn btn-sm" @click="closeModal">✕</button>
+          <h3>Событие #{{ editModal.event.id }}</h3>
+          <button class="btn btn-sm" type="button" @click="closeEdit">✕</button>
         </header>
-        <div class="modal-body">
-          <dl class="detail-list">
-            <dt>Заголовок</dt>
-            <dd>{{ selectedEvent.title || "—" }}</dd>
-            <dt>Описание</dt>
-            <dd>{{ selectedEvent.description || "—" }}</dd>
-            <dt>Источник</dt>
-            <dd>{{ selectedEvent.source_type }} — {{ selectedEvent.source_url }}</dd>
-            <dt>Регион / Тема</dt>
-            <dd>{{ selectedEvent.region ?? "—" }} / {{ selectedEvent.topic ?? "—" }}</dd>
-            <dt>Населённый пункт</dt>
-            <dd>{{ selectedEvent.locality || "—" }}</dd>
-            <dt>Дата события</dt>
-            <dd>{{ formatDate(selectedEvent.event_date) }}</dd>
-            <dt>Ingested</dt>
-            <dd>{{ formatDate(selectedEvent.ingested_at) }}</dd>
-            <dt>Координаты</dt>
-            <dd>
-              <template v-if="selectedEvent.latitude != null">
-                {{ selectedEvent.latitude }}, {{ selectedEvent.longitude }}
-              </template>
-              <template v-else>—</template>
-            </dd>
-            <dt>Текст</dt>
-            <dd class="raw-text">{{ selectedEvent.raw_text || "—" }}</dd>
-          </dl>
-        </div>
-        <footer class="modal-footer">
-          <button
-            v-if="selectedEvent.latitude != null"
-            class="btn btn-primary"
-            @click="goToMap(selectedEvent)"
-          >
-            Показать на карте
-          </button>
-          <button class="btn" @click="closeModal">Закрыть</button>
-        </footer>
+        <form class="modal-body admin-form" @submit.prevent="handleSaveEdit">
+          <div class="form-row">
+            <label>
+              Заголовок
+              <input v-model="editModal.form.title" type="text" />
+            </label>
+            <label>
+              Источник (type)
+              <input v-model="editModal.form.source_type" type="text" />
+            </label>
+            <label class="span-2">
+              source_url
+              <input v-model="editModal.form.source_url" type="text" required />
+            </label>
+            <label class="span-2">
+              Описание
+              <textarea v-model="editModal.form.description" rows="2" />
+            </label>
+            <label>
+              Регион
+              <input v-model="editModal.form.region" type="text" />
+            </label>
+            <label>
+              Тема
+              <input v-model="editModal.form.topic" type="text" />
+            </label>
+            <label>
+              Населённый пункт
+              <input v-model="editModal.form.locality" type="text" />
+            </label>
+            <label>
+              Дата события
+              <input v-model="editModal.form.event_date" type="datetime-local" />
+            </label>
+            <label>
+              Широта
+              <input v-model="editModal.form.latitude" type="text" />
+            </label>
+            <label>
+              Долгота
+              <input v-model="editModal.form.longitude" type="text" />
+            </label>
+            <label>
+              Теги
+              <input v-model="editModal.form.tags" type="text" />
+            </label>
+            <label class="span-2">
+              Сырой текст
+              <textarea v-model="editModal.form.raw_text" rows="4" />
+            </label>
+            <p class="hint span-2">
+              Ingested: {{ formatDate(editModal.event.ingested_at) }}. При координатах создаётся/обновляется
+              точка на карте; без координат связанная точка удаляется.
+            </p>
+          </div>
+          <div class="form-actions">
+            <button type="button" class="btn" @click="closeEdit">Отмена</button>
+            <button
+              v-if="editModal.event.latitude != null"
+              type="button"
+              class="btn"
+              @click="goToMap(editModal.event)"
+            >
+              На карте
+            </button>
+            <button type="submit" class="btn btn-primary" :disabled="submitting">
+              {{ submitting ? "Сохранение..." : "Сохранить" }}
+            </button>
+          </div>
+        </form>
       </div>
     </div>
   </div>
 </template>
+
+<style scoped>
+.form-row .span-2 {
+  grid-column: span 2;
+}
+
+.form-row textarea {
+  width: 100%;
+  font: inherit;
+  padding: 0.4rem 0.5rem;
+  border: 1px solid #d1d5db;
+  border-radius: 4px;
+  resize: vertical;
+}
+
+.actions {
+  display: flex;
+  gap: 0.35rem;
+  flex-wrap: wrap;
+  white-space: nowrap;
+}
+
+.hint {
+  margin: 0;
+  font-size: 0.85rem;
+  color: #666;
+  line-height: 1.4;
+}
+
+.modal-wide {
+  width: min(720px, 94vw);
+}
+
+.mono {
+  font-family: ui-monospace, SFMono-Regular, Menlo, Consolas, monospace;
+  font-size: 0.8rem;
+}
+</style>

centers/analytics/frontend/src/views/LoginView.vue

@@ -0,0 +1,146 @@
+<script setup lang="ts">
+import { ref } from "vue";
+import { useRoute, useRouter } from "vue-router";
+
+import { useAuth } from "../composables/useAuth";
+
+const route = useRoute();
+const router = useRouter();
+const { login, busy, error, authenticated } = useAuth();
+
+const username = ref("admin");
+const password = ref("");
+
+if (authenticated.value) {
+  const next = typeof route.query.next === "string" ? route.query.next : "/";
+  void router.replace(next || "/");
+}
+
+async function onSubmit() {
+  const ok = await login(username.value, password.value);
+  if (!ok) return;
+  const next = typeof route.query.next === "string" ? route.query.next : "/";
+  await router.replace(next || "/");
+}
+</script>
+
+<template>
+  <div class="login-page">
+    <form class="login-card" @submit.prevent="onSubmit">
+      <h1>Вход в админку</h1>
+      <p class="hint">Карта доступна без входа. Остальные разделы — после авторизации.</p>
+
+      <label>
+        Логин
+        <input v-model="username" type="text" autocomplete="username" required />
+      </label>
+
+      <label>
+        Пароль
+        <input v-model="password" type="password" autocomplete="current-password" required />
+      </label>
+
+      <p v-if="error" class="error">{{ error }}</p>
+
+      <button type="submit" class="btn-primary" :disabled="busy">
+        {{ busy ? "Вход..." : "Войти" }}
+      </button>
+
+      <router-link class="back-link" to="/">← К карте</router-link>
+    </form>
+  </div>
+</template>
+
+<style scoped>
+.login-page {
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  min-height: 100%;
+  padding: 2rem 1rem;
+  background: linear-gradient(160deg, #eef2f7 0%, #f5f5f5 45%, #e8edf3 100%);
+}
+
+.login-card {
+  width: 100%;
+  max-width: 380px;
+  display: flex;
+  flex-direction: column;
+  gap: 0.85rem;
+  padding: 1.75rem;
+  background: #fff;
+  border: 1px solid #e0e0e0;
+  border-radius: 10px;
+  box-shadow: 0 8px 24px rgba(0, 0, 0, 0.06);
+}
+
+h1 {
+  margin: 0;
+  font-size: 1.25rem;
+  font-weight: 650;
+}
+
+.hint {
+  margin: 0;
+  font-size: 0.875rem;
+  color: #666;
+  line-height: 1.4;
+}
+
+label {
+  display: flex;
+  flex-direction: column;
+  gap: 0.35rem;
+  font-size: 0.875rem;
+  color: #444;
+}
+
+input {
+  padding: 0.55rem 0.7rem;
+  border: 1px solid #d0d0d0;
+  border-radius: 6px;
+  background: #fff;
+}
+
+input:focus {
+  outline: 2px solid #bbdefb;
+  border-color: #64b5f6;
+}
+
+.error {
+  margin: 0;
+  font-size: 0.85rem;
+  color: #b91c1c;
+}
+
+.btn-primary {
+  margin-top: 0.25rem;
+  padding: 0.65rem 1rem;
+  border: none;
+  border-radius: 6px;
+  background: #1565c0;
+  color: #fff;
+  font-weight: 500;
+  cursor: pointer;
+}
+
+.btn-primary:disabled {
+  opacity: 0.65;
+  cursor: not-allowed;
+}
+
+.btn-primary:not(:disabled):hover {
+  background: #0d47a1;
+}
+
+.back-link {
+  text-align: center;
+  font-size: 0.875rem;
+  color: #1565c0;
+  text-decoration: none;
+}
+
+.back-link:hover {
+  text-decoration: underline;
+}
+</style>

centers/analytics/frontend/src/views/MapViewPage.vue

@@ -16,12 +16,14 @@ import CoordsTools from "../components/map/CoordsTools.vue";
 import MapToolbar from "../components/map/MapToolbar.vue";
 import PlaceSearch from "../components/map/PlaceSearch.vue";
 import ObjectPanel from "../components/ObjectPanel.vue";
+import { useAuth } from "../composables/useAuth";
 import type { LeafletMapApi } from "../composables/useLeafletMap";
 import type { DateRangePreset, MapFilters, MapObjectWithEvent, MapQueryParams } from "../types/map";
 import type { MapObjectCreate, ObjectType } from "../types/object";
 
 const route = useRoute();
 const router = useRouter();
+const { authenticated } = useAuth();
 
 const objects = ref<MapObjectWithEvent[]>([]);
 const mapFilters = ref<MapFilters | null>(null);
@@ -162,6 +164,7 @@ function handleMapContextMenu(payload: {
   longitude: number;
   object: MapObjectWithEvent | null;
 }) {
+  if (!authenticated.value) return;
   contextMenu.value = {
     visible: true,
     x: payload.x,
@@ -257,6 +260,7 @@ async function handleMoveObject(payload: {
   latitude: number;
   longitude: number;
 }) {
+  if (!authenticated.value) return;
   if (payload.object.event_id) return;
   try {
     await updateObject(payload.object.id, {
@@ -327,17 +331,22 @@ onUnmounted(() => {
         :objects="objects"
         :selected-id="selectedId"
         :open-popup-id="openPopupId"
+        :editable="authenticated"
         @ready="onMapReady"
         @center-change="onCenterChange"
         @select="handleSelectObject"
         @contextmenu="handleMapContextMenu"
         @move="handleMoveObject"
       />
-      <ObjectPanel :object="selectedObject" @open-events="openInEvents" />
+      <ObjectPanel
+        :object="selectedObject"
+        :can-edit="authenticated"
+        @open-events="openInEvents"
+      />
     </div>
 
     <ContextMenu
-      v-if="contextMenu.visible"
+      v-if="contextMenu.visible && authenticated"
       :x="contextMenu.x"
       :y="contextMenu.y"
       :target="contextMenu.target"
@@ -350,7 +359,7 @@ onUnmounted(() => {
     />
 
     <CreateObjectModal
-      v-if="createModal.visible"
+      v-if="createModal.visible && authenticated"
       :latitude="createModal.latitude"
       :longitude="createModal.longitude"
       @close="closeCreateModal"
@@ -358,7 +367,7 @@ onUnmounted(() => {
     />
 
     <EditObjectModal
-      v-if="editModal.visible && editModal.object"
+      v-if="editModal.visible && editModal.object && authenticated"
       :object="editModal.object"
       @close="closeEditModal"
       @submit="handleEditObject"

centers/analytics/frontend/src/views/ParserProfilesView.vue

@@ -0,0 +1,887 @@
+<script setup lang="ts">
+import { onMounted, ref, watch } from "vue";
+import {
+  createProfile,
+  deleteProfile,
+  fetchLlmDefaults,
+  fetchProfiles,
+  fetchTargetFields,
+  generateParserProfile,
+  previewLlmProfile,
+  previewParserProfile,
+  updateProfile,
+  type HeuristicProfile,
+  type LlmProfile,
+  type TargetField,
+} from "../api/entities";
+import type { ParserProfile } from "../types/admin";
+
+const profiles = ref<ParserProfile[]>([]);
+const loading = ref(true);
+const error = ref("");
+const success = ref("");
+const submitting = ref(false);
+
+const profileKind = ref<"heuristic" | "llm">("heuristic");
+const samplePost = ref("");
+const generateHint = ref("");
+const profileName = ref("");
+const targetFields = ref<TargetField[]>([]);
+const profileJson = ref("");
+const llmInstruction = ref("");
+const llmSchemaJson = ref("");
+const llmMultiEvent = ref(false);
+const previewFields = ref<Record<string, string> | null>(null);
+const previewEvents = ref<Record<string, string>[]>([]);
+const emptyFields = ref<string[]>([]);
+const requiredFields = ref<string[]>([]);
+const previewMatched = ref(true);
+const previewMissing = ref<string[]>([]);
+const previewIsEvent = ref(true);
+const previewMatchedCount = ref(0);
+const generating = ref(false);
+const previewing = ref(false);
+const loadingFields = ref(true);
+const editingId = ref<number | null>(null);
+const profileStatus = ref<"draft" | "ready">("draft");
+const llmDefaults = ref<LlmProfile | null>(null);
+
+async function loadProfiles() {
+  loading.value = true;
+  try {
+    profiles.value = await fetchProfiles();
+    error.value = "";
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось загрузить профили";
+  } finally {
+    loading.value = false;
+  }
+}
+
+async function loadTargetFields() {
+  loadingFields.value = true;
+  try {
+    const data = await fetchTargetFields();
+    targetFields.value = data.fields;
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось загрузить целевые поля";
+  } finally {
+    loadingFields.value = false;
+  }
+}
+
+async function loadLlmDefaults() {
+  try {
+    llmDefaults.value = await fetchLlmDefaults();
+  } catch {
+    llmDefaults.value = null;
+  }
+}
+
+function applyLlmDefaults() {
+  const d = llmDefaults.value;
+  if (!d) return;
+  llmInstruction.value = d.instruction || "";
+  llmSchemaJson.value = JSON.stringify(d.extract_schema || {}, null, 2);
+  if (!requiredFields.value.length) {
+    requiredFields.value = Array.isArray(d.required_fields) ? [...d.required_fields] : [];
+  }
+  if (typeof d.multi_event === "boolean") {
+    llmMultiEvent.value = d.multi_event;
+  }
+}
+
+function syncHeuristicFromJson(): HeuristicProfile | null {
+  if (!profileJson.value.trim()) return null;
+  try {
+    return JSON.parse(profileJson.value) as HeuristicProfile;
+  } catch {
+    error.value = "Некорректный JSON профиля";
+    return null;
+  }
+}
+
+function syncLlmFromForm(): LlmProfile | null {
+  let schema: Record<string, string>;
+  try {
+    schema = JSON.parse(llmSchemaJson.value || "{}") as Record<string, string>;
+  } catch {
+    error.value = "Некорректный JSON extract_schema";
+    return null;
+  }
+  if (!schema || typeof schema !== "object" || !Object.keys(schema).length) {
+    error.value = "extract_schema не должен быть пустым";
+    return null;
+  }
+  return {
+    instruction: llmInstruction.value.trim() || null,
+    extract_schema: schema,
+    required_fields: [...requiredFields.value],
+    multi_event: llmMultiEvent.value,
+  };
+}
+
+const DEFAULT_REQUIRED = ["event_date", "coords"];
+
+function uniqueFields(names: string[]): string[] {
+  return names.filter((name, idx) => names.indexOf(name) === idx);
+}
+
+function writeRequiredToHeuristicJson(names: string[]) {
+  const current = syncHeuristicFromJson();
+  if (!current) return;
+  current.required_fields = names;
+  profileJson.value = JSON.stringify(current, null, 2);
+}
+
+function requiredFromHeuristic(
+  profile: HeuristicProfile,
+  preview: Record<string, string> | null,
+): string[] {
+  if (Array.isArray(profile.required_fields)) {
+    return uniqueFields(profile.required_fields);
+  }
+  if (!preview) return [];
+  return DEFAULT_REQUIRED.filter((name) => !!(preview[name] || "").trim());
+}
+
+function isRequired(name: string): boolean {
+  return requiredFields.value.includes(name);
+}
+
+async function toggleRequired(name: string) {
+  const next = isRequired(name)
+    ? requiredFields.value.filter((n) => n !== name)
+    : [...requiredFields.value, name];
+  requiredFields.value = next;
+  if (profileKind.value === "heuristic") {
+    writeRequiredToHeuristicJson(next);
+  }
+  if (previewFields.value && samplePost.value.trim()) {
+    await handlePreview();
+  }
+}
+
+watch(profileKind, (kind, prev) => {
+  if (kind === prev) return;
+  previewFields.value = null;
+  previewEvents.value = [];
+  emptyFields.value = [];
+  previewMatched.value = true;
+  previewMissing.value = [];
+  previewIsEvent.value = true;
+  previewMatchedCount.value = 0;
+  success.value = "";
+  if (kind === "llm" && !llmSchemaJson.value.trim()) {
+    applyLlmDefaults();
+  }
+});
+
+async function handleGenerate() {
+  if (profileKind.value === "llm") {
+    applyLlmDefaults();
+    success.value = "Подставлены дефолты LLM schema/instruction. Отредактируйте и Preview.";
+    return;
+  }
+  if (!samplePost.value.trim()) {
+    error.value = "Вставьте образец поста";
+    return;
+  }
+  generating.value = true;
+  error.value = "";
+  success.value = "";
+  previewFields.value = null;
+  emptyFields.value = [];
+  try {
+    let current: HeuristicProfile | null = null;
+    if (profileJson.value.trim()) {
+      current = syncHeuristicFromJson();
+      if (!current) return;
+    }
+    const data = await generateParserProfile({
+      sample_post: samplePost.value,
+      hint: generateHint.value.trim() || undefined,
+      current_profile: current,
+    });
+    const required =
+      current && Array.isArray(current.required_fields)
+        ? uniqueFields(current.required_fields)
+        : DEFAULT_REQUIRED.filter((name) => !!(data.preview[name] || "").trim());
+    data.profile.required_fields = required;
+    profileJson.value = JSON.stringify(data.profile, null, 2);
+    requiredFields.value = required;
+    previewFields.value = data.preview;
+    emptyFields.value = data.empty_fields || [];
+    const previewData = await previewParserProfile(samplePost.value, data.profile);
+    previewMatched.value = previewData.matched;
+    previewMissing.value = previewData.missing_required || [];
+    if (emptyFields.value.length) {
+      success.value =
+        `Профиль обновлён, но пустые поля: ${emptyFields.value.join(", ")}. ` +
+        "Допишите подсказку ниже и снова нажмите Generate.";
+      if (!generateHint.value.trim() && emptyFields.value.includes("description")) {
+        generateHint.value =
+          "description должен содержать многострочный текст тела поста после заголовка и до координат (47.…, 36.…), без строки заголовка и без хвоста Источник/Геопривязка/#теги";
+      }
+    } else {
+      success.value = "Профиль сгенерирован. Preview и runtime — только по правилам, без LLM.";
+    }
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Ошибка генерации";
+  } finally {
+    generating.value = false;
+  }
+}
+
+async function handlePreview() {
+  if (!samplePost.value.trim()) {
+    error.value = "Нужен образец поста для preview";
+    return;
+  }
+  previewing.value = true;
+  error.value = "";
+  try {
+    if (profileKind.value === "llm") {
+      const llm = syncLlmFromForm();
+      if (!llm) return;
+      const data = await previewLlmProfile(samplePost.value, llm);
+      const rows =
+        Array.isArray(data.events) && data.events.length
+          ? data.events
+          : data.fields
+            ? [data.fields]
+            : [];
+      previewEvents.value = rows;
+      previewFields.value = rows[0] || data.fields || null;
+      emptyFields.value = previewFields.value
+        ? Object.entries(previewFields.value)
+            .filter(([, v]) => !v)
+            .map(([k]) => k)
+        : [];
+      previewMatched.value = data.matched;
+      previewMissing.value = data.missing_required || [];
+      previewIsEvent.value = data.is_event;
+      previewMatchedCount.value = data.matched_count ?? rows.length;
+      if (!data.is_event) {
+        success.value = "Модель пометила текст как не-событие (is_event=false) — runtime пропустит.";
+      } else if (rows.length > 1) {
+        success.value = `Preview: ${rows.length} событий из поста (пройдут фильтр: ${previewMatchedCount.value}).`;
+      }
+    } else {
+      const current = syncHeuristicFromJson();
+      if (!current) {
+        if (!error.value) error.value = "Сначала сгенерируйте или вставьте профиль";
+        return;
+      }
+      const data = await previewParserProfile(samplePost.value, current);
+      previewFields.value = data.fields;
+      emptyFields.value = Object.entries(data.fields)
+        .filter(([, v]) => !v)
+        .map(([k]) => k);
+      requiredFields.value = requiredFromHeuristic(current, data.fields);
+      if (!Array.isArray(current.required_fields)) {
+        writeRequiredToHeuristicJson(requiredFields.value);
+      }
+      previewMatched.value = data.matched;
+      previewMissing.value = data.missing_required || [];
+      previewIsEvent.value = true;
+      previewEvents.value = data.fields ? [data.fields] : [];
+      previewMatchedCount.value = data.matched ? 1 : 0;
+    }
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Ошибка preview";
+  } finally {
+    previewing.value = false;
+  }
+}
+
+function resetEditor() {
+  editingId.value = null;
+  profileKind.value = "heuristic";
+  profileName.value = "";
+  samplePost.value = "";
+  generateHint.value = "";
+  profileJson.value = "";
+  llmInstruction.value = "";
+  llmSchemaJson.value = "";
+  llmMultiEvent.value = false;
+  profileStatus.value = "draft";
+  previewFields.value = null;
+  previewEvents.value = [];
+  emptyFields.value = [];
+  requiredFields.value = [];
+  previewMatched.value = true;
+  previewMissing.value = [];
+  previewIsEvent.value = true;
+  previewMatchedCount.value = 0;
+  success.value = "";
+}
+
+function openEdit(p: ParserProfile) {
+  editingId.value = p.id;
+  profileKind.value = p.kind === "llm" ? "llm" : "heuristic";
+  profileName.value = p.name;
+  samplePost.value = p.sample_post || "";
+  generateHint.value = "";
+  profileStatus.value = p.status === "ready" ? "ready" : "draft";
+  previewFields.value = null;
+  previewEvents.value = [];
+  emptyFields.value = [];
+  previewMatched.value = true;
+  previewMissing.value = [];
+  previewIsEvent.value = true;
+  previewMatchedCount.value = 0;
+  success.value = "";
+  error.value = "";
+
+  if (p.kind === "llm") {
+    profileJson.value = "";
+    const lp = (p.llm_profile || {}) as LlmProfile;
+    llmInstruction.value = lp.instruction || "";
+    llmSchemaJson.value = JSON.stringify(lp.extract_schema || {}, null, 2);
+    llmMultiEvent.value = !!lp.multi_event;
+    requiredFields.value = Array.isArray(lp.required_fields)
+      ? uniqueFields(lp.required_fields)
+      : [];
+    if (!llmSchemaJson.value.trim() || llmSchemaJson.value === "{}") {
+      applyLlmDefaults();
+    }
+  } else {
+    llmInstruction.value = "";
+    llmSchemaJson.value = "";
+    llmMultiEvent.value = false;
+    profileJson.value = p.heuristic_profile
+      ? JSON.stringify(p.heuristic_profile, null, 2)
+      : "";
+    const hp = p.heuristic_profile;
+    if (hp && Array.isArray(hp.required_fields)) {
+      requiredFields.value = uniqueFields(hp.required_fields as string[]);
+    } else if (hp) {
+      requiredFields.value = [...DEFAULT_REQUIRED];
+      writeRequiredToHeuristicJson(requiredFields.value);
+    } else {
+      requiredFields.value = [];
+    }
+  }
+  window.scrollTo({ top: 0, behavior: "smooth" });
+}
+
+async function handleSave() {
+  if (!profileName.value.trim()) {
+    error.value = "Укажите название профиля";
+    return;
+  }
+  submitting.value = true;
+  error.value = "";
+  success.value = "";
+  try {
+    if (profileKind.value === "llm") {
+      const llm = syncLlmFromForm();
+      if (!llm) return;
+      const payload = {
+        name: profileName.value.trim(),
+        kind: "llm" as const,
+        sample_post: samplePost.value,
+        llm_profile: llm,
+        heuristic_profile: null,
+        status: profileStatus.value,
+      };
+      if (editingId.value != null) {
+        await updateProfile(editingId.value, payload);
+        success.value = `LLM-профиль #${editingId.value} обновлён`;
+      } else {
+        const created = await createProfile({
+          ...payload,
+          status: profileStatus.value === "draft" ? "draft" : "ready",
+        });
+        success.value = `LLM-профиль #${created.id} сохранён`;
+        editingId.value = created.id;
+        profileStatus.value = created.status === "ready" ? "ready" : "draft";
+      }
+    } else {
+      writeRequiredToHeuristicJson(requiredFields.value);
+      const current = syncHeuristicFromJson();
+      if (!current) {
+        if (!error.value) error.value = "Нужен JSON профиля (Generate или вручную)";
+        return;
+      }
+      const payload = {
+        name: profileName.value.trim(),
+        kind: "heuristic" as const,
+        sample_post: samplePost.value,
+        heuristic_profile: current,
+        llm_profile: null,
+        status: profileStatus.value,
+      };
+      if (editingId.value != null) {
+        await updateProfile(editingId.value, payload);
+        success.value = `Профиль #${editingId.value} обновлён`;
+      } else {
+        const created = await createProfile({
+          ...payload,
+          status: profileStatus.value === "draft" ? "draft" : "ready",
+        });
+        success.value = `Профиль #${created.id} сохранён`;
+        editingId.value = created.id;
+        profileStatus.value = created.status === "ready" ? "ready" : "draft";
+      }
+    }
+    await loadProfiles();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось сохранить профиль";
+  } finally {
+    submitting.value = false;
+  }
+}
+
+async function handleDelete(p: ParserProfile) {
+  if (!window.confirm(`Удалить профиль «${p.name}»?`)) return;
+  error.value = "";
+  try {
+    await deleteProfile(p.id);
+    if (editingId.value === p.id) resetEditor();
+    await loadProfiles();
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось удалить";
+  }
+}
+
+function formatDate(value: string): string {
+  return new Date(value).toLocaleString("ru-RU");
+}
+
+function kindLabel(kind: string | undefined): string {
+  return kind === "llm" ? "llm" : "heuristic";
+}
+
+function canSave(): boolean {
+  if (!profileName.value.trim()) return false;
+  if (profileKind.value === "llm") return !!llmSchemaJson.value.trim();
+  return !!profileJson.value.trim();
+}
+
+onMounted(async () => {
+  await Promise.all([loadProfiles(), loadTargetFields(), loadLlmDefaults()]);
+});
+</script>
+
+<template>
+  <div class="page">
+    <h2 class="page-heading">Профили парсера</h2>
+    <p class="intro">
+      Два вида профилей под фиксированные поля Event:
+      <strong>heuristic</strong> — статичные правила (Generate один раз, runtime без LLM);
+      <strong>llm</strong> — instruction + schema, DeepSeek на каждый пост в batch.
+      Связка с каналом — на
+      <router-link to="/parsers">Парсеры</router-link>.
+    </p>
+
+    <section class="card">
+      <div class="card-head">
+        <h3>{{ editingId != null ? `Редактирование #${editingId}` : "Новый профиль" }}</h3>
+        <button v-if="editingId != null" type="button" class="btn btn-sm" @click="resetEditor">
+          Сбросить
+        </button>
+      </div>
+      <div class="form-row">
+        <label>
+          Название
+          <input v-model="profileName" type="text" placeholder="Сводка: дата + НП + coords" />
+        </label>
+        <label>
+          Тип
+          <select v-model="profileKind" :disabled="editingId != null">
+            <option value="heuristic">heuristic (правила)</option>
+            <option value="llm">llm (DeepSeek runtime)</option>
+          </select>
+        </label>
+        <label>
+          Статус
+          <select v-model="profileStatus">
+            <option value="draft">draft</option>
+            <option value="ready">ready</option>
+          </select>
+        </label>
+      </div>
+      <label>
+        Образец поста
+        <textarea
+          v-model="samplePost"
+          class="sample-area"
+          rows="10"
+          placeholder="15.03.2024 Населённый пункт&#10;&#10;Описание…&#10;&#10;48.123456, 37.654321"
+        />
+      </label>
+
+      <template v-if="profileKind === 'heuristic'">
+        <label>
+          Подсказка агенту
+          <span class="muted"> (если Generate ошибся — опишите, что исправить, и нажмите Generate снова)</span>
+          <textarea
+            v-model="generateHint"
+            class="hint-area"
+            rows="3"
+            placeholder="например: description — абзацы между заголовком и координатами, без #хештегов"
+          />
+        </label>
+      </template>
+
+      <template v-else>
+        <label>
+          Instruction
+          <textarea
+            v-model="llmInstruction"
+            class="hint-area"
+            rows="3"
+            placeholder="Инструкция для DeepSeek на каждый пост"
+          />
+        </label>
+        <label class="required-item multi-flag">
+          <input v-model="llmMultiEvent" type="checkbox" />
+          Несколько событий в посте
+          <span class="muted"> (LLM вернёт массив; ingest с URL #e1, #e2…)</span>
+        </label>
+      </template>
+
+      <div class="form-actions">
+        <button
+          type="button"
+          class="btn btn-primary"
+          :disabled="generating || (profileKind === 'heuristic' && !samplePost.trim())"
+          @click="handleGenerate"
+        >
+          <template v-if="profileKind === 'llm'">
+            {{ generating ? "…" : "Подставить дефолты schema" }}
+          </template>
+          <template v-else>
+            {{
+              generating
+                ? "Генерация..."
+                : generateHint.trim() || profileJson.trim()
+                  ? "Generate / Refine"
+                  : "Generate"
+            }}
+          </template>
+        </button>
+        <button
+          type="button"
+          class="btn"
+          :disabled="
+            previewing ||
+            (profileKind === 'heuristic' ? !profileJson.trim() : !llmSchemaJson.trim())
+          "
+          @click="handlePreview"
+        >
+          {{ previewing ? "Preview..." : "Preview" }}
+        </button>
+        <button
+          type="button"
+          class="btn btn-primary"
+          :disabled="submitting || !canSave()"
+          @click="handleSave"
+        >
+          {{ submitting ? "Сохранение..." : editingId != null ? "Обновить" : "Сохранить профиль" }}
+        </button>
+      </div>
+    </section>
+
+    <div class="grid-2">
+      <section class="card">
+        <h3>Целевые поля <span class="muted">(Event)</span></h3>
+        <p v-if="loadingFields" class="muted">Загрузка…</p>
+        <div v-else class="table-wrap">
+          <table class="admin-table">
+            <thead>
+              <tr>
+                <th>Поле</th>
+                <th>Тип</th>
+                <th>Описание</th>
+              </tr>
+            </thead>
+            <tbody>
+              <tr v-for="field in targetFields" :key="field.name">
+                <td><code>{{ field.name }}</code></td>
+                <td>{{ field.type }}</td>
+                <td>{{ field.description }}</td>
+              </tr>
+            </tbody>
+          </table>
+        </div>
+      </section>
+
+      <section class="card">
+        <h3 v-if="profileKind === 'heuristic'">Профиль (JSON)</h3>
+        <h3 v-else>extract_schema (JSON)</h3>
+        <textarea
+          v-if="profileKind === 'heuristic'"
+          v-model="profileJson"
+          class="profile-area"
+          rows="16"
+          spellcheck="false"
+          placeholder="Появится после Generate"
+        />
+        <textarea
+          v-else
+          v-model="llmSchemaJson"
+          class="profile-area"
+          rows="16"
+          spellcheck="false"
+          placeholder="Ключ → описание поля для LLM"
+        />
+      </section>
+    </div>
+
+    <section v-if="previewFields" class="card">
+      <h3>
+        Preview
+        <span v-if="previewEvents.length > 1" class="muted">
+          ({{ previewEvents.length }} событий)
+        </span>
+        <span v-else>строки</span>
+      </h3>
+      <p v-if="profileKind === 'llm' && !previewIsEvent" class="form-warn">
+        is_event=false — пост не будет ингеститься.
+      </p>
+      <p v-if="profileKind === 'llm' && previewEvents.length > 1" class="form-success match-ok">
+        Из поста извлечено {{ previewEvents.length }} событий;
+        пройдут required_fields: {{ previewMatchedCount }}.
+        Runtime URL: пост#e1 … #e{{ previewEvents.length }}.
+      </p>
+      <p v-if="emptyFields.length" class="form-warn">
+        Пустые поля (первая строка): <code>{{ emptyFields.join(", ") }}</code>
+        <template v-if="profileKind === 'heuristic'">
+          — уточните подсказку и Generate / Refine.
+        </template>
+      </p>
+      <p v-if="!requiredFields.length" class="form-warn">
+        Нет обязательных полей —
+        <template v-if="profileKind === 'llm'">
+          фильтр только по is_event.
+        </template>
+        <template v-else>
+          любой пост канала будет сохранён. Отметьте поля, без которых пост нужно отбрасывать.
+        </template>
+      </p>
+      <p v-else-if="!previewMatched && previewEvents.length <= 1" class="form-warn">
+        Этот образец не прошёл бы фильтр. Не хватает:
+        <code>{{ previewMissing.join(", ") }}</code>
+      </p>
+      <p v-else-if="previewMatched" class="form-success match-ok">
+        <template v-if="previewEvents.length <= 1">
+          Образец проходит фильтр. Runtime сохранит только посты с заполненными
+          <code>{{ requiredFields.join(", ") }}</code>.
+        </template>
+        <template v-else>
+          Обязательные поля: <code>{{ requiredFields.join(", ") }}</code>
+          (проверка на каждое событие).
+        </template>
+      </p>
+      <div class="table-wrap">
+        <table class="admin-table">
+          <thead>
+            <tr>
+              <th v-if="previewEvents.length > 1">#</th>
+              <th v-for="key in Object.keys(previewFields)" :key="key">{{ key }}</th>
+            </tr>
+          </thead>
+          <tbody>
+            <tr v-for="(row, idx) in (previewEvents.length ? previewEvents : [previewFields])" :key="idx">
+              <td v-if="previewEvents.length > 1">{{ idx + 1 }}</td>
+              <td
+                v-for="key in Object.keys(previewFields)"
+                :key="key"
+                class="preview-cell"
+                :class="{ empty: !row?.[key] }"
+              >
+                {{ row?.[key] || "—" }}
+              </td>
+            </tr>
+          </tbody>
+        </table>
+      </div>
+      <div class="required-box">
+        <h4>Обязательно для сохранения</h4>
+        <p class="muted required-hint">
+          <template v-if="profileKind === 'llm'">
+            После LLM: событие отбрасывается, если обязательное поле пустое (поверх is_event).
+          </template>
+          <template v-else>
+            Пост отбрасывается, если хоть одно отмеченное поле не извлеклось (для coords — ещё и не парсится lat,lon).
+          </template>
+        </p>
+        <div class="required-grid">
+          <label
+            v-for="key in Object.keys(previewFields)"
+            :key="key"
+            class="required-item"
+          >
+            <input
+              type="checkbox"
+              :checked="isRequired(key)"
+              @change="toggleRequired(key)"
+            />
+            <code>{{ key }}</code>
+          </label>
+        </div>
+      </div>
+    </section>
+
+    <p v-if="error" class="form-error">{{ error }}</p>
+    <p v-if="success" class="form-success">{{ success }}</p>
+
+    <section class="card">
+      <h3>Сохранённые профили <span v-if="loading" class="muted">(загрузка...)</span></h3>
+      <div class="table-wrap">
+        <table class="admin-table">
+          <thead>
+            <tr>
+              <th>ID</th>
+              <th>Название</th>
+              <th>Тип</th>
+              <th>Статус</th>
+              <th>Создан</th>
+              <th></th>
+            </tr>
+          </thead>
+          <tbody>
+            <tr v-for="p in profiles" :key="p.id">
+              <td>{{ p.id }}</td>
+              <td>{{ p.name }}</td>
+              <td><code>{{ kindLabel(p.kind) }}</code></td>
+              <td>{{ p.status }}</td>
+              <td>{{ formatDate(p.created_at) }}</td>
+              <td class="actions">
+                <button type="button" class="btn btn-sm" @click="openEdit(p)">Изменить</button>
+                <button type="button" class="btn btn-sm btn-danger" @click="handleDelete(p)">
+                  Удалить
+                </button>
+              </td>
+            </tr>
+            <tr v-if="!loading && !profiles.length">
+              <td colspan="6" class="muted">Пока нет профилей</td>
+            </tr>
+          </tbody>
+        </table>
+      </div>
+    </section>
+  </div>
+</template>
+
+<style scoped>
+.intro {
+  margin: 0 0 1rem;
+  font-size: 0.9rem;
+  color: #555;
+  line-height: 1.5;
+  max-width: 72ch;
+}
+
+.card-head {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 0.75rem;
+  margin-bottom: 0.5rem;
+}
+
+.card-head h3 {
+  margin: 0;
+}
+
+.sample-area,
+.profile-area,
+.hint-area {
+  width: 100%;
+  font: inherit;
+  font-family: ui-monospace, SFMono-Regular, Menlo, Consolas, monospace;
+  font-size: 0.8rem;
+  padding: 0.6rem 0.75rem;
+  border: 1px solid #d1d5db;
+  border-radius: 6px;
+  resize: vertical;
+  line-height: 1.45;
+  margin-top: 0.35rem;
+}
+
+.hint-area {
+  font-family: inherit;
+}
+
+.grid-2 {
+  display: grid;
+  grid-template-columns: 1fr 1fr;
+  gap: 1rem;
+}
+
+@media (max-width: 960px) {
+  .grid-2 {
+    grid-template-columns: 1fr;
+  }
+}
+
+.preview-cell {
+  max-width: 180px;
+  white-space: pre-wrap;
+  word-break: break-word;
+  font-size: 0.8rem;
+}
+
+.preview-cell.empty {
+  background: #fff8e1;
+  color: #b26a00;
+}
+
+.form-success {
+  color: #2e7d32;
+  font-size: 0.875rem;
+  margin-top: 0.5rem;
+}
+
+.form-warn {
+  color: #b26a00;
+  font-size: 0.875rem;
+  margin: 0 0 0.75rem;
+}
+
+.match-ok {
+  margin: 0 0 0.75rem;
+}
+
+.required-box {
+  margin-top: 1rem;
+}
+
+.required-box h4 {
+  margin: 0 0 0.25rem;
+  font-size: 0.9rem;
+}
+
+.required-hint {
+  margin: 0 0 0.6rem;
+  font-size: 0.8rem;
+}
+
+.required-grid {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 0.5rem 1rem;
+}
+
+.required-item {
+  display: flex;
+  align-items: center;
+  gap: 0.35rem;
+  font-size: 0.85rem;
+}
+
+.multi-flag {
+  margin: 0.75rem 0 0;
+  flex-wrap: wrap;
+}
+
+.actions {
+  display: flex;
+  gap: 0.35rem;
+  flex-wrap: wrap;
+}
+</style>

centers/analytics/frontend/src/views/ParsersView.vue

@@ -1,7 +1,8 @@
 <script setup lang="ts">
 import { computed, onMounted, onUnmounted, ref } from "vue";
 import { createJob, deleteJob, fetchJobs, retryJob, updateJob } from "../api/admin";
-import type { ParseJob } from "../types/admin";
+import { fetchChannels, fetchProfiles } from "../api/entities";
+import type { ParseChannel, ParseJob, ParserProfile } from "../types/admin";
 
 const INTERVAL_OPTIONS = [
   { value: 900, label: "15 мин" },
@@ -12,83 +13,69 @@ const INTERVAL_OPTIONS = [
   { value: 86400, label: "24 ч" },
 ];
 
-const SOURCE_TYPES = [
-  { value: "telegram", label: "Telegram" },
-  { value: "crawl4ai", label: "Crawl4AI (web)" },
-  { value: "viina", label: "VIINA (news NLP)" },
-] as const;
-
 const jobs = ref<ParseJob[]>([]);
+const channels = ref<ParseChannel[]>([]);
+const profiles = ref<ParserProfile[]>([]);
 const loading = ref(true);
 const error = ref("");
 const submitting = ref(false);
+const forcingJobId = ref<number | null>(null);
 const editingJob = ref<ParseJob | null>(null);
 
+const pairForm = ref({
+  channel_id: null as number | null,
+  profile_id: null as number | null,
+  limit: 10,
+  interval_seconds: 3600,
+});
+
 const editForm = ref({
-  channel: "",
-  limit: 50,
-  urlsText: "",
-  textsText: "",
-  domain_profile: "generic_news",
-  input_mode: "urls" as "urls" | "texts" | "mixed",
-  extract_mode: "heuristic" as "heuristic" | "llm",
+  channel_id: null as number | null,
+  profile_id: null as number | null,
+  limit: 10,
   interval_seconds: 3600,
   is_active: true,
 });
 
-const form = ref({
-  source_type: "telegram" as "telegram" | "crawl4ai" | "viina",
-  channel: "",
-  limit: 50,
-  urlsText: "",
-  textsText: "",
-  domain_profile: "generic_news",
-  input_mode: "urls" as "urls" | "texts" | "mixed",
-  extract_mode: "heuristic" as "heuristic" | "llm",
-  interval_seconds: 3600,
+const readyProfiles = computed(() =>
+  profiles.value.filter((p) => {
+    if (p.status !== "ready") return false;
+    if (p.kind === "llm") return !!p.llm_profile;
+    return !!p.heuristic_profile;
+  }),
+);
+const activeChannels = computed(() => channels.value.filter((c) => c.is_active));
+
+const editChannelOptions = computed(() => channels.value);
+
+const editProfileOptions = computed(() => {
+  const list = [...readyProfiles.value];
+  const id = editForm.value.profile_id;
+  if (id != null && !list.some((p) => p.id === id)) {
+    const found = profiles.value.find((p) => p.id === id);
+    if (found) list.unshift(found);
+  }
+  return list;
 });
 
 let pollTimer: ReturnType<typeof setInterval> | null = null;
 
-function channelFromConfig(config: Record<string, unknown>): string {
-  return typeof config.channel === "string" ? config.channel : "";
+function isPairJob(job: ParseJob | null): boolean {
+  return !!job && job.profile_id != null && job.channel_id != null;
 }
 
 function limitFromConfig(config: Record<string, unknown>): number {
   const limit = config.limit;
-  return typeof limit === "number" ? limit : 50;
-}
-
-function urlsFromConfig(config: Record<string, unknown>): string[] {
-  return Array.isArray(config.urls)
-    ? config.urls.filter((u): u is string => typeof u === "string")
-    : [];
-}
-
-function textsFromConfig(config: Record<string, unknown>): string[] {
-  return Array.isArray(config.texts)
-    ? config.texts.filter((t): t is string => typeof t === "string")
-    : [];
-}
-
-function parseLines(text: string): string[] {
-  return text
-    .split(/\r?\n/)
-    .map((line) => line.trim())
-    .filter(Boolean);
+  return typeof limit === "number" ? limit : 10;
 }
 
 function sourceSummary(job: ParseJob): string {
-  const cfg = job.source_config;
-  const mode = cfg.extract_mode === "llm" ? " [LLM]" : "";
-  if (job.source_type === "telegram") {
-    return (channelFromConfig(cfg) || "—") + mode;
+  if (isPairJob(job)) {
+    const ch = job.channel_name || job.channel_handle || `#${job.channel_id}`;
+    const pr = job.profile_name || `#${job.profile_id}`;
+    return `${ch} × ${pr}`;
   }
-  const urls = urlsFromConfig(cfg);
-  if (urls.length) return (urls.length === 1 ? urls[0] : `${urls.length} URL`) + mode;
-  const texts = textsFromConfig(cfg);
-  if (texts.length) return `${texts.length} текст(ов)`;
-  return "—";
+  return `${job.source_type} (без связки)`;
 }
 
 function formatInterval(seconds: number): string {
@@ -99,53 +86,6 @@ function formatInterval(seconds: number): string {
   return `${Math.round(seconds / 86400)} д`;
 }
 
-function buildSourceConfig(
-  sourceType: string,
-  data: {
-    channel: string;
-    limit: number;
-    urlsText: string;
-    textsText: string;
-    domain_profile: string;
-    input_mode: "urls" | "texts" | "mixed";
-    extract_mode: "heuristic" | "llm";
-  },
-): Record<string, unknown> {
-  if (sourceType === "telegram") {
-    return {
-      channel: data.channel.trim(),
-      limit: data.limit,
-      extract_mode: data.extract_mode,
-    };
-  }
-  if (sourceType === "crawl4ai") {
-    return {
-      urls: parseLines(data.urlsText),
-      domain_profile: data.domain_profile.trim() || "generic_news",
-      extract_mode: data.extract_mode,
-    };
-  }
-  return {
-    urls: parseLines(data.urlsText),
-    texts: parseLines(data.textsText),
-    input_mode: data.input_mode,
-  };
-}
-
-const formHint = computed(() => {
-  if (form.value.source_type === "telegram") {
-    return form.value.extract_mode === "llm"
-      ? "LLM (DeepSeek): неструктурированные посты канала разбираются в locality/date/coords/description."
-      : "Активные Telegram-парсеры подхватываются real-time listener; batch забирает последние N постов (эвристики).";
-  }
-  if (form.value.source_type === "crawl4ai") {
-    return form.value.extract_mode === "llm"
-      ? "Crawl4AI + LLM (DeepSeek): страница → structured event. Нужен DEEPSEEK_API_KEY и cp-workers-web."
-      : "Crawl4AI обходит URL и нормализует страницы эвристиками. Обрабатывает cp-workers-web.";
-  }
-  return "VIINA-адаптер извлекает инциденты из новостных URL/текстов. Обрабатывает cp-workers-nlp.";
-});
-
 async function loadJobs() {
   try {
     jobs.value = await fetchJobs();
@@ -157,68 +97,51 @@ async function loadJobs() {
   }
 }
 
-async function handleSubmit() {
-  if (form.value.source_type === "telegram" && !form.value.channel.trim()) {
-    error.value = "Укажите канал Telegram";
-    return;
-  }
-  if (form.value.source_type === "crawl4ai" && !parseLines(form.value.urlsText).length) {
-    error.value = "Укажите хотя бы один URL";
-    return;
-  }
-  if (form.value.source_type === "viina") {
-    const urls = parseLines(form.value.urlsText);
-    const texts = parseLines(form.value.textsText);
-    if (form.value.input_mode === "urls" && !urls.length) {
-      error.value = "Укажите URL для VIINA";
-      return;
-    }
-    if (form.value.input_mode === "texts" && !texts.length) {
-      error.value = "Укажите тексты для VIINA";
-      return;
-    }
-    if (form.value.input_mode === "mixed" && !urls.length && !texts.length) {
-      error.value = "Укажите URL или тексты";
-      return;
-    }
+async function loadEntities() {
+  try {
+    const [chs, prs] = await Promise.all([fetchChannels(), fetchProfiles()]);
+    channels.value = chs;
+    profiles.value = prs;
+  } catch {
+    /* selects stay empty */
   }
+}
 
+async function handleCreatePair() {
+  if (!pairForm.value.channel_id || !pairForm.value.profile_id) {
+    error.value = "Выберите канал и профиль";
+    return;
+  }
   submitting.value = true;
   error.value = "";
   try {
     await createJob({
-      source_type: form.value.source_type,
-      source_config: buildSourceConfig(form.value.source_type, form.value),
-      interval_seconds: form.value.interval_seconds,
+      channel_id: pairForm.value.channel_id,
+      profile_id: pairForm.value.profile_id,
+      limit: pairForm.value.limit,
+      interval_seconds: pairForm.value.interval_seconds,
       is_active: true,
     });
-    form.value.channel = "";
-    form.value.urlsText = "";
-    form.value.textsText = "";
+    pairForm.value.channel_id = null;
+    pairForm.value.profile_id = null;
     await loadJobs();
   } catch (err) {
-    error.value = err instanceof Error ? err.message : "Не удалось создать задание";
+    error.value = err instanceof Error ? err.message : "Не удалось создать связку";
   } finally {
     submitting.value = false;
   }
 }
 
 function openEdit(job: ParseJob) {
+  if (!isPairJob(job)) {
+    error.value = "Редактируются только связки канал+профиль. Остальные задания можно удалить.";
+    return;
+  }
   editingJob.value = job;
   editForm.value = {
-    channel: channelFromConfig(job.source_config),
+    channel_id: job.channel_id ?? null,
+    profile_id: job.profile_id ?? null,
     limit: limitFromConfig(job.source_config),
-    urlsText: urlsFromConfig(job.source_config).join("\n"),
-    textsText: textsFromConfig(job.source_config).join("\n"),
-    domain_profile:
-      typeof job.source_config.domain_profile === "string"
-        ? job.source_config.domain_profile
-        : "generic_news",
-    input_mode:
-      job.source_config.input_mode === "texts" || job.source_config.input_mode === "mixed"
-        ? job.source_config.input_mode
-        : "urls",
-    extract_mode: job.source_config.extract_mode === "llm" ? "llm" : "heuristic",
     interval_seconds: job.interval_seconds,
     is_active: job.is_active,
   };
@@ -229,19 +152,18 @@ function closeEdit() {
 }
 
 async function handleSaveEdit() {
-  if (!editingJob.value) return;
-  const sourceType = editingJob.value.source_type;
-
-  if (sourceType === "telegram" && !editForm.value.channel.trim()) {
-    error.value = "Укажите канал Telegram";
+  if (!editingJob.value || !isPairJob(editingJob.value)) return;
+  if (!editForm.value.channel_id || !editForm.value.profile_id) {
+    error.value = "Выберите канал и профиль";
     return;
   }
-
   submitting.value = true;
   error.value = "";
   try {
     await updateJob(editingJob.value.id, {
-      source_config: buildSourceConfig(sourceType, editForm.value),
+      channel_id: editForm.value.channel_id,
+      profile_id: editForm.value.profile_id,
+      limit: editForm.value.limit,
       interval_seconds: editForm.value.interval_seconds,
       is_active: editForm.value.is_active,
     });
@@ -268,12 +190,30 @@ async function handleDelete(job: ParseJob) {
   }
 }
 
-async function handleRetry(jobId: number) {
+const STALE_JOB_MS = 15 * 60 * 1000;
+
+function isStaleJob(job: ParseJob): boolean {
+  if (job.status !== "queued" && job.status !== "running") return false;
+  if (!job.last_run_at) return job.status === "running";
+  return Date.now() - new Date(job.last_run_at).getTime() >= STALE_JOB_MS;
+}
+
+function canForceRun(job: ParseJob): boolean {
+  if (job.status !== "queued" && job.status !== "running") return true;
+  return isStaleJob(job);
+}
+
+async function handleForceRun(jobId: number) {
+  forcingJobId.value = jobId;
+  error.value = "";
   try {
     await retryJob(jobId);
     await loadJobs();
   } catch (err) {
-    error.value = err instanceof Error ? err.message : "Не удалось перезапустить задание";
+    error.value =
+      err instanceof Error ? err.message : "Не удалось запустить запрос постов";
+  } finally {
+    forcingJobId.value = null;
   }
 }
 
@@ -291,6 +231,7 @@ function formatDate(value: string | null): string {
 
 onMounted(() => {
   loadJobs();
+  loadEntities();
   pollTimer = setInterval(loadJobs, 5000);
 });
 
@@ -304,80 +245,43 @@ onUnmounted(() => {
     <h2 class="page-heading">Парсеры</h2>
 
     <section class="card">
-      <h3>Новый парсер</h3>
+      <h3>Новая связка (канал + профиль)</h3>
       <p class="hint">
-        {{ formHint }}
-        Дубликаты по <code>source_url</code> не записываются.
+        Создайте
+        <router-link to="/channels">канал</router-link>
+        и
+        <router-link to="/parser-profiles">профиль</router-link>
+        (heuristic или llm), затем свяжите их здесь. Flatten в Redis:
+        heuristic → <code>extract_mode=profile</code>;
+        llm → <code>extract_mode=llm</code> (batch; listener для llm пока fallback).
       </p>
-      <form class="admin-form" @submit.prevent="handleSubmit">
+      <form class="admin-form" @submit.prevent="handleCreatePair">
         <div class="form-row">
           <label>
-            Тип источника
-            <select v-model="form.source_type">
-              <option v-for="opt in SOURCE_TYPES" :key="opt.value" :value="opt.value">
-                {{ opt.label }}
+            Канал
+            <select v-model.number="pairForm.channel_id" required>
+              <option :value="null" disabled>Выберите…</option>
+              <option v-for="ch in activeChannels" :key="ch.id" :value="ch.id">
+                {{ ch.name }} (@{{ ch.channel }})
               </option>
             </select>
           </label>
-
-          <template v-if="form.source_type === 'telegram'">
-            <label>
-              Канал
-              <input v-model="form.channel" type="text" placeholder="creamy_caprice" required />
-            </label>
-            <label>
-              Лимит
-              <input v-model.number="form.limit" type="number" min="1" max="1000" />
-            </label>
-            <label>
-              Извлечение
-              <select v-model="form.extract_mode">
-                <option value="heuristic">Эвристики (структурированные посты)</option>
-                <option value="llm">LLM DeepSeek (неструктурированные)</option>
-              </select>
-            </label>
-          </template>
-
-          <template v-else-if="form.source_type === 'crawl4ai'">
-            <label class="span-2">
-              URL (по одному в строке)
-              <textarea v-model="form.urlsText" rows="3" placeholder="https://example.com/news/…" required />
-            </label>
-            <label>
-              Domain profile
-              <input v-model="form.domain_profile" type="text" placeholder="generic_news" />
-            </label>
-            <label>
-              Извлечение
-              <select v-model="form.extract_mode">
-                <option value="heuristic">Эвристики (regex)</option>
-                <option value="llm">LLM DeepSeek (Crawl4AI)</option>
-              </select>
-            </label>
-          </template>
-
-          <template v-else>
-            <label>
-              Режим ввода
-              <select v-model="form.input_mode">
-                <option value="urls">URLs</option>
-                <option value="texts">Texts</option>
-                <option value="mixed">Mixed</option>
-              </select>
-            </label>
-            <label v-if="form.input_mode !== 'texts'" class="span-2">
-              URL (по одному в строке)
-              <textarea v-model="form.urlsText" rows="3" placeholder="https://example.com/article…" />
-            </label>
-            <label v-if="form.input_mode !== 'urls'" class="span-2">
-              Тексты (по одному блоку в строке)
-              <textarea v-model="form.textsText" rows="3" placeholder="Текст статьи…" />
-            </label>
-          </template>
-
+          <label>
+            Профиль
+            <select v-model.number="pairForm.profile_id" required>
+              <option :value="null" disabled>Выберите…</option>
+              <option v-for="p in readyProfiles" :key="p.id" :value="p.id">
+                {{ p.name }} (#{{ p.id }}, {{ p.kind === "llm" ? "llm" : "heuristic" }})
+              </option>
+            </select>
+          </label>
+          <label>
+            Лимит batch
+            <input v-model.number="pairForm.limit" type="number" min="1" max="1000" />
+          </label>
           <label>
             Интервал
-            <select v-model.number="form.interval_seconds">
+            <select v-model.number="pairForm.interval_seconds">
               <option v-for="opt in INTERVAL_OPTIONS" :key="opt.value" :value="opt.value">
                 {{ opt.label }}
               </option>
@@ -385,16 +289,24 @@ onUnmounted(() => {
           </label>
         </div>
         <div class="form-actions">
-          <button type="submit" class="btn btn-primary" :disabled="submitting">
-            {{ submitting ? "Создание..." : "Создать парсер" }}
+          <button
+            type="submit"
+            class="btn btn-primary"
+            :disabled="submitting || !pairForm.channel_id || !pairForm.profile_id"
+          >
+            {{ submitting ? "Создание..." : "Создать связку" }}
           </button>
         </div>
       </form>
+      <p v-if="!activeChannels.length || !readyProfiles.length" class="hint">
+        Нет активных каналов или готовых профилей —
+        сначала добавьте их на страницах Каналы / Профили.
+      </p>
       <p v-if="error" class="form-error">{{ error }}</p>
     </section>
 
     <section class="card">
-      <h3>Парсеры <span v-if="loading" class="muted">(загрузка...)</span></h3>
+      <h3>Связки <span v-if="loading" class="muted">(загрузка...)</span></h3>
       <div class="table-wrap">
         <table class="admin-table">
           <thead>
@@ -426,19 +338,26 @@ onUnmounted(() => {
               <td>{{ formatDate(job.last_run_at) }}</td>
               <td class="error-cell">{{ job.last_error ?? "—" }}</td>
               <td class="actions-cell">
-                <button class="btn btn-sm" type="button" @click="openEdit(job)">Изменить</button>
                 <button
-                  v-if="job.status === 'failed' || job.status === 'error'"
+                  v-if="isPairJob(job)"
                   class="btn btn-sm"
                   type="button"
-                  @click="handleRetry(job.id)"
+                  @click="openEdit(job)"
                 >
-                  Повторить
+                  Изменить
+                </button>
+                <button
+                  class="btn btn-sm"
+                  type="button"
+                  :disabled="!canForceRun(job) || forcingJobId === job.id"
+                  @click="handleForceRun(job.id)"
+                >
+                  {{ forcingJobId === job.id ? "Запуск..." : "Запустить" }}
                 </button>
                 <button
                   class="btn btn-sm btn-danger"
                   type="button"
-                  :disabled="job.status === 'running'"
+                  :disabled="job.status === 'running' && !isStaleJob(job)"
                   @click="handleDelete(job)"
                 >
                   Удалить
@@ -452,61 +371,29 @@ onUnmounted(() => {
 
     <div v-if="editingJob" class="modal-backdrop" @click.self="closeEdit">
       <div class="modal card">
-        <h3>Редактировать парсер #{{ editingJob.id }} ({{ editingJob.source_type }})</h3>
+        <h3>Редактировать связку #{{ editingJob.id }}</h3>
         <form class="admin-form" @submit.prevent="handleSaveEdit">
           <div class="form-row">
-            <template v-if="editingJob.source_type === 'telegram'">
-              <label>
-                Канал
-                <input v-model="editForm.channel" type="text" required />
-              </label>
-              <label>
-                Лимит
-                <input v-model.number="editForm.limit" type="number" min="1" max="1000" />
-              </label>
-              <label>
-                Извлечение
-                <select v-model="editForm.extract_mode">
-                  <option value="heuristic">Эвристики</option>
-                  <option value="llm">LLM DeepSeek</option>
-                </select>
-              </label>
-            </template>
-            <template v-else-if="editingJob.source_type === 'crawl4ai'">
-              <label class="span-2">
-                URL (по одному в строке)
-                <textarea v-model="editForm.urlsText" rows="3" required />
-              </label>
-              <label>
-                Domain profile
-                <input v-model="editForm.domain_profile" type="text" />
-              </label>
-              <label>
-                Извлечение
-                <select v-model="editForm.extract_mode">
-                  <option value="heuristic">Эвристики</option>
-                  <option value="llm">LLM DeepSeek</option>
-                </select>
-              </label>
-            </template>
-            <template v-else>
-              <label>
-                Режим ввода
-                <select v-model="editForm.input_mode">
-                  <option value="urls">URLs</option>
-                  <option value="texts">Texts</option>
-                  <option value="mixed">Mixed</option>
-                </select>
-              </label>
-              <label v-if="editForm.input_mode !== 'texts'" class="span-2">
-                URL
-                <textarea v-model="editForm.urlsText" rows="3" />
-              </label>
-              <label v-if="editForm.input_mode !== 'urls'" class="span-2">
-                Тексты
-                <textarea v-model="editForm.textsText" rows="3" />
-              </label>
-            </template>
+            <label>
+              Канал
+              <select v-model.number="editForm.channel_id" required>
+                <option v-for="ch in editChannelOptions" :key="ch.id" :value="ch.id">
+                  {{ ch.name }} (@{{ ch.channel }}){{ ch.is_active ? "" : " — выкл." }}
+                </option>
+              </select>
+            </label>
+            <label>
+              Профиль
+              <select v-model.number="editForm.profile_id" required>
+                <option v-for="p in editProfileOptions" :key="p.id" :value="p.id">
+                  {{ p.name }} (#{{ p.id }})
+                </option>
+              </select>
+            </label>
+            <label>
+              Лимит batch
+              <input v-model.number="editForm.limit" type="number" min="1" max="1000" />
+            </label>
             <label>
               Интервал
               <select v-model.number="editForm.interval_seconds">
@@ -540,19 +427,6 @@ onUnmounted(() => {
   line-height: 1.5;
 }
 
-.form-row .span-2 {
-  grid-column: span 2;
-}
-
-.form-row textarea {
-  width: 100%;
-  font: inherit;
-  padding: 0.4rem 0.5rem;
-  border: 1px solid #d1d5db;
-  border-radius: 4px;
-  resize: vertical;
-}
-
 .source-cell {
   max-width: 280px;
   overflow: hidden;
@@ -568,11 +442,6 @@ onUnmounted(() => {
   margin-right: 0.25rem;
 }
 
-.btn-danger {
-  color: #b91c1c;
-  border-color: #fecaca;
-}
-
 .modal-backdrop {
   position: fixed;
   inset: 0;
@@ -586,8 +455,6 @@ onUnmounted(() => {
 .modal {
   width: min(560px, 92vw);
   margin: 0;
-  max-height: 90vh;
-  overflow: auto;
 }
 
 .checkbox-row {
@@ -596,4 +463,13 @@ onUnmounted(() => {
   gap: 0.5rem;
   margin-top: 1.5rem;
 }
+
+.error-cell {
+  max-width: 200px;
+  overflow: hidden;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+  font-size: 0.8rem;
+  color: #b91c1c;
+}
 </style>

centers/analytics/frontend/src/views/VpnView.vue

@@ -0,0 +1,278 @@
+<script setup lang="ts">
+import { computed, onMounted, ref } from "vue";
+
+import { fetchVpnSettings, updateVpnSettings } from "../api/admin";
+import type { VpnMode, VpnSettings } from "../types/admin";
+
+const loading = ref(true);
+const submitting = ref(false);
+const error = ref("");
+const success = ref("");
+const settings = ref<VpnSettings | null>(null);
+
+const form = ref({
+  enabled: false,
+  mode: "subscription" as VpnMode,
+  subscription_url: "",
+  clear_subscription_url: false,
+  subscription_interval_seconds: 3600,
+  host: "",
+  port: 1080 as number | null,
+  username: "",
+  password: "",
+  clear_password: false,
+  proxied_source_types: [] as string[],
+});
+
+const sourceLabels: Record<string, string> = {
+  telegram: "Telegram",
+  crawl4ai: "Crawl4AI (web)",
+  viina: "VIINA (nlp)",
+};
+
+const isSubscription = computed(() => form.value.mode === "subscription");
+const isDirectProxy = computed(
+  () => form.value.mode === "socks5" || form.value.mode === "http",
+);
+
+function applySettings(data: VpnSettings) {
+  settings.value = data;
+  form.value = {
+    enabled: data.enabled,
+    mode: data.mode,
+    subscription_url: "",
+    clear_subscription_url: false,
+    subscription_interval_seconds: data.subscription_interval_seconds,
+    host: data.host ?? "",
+    port: data.port ?? 1080,
+    username: data.username ?? "",
+    password: "",
+    clear_password: false,
+    proxied_source_types: [...data.proxied_source_types],
+  };
+}
+
+async function load() {
+  loading.value = true;
+  error.value = "";
+  success.value = "";
+  try {
+    applySettings(await fetchVpnSettings());
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось загрузить VPN";
+  } finally {
+    loading.value = false;
+  }
+}
+
+function toggleSource(sourceType: string, checked: boolean) {
+  const set = new Set(form.value.proxied_source_types);
+  if (checked) set.add(sourceType);
+  else set.delete(sourceType);
+  form.value.proxied_source_types = [...set];
+}
+
+async function handleSave() {
+  submitting.value = true;
+  error.value = "";
+  success.value = "";
+  try {
+    const payload: Parameters<typeof updateVpnSettings>[0] = {
+      enabled: form.value.enabled,
+      mode: form.value.mode,
+      subscription_interval_seconds: form.value.subscription_interval_seconds,
+      host: form.value.host.trim() || null,
+      port: form.value.port,
+      username: form.value.username.trim() || null,
+      proxied_source_types: form.value.proxied_source_types,
+    };
+
+    if (form.value.clear_subscription_url) {
+      payload.clear_subscription_url = true;
+    } else if (form.value.subscription_url.trim()) {
+      payload.subscription_url = form.value.subscription_url.trim();
+    }
+
+    if (form.value.clear_password) {
+      payload.clear_password = true;
+    } else if (form.value.password) {
+      payload.password = form.value.password;
+    }
+
+    applySettings(await updateVpnSettings(payload));
+    success.value = "Настройки VPN сохранены";
+  } catch (err) {
+    error.value = err instanceof Error ? err.message : "Не удалось сохранить";
+  } finally {
+    submitting.value = false;
+  }
+}
+
+onMounted(load);
+</script>
+
+<template>
+  <div class="page">
+    <h2 class="page-heading">VPN</h2>
+    <p class="intro">
+      Проксирование трафика выбранных источников парсинга. Режим
+      <strong>subscription</strong> поднимает SOCKS через сервис
+      <code>cp-vpn</code> (mihomo). Режимы SOCKS5/HTTP задают адрес напрямую для
+      воркеров. Без включённого VPN остаётся fallback из
+      <code>TELEGRAM_PROXY_*</code> в <code>.env</code>.
+    </p>
+
+    <section class="card">
+      <h3>Настройки <span v-if="loading" class="muted">(загрузка...)</span></h3>
+      <form v-if="settings" class="admin-form" @submit.prevent="handleSave">
+        <label class="checkbox-row">
+          <input v-model="form.enabled" type="checkbox" />
+          <span>Прокси включён</span>
+        </label>
+
+        <div class="form-row vpn-row">
+          <label>
+            Способ проксирования
+            <select v-model="form.mode">
+              <option value="subscription">Subscription URL</option>
+              <option value="socks5">SOCKS5</option>
+              <option value="http">HTTP</option>
+            </select>
+          </label>
+        </div>
+
+        <template v-if="isSubscription">
+          <label>
+            Адрес подписки (subscription URL)
+            <input
+              v-model="form.subscription_url"
+              type="url"
+              placeholder="https://sub.example.com/…"
+              autocomplete="off"
+            />
+          </label>
+          <p v-if="settings.subscription_url_set" class="hint">
+            Сохранён:
+            <code>{{ settings.subscription_url_hint }}</code>
+            — оставьте поле пустым, чтобы не менять.
+          </p>
+          <label class="checkbox-row">
+            <input v-model="form.clear_subscription_url" type="checkbox" />
+            <span>Очистить сохранённый URL</span>
+          </label>
+          <label>
+            Интервал обновления подписки (сек)
+            <input
+              v-model.number="form.subscription_interval_seconds"
+              type="number"
+              min="60"
+              max="86400"
+            />
+          </label>
+        </template>
+
+        <template v-if="isDirectProxy">
+          <div class="form-row">
+            <label>
+              Host
+              <input v-model="form.host" type="text" placeholder="127.0.0.1" />
+            </label>
+            <label>
+              Port
+              <input v-model.number="form.port" type="number" min="1" max="65535" />
+            </label>
+            <label>
+              Username
+              <input v-model="form.username" type="text" autocomplete="off" />
+            </label>
+            <label>
+              Password
+              <input
+                v-model="form.password"
+                type="password"
+                autocomplete="new-password"
+                :placeholder="settings.password_set ? '•••• (оставьте пустым)' : ''"
+              />
+            </label>
+          </div>
+          <label v-if="settings.password_set" class="checkbox-row">
+            <input v-model="form.clear_password" type="checkbox" />
+            <span>Очистить пароль</span>
+          </label>
+        </template>
+
+        <fieldset class="sources">
+          <legend>Источники через прокси</legend>
+          <label
+            v-for="st in settings.available_source_types"
+            :key="st"
+            class="checkbox-row"
+          >
+            <input
+              type="checkbox"
+              :checked="form.proxied_source_types.includes(st)"
+              @change="
+                toggleSource(st, ($event.target as HTMLInputElement).checked)
+              "
+            />
+            <span>{{ sourceLabels[st] || st }} <code>({{ st }})</code></span>
+          </label>
+        </fieldset>
+
+        <div class="form-actions">
+          <button type="submit" class="btn btn-primary" :disabled="submitting || loading">
+            {{ submitting ? "Сохранение..." : "Сохранить" }}
+          </button>
+        </div>
+        <p v-if="error" class="form-error">{{ error }}</p>
+        <p v-if="success" class="form-success">{{ success }}</p>
+      </form>
+      <p v-else-if="error" class="form-error">{{ error }}</p>
+    </section>
+  </div>
+</template>
+
+<style scoped>
+.intro {
+  margin: 0 0 1rem;
+  font-size: 0.9rem;
+  color: #555;
+  line-height: 1.5;
+}
+
+.checkbox-row {
+  display: flex;
+  align-items: center;
+  gap: 0.5rem;
+  margin: 0.5rem 0 1rem;
+}
+
+.vpn-row {
+  margin-bottom: 0.75rem;
+}
+
+.hint {
+  margin: -0.5rem 0 0.75rem;
+  font-size: 0.85rem;
+  color: #666;
+}
+
+.sources {
+  border: 1px solid #e5e5e5;
+  border-radius: 6px;
+  padding: 0.75rem 1rem 0.25rem;
+  margin: 0.5rem 0 1rem;
+}
+
+.sources legend {
+  padding: 0 0.35rem;
+  font-size: 0.9rem;
+  color: #444;
+}
+
+.form-success {
+  margin: 0.75rem 0 0;
+  color: #1a7f37;
+  font-size: 0.9rem;
+}
+</style>

centers/parsing/ARCHITECTURE.md

@@ -54,14 +54,42 @@ async def run(job_id, source_config, *, ctx) -> tuple[list[dict], str | None]
 - `source_type` события = тип адаптера;
 - `source_config` валидируется схемами из `contracts/sources.py`.
 
+### Режимы извлечения (Telegram)
+
+Цель всегда одна: фиксированные поля `IngestEventItem` / Event в ЦА (`title`, `description`, `locality`, `coords`→lat/lng, `event_date`, `topic`, …). Кастомные пользовательские таблицы — roadmap.
+
+| `extract_mode` | Что делает | Откуда в UI |
+|----------------|------------|-------------|
+| `heuristic` | Legacy regex-парсер постов | raw job API (не пара) |
+| `llm` | DeepSeek **на каждый** пост (batch) | профиль `kind=llm` → pair flatten |
+| `profile` | Статичные правила `HeuristicProfile` | профиль `kind=heuristic` → pair flatten |
+
 ### LLM-режим (`extract_mode: llm`)
 
 Для неструктурированных Telegram-постов и Crawl4AI:
 
-- ключ `DEEPSEEK_API_KEY` в `.env` (воркеры `cp-workers` / `cp-workers-web`);
-- Telegram: текст поста → DeepSeek JSON → `IngestEventItem`;
+- ключ `DEEPSEEK_API_KEY` в `.env` (воркеры `cp-workers` / `cp-workers-web`; preview в `ca-api`);
+- Telegram batch: текст → DeepSeek JSON → один или несколько `IngestEventItem` (`workers/llm_extract.py`);
+- опционально `extract_schema`, `instruction`, `required_fields` (post-extract gate поверх `is_event`);
+- **`multi_event`** (opt-in в `llm_profile` / `TelegramSourceConfig`): модель возвращает массив `events`; каждое событие — отдельный ingest. URL: один event → `post.url`; несколько → `post.url#e1`, `#e2`, … (дедуп ЦА по `source_url`);
+- heuristic / profile по-прежнему **1 пост → 1 событие**;
 - Crawl4AI: страница → `LLMExtractionStrategy` (DeepSeek) с fallback на тот же DeepSeek по markdown;
-- в UI «Парсеры»: поле **Извлечение** = LLM DeepSeek.
+- **listener:** для `llm` пока fallback на heuristic (LLM — batch-only by design).
+
+Reusable LLM-профиль в ЦА: `ParserProfile.kind=llm` + JSON `llm_profile` (`contracts/llm_profile.py`). При enqueue flatten → `extract_mode=llm` + schema/instruction/required_fields/`multi_event`.
+
+### Profile-режим (`extract_mode: profile`)
+
+Статичный парсер из конструктора ЦА:
+
+- `heuristic_profile` в `source_config` (схема `contracts/heuristic_profile.py`);
+- интерпретатор: `workers/heuristic_profile.py` (те же правила, что preview в ЦА);
+- `required_fields`: пост без заполненных обязательных полей не ингестится (batch + listener);
+- генерация правил — один раз в админке (`/admin/parser-profiles/generate`); runtime **без** LLM.
+
+Reusable heuristic-профиль: `ParserProfile.kind=heuristic` + `heuristic_profile`. Pair на «Парсеры» → flatten `extract_mode=profile`.
+
+UI «Профили»: выбор `kind` (heuristic | llm). Связка канал+профиль на «Парсеры».
 
 ```mermaid
 flowchart LR
@@ -118,7 +146,10 @@ Legacy-ключ `cp:jobs` по-прежнему дренируется telegram-
 
 ## Telegram real-time
 
-`TelegramListener` без изменений: подписки из ЦА, ingest с `listener: true` (статус `ParseJob` не трогается).
+`TelegramListener`: подписки из ЦА, ingest с `listener: true` (статус `ParseJob` не трогается).
+
+- `extract_mode=profile` — те же heuristic-правила, что batch;
+- `extract_mode=llm` — **не** вызывает DeepSeek; fallback на legacy heuristic (LLM только в batch).
 
 ## Переменные окружения
 

centers/parsing/vpn/Dockerfile

@@ -0,0 +1,31 @@
+FROM python:3.12-alpine
+
+ARG MIHOMO_VERSION=1.19.3
+ARG TARGETARCH
+
+RUN apk add --no-cache ca-certificates curl gzip \
+    && arch="${TARGETARCH:-amd64}" \
+    && case "$arch" in \
+         amd64|x86_64) mihomo_arch=amd64 ;; \
+         arm64|aarch64) mihomo_arch=arm64 ;; \
+         *) mihomo_arch=amd64 ;; \
+       esac \
+    && curl -fsSL \
+         "https://github.com/MetaCubeX/mihomo/releases/download/v${MIHOMO_VERSION}/mihomo-linux-${mihomo_arch}-compatible-v${MIHOMO_VERSION}.gz" \
+         -o /tmp/mihomo.gz \
+    && gunzip -c /tmp/mihomo.gz > /usr/local/bin/mihomo \
+    && chmod +x /usr/local/bin/mihomo \
+    && rm -f /tmp/mihomo.gz \
+    && mkdir -p /etc/mihomo/providers
+
+WORKDIR /app
+COPY controller.py /app/controller.py
+
+ENV MIHOMO_BIN=/usr/local/bin/mihomo \
+    MIHOMO_CONFIG_DIR=/etc/mihomo \
+    VPN_SOCKS_PORT=1080 \
+    VPN_POLL_SECONDS=15
+
+EXPOSE 1080
+
+CMD ["python", "-u", "/app/controller.py"]

centers/parsing/vpn/controller.py

@@ -0,0 +1,204 @@
+"""Poll CA /internal/vpn and keep mihomo config in sync for subscription mode."""
+
+from __future__ import annotations
+
+import json
+import logging
+import os
+import signal
+import subprocess
+import sys
+import time
+from pathlib import Path
+from urllib.error import HTTPError, URLError
+from urllib.request import Request, urlopen
+
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s %(levelname)s %(message)s",
+)
+logger = logging.getLogger("cp-vpn")
+
+CA_API_URL = os.getenv("CA_API_URL", "http://ca-api:8000").rstrip("/")
+INTERNAL_TOKEN = os.getenv("INTERNAL_TOKEN", "dev-internal-token")
+POLL_SECONDS = int(os.getenv("VPN_POLL_SECONDS", "15"))
+CONFIG_DIR = Path(os.getenv("MIHOMO_CONFIG_DIR", "/etc/mihomo"))
+CONFIG_PATH = CONFIG_DIR / "config.yaml"
+MIHOMO_BIN = os.getenv("MIHOMO_BIN", "/usr/local/bin/mihomo")
+MIXED_PORT = int(os.getenv("VPN_SOCKS_PORT", "1080"))
+CONTROLLER = "127.0.0.1:9090"
+
+_mihomo_proc: subprocess.Popen | None = None
+_last_fingerprint: str | None = None
+
+
+def fetch_vpn() -> dict:
+    req = Request(
+        f"{CA_API_URL}/internal/vpn",
+        headers={"X-Internal-Token": INTERNAL_TOKEN},
+        method="GET",
+    )
+    with urlopen(req, timeout=15) as resp:
+        return json.loads(resp.read().decode("utf-8"))
+
+
+def fingerprint(cfg: dict) -> str:
+    return json.dumps(
+        {
+            "enabled": cfg.get("enabled"),
+            "mode": cfg.get("mode"),
+            "subscription_url": cfg.get("subscription_url"),
+            "subscription_interval_seconds": cfg.get("subscription_interval_seconds"),
+        },
+        sort_keys=True,
+    )
+
+
+def write_direct_config() -> None:
+    CONFIG_DIR.mkdir(parents=True, exist_ok=True)
+    (CONFIG_DIR / "providers").mkdir(parents=True, exist_ok=True)
+    CONFIG_PATH.write_text(
+        f"""# managed by cp-vpn controller — DIRECT (subscription inactive)
+mixed-port: {MIXED_PORT}
+allow-lan: true
+bind-address: "*"
+mode: direct
+log-level: warning
+external-controller: {CONTROLLER}
+""",
+        encoding="utf-8",
+    )
+
+
+def write_subscription_config(url: str, interval: int) -> None:
+    CONFIG_DIR.mkdir(parents=True, exist_ok=True)
+    (CONFIG_DIR / "providers").mkdir(parents=True, exist_ok=True)
+    # YAML: quote URL; escape double quotes in URL if any
+    safe_url = url.replace("\\", "\\\\").replace('"', '\\"')
+    CONFIG_PATH.write_text(
+        f"""# managed by cp-vpn controller — subscription
+mixed-port: {MIXED_PORT}
+allow-lan: true
+bind-address: "*"
+mode: rule
+log-level: info
+external-controller: {CONTROLLER}
+proxy-providers:
+  sub:
+    type: http
+    url: "{safe_url}"
+    interval: {interval}
+    path: ./providers/sub.yaml
+    health-check:
+      enable: true
+      url: https://www.gstatic.com/generate_204
+      interval: 600
+proxy-groups:
+  - name: PROXY
+    type: select
+    use:
+      - sub
+rules:
+  - MATCH,PROXY
+""",
+        encoding="utf-8",
+    )
+
+
+def start_mihomo() -> None:
+    global _mihomo_proc
+    if _mihomo_proc is not None and _mihomo_proc.poll() is None:
+        return
+    logger.info("Starting mihomo (%s)", MIHOMO_BIN)
+    _mihomo_proc = subprocess.Popen(
+        [MIHOMO_BIN, "-d", str(CONFIG_DIR)],
+        stdout=sys.stdout,
+        stderr=sys.stderr,
+    )
+
+
+def stop_mihomo() -> None:
+    global _mihomo_proc
+    if _mihomo_proc is None:
+        return
+    if _mihomo_proc.poll() is None:
+        logger.info("Stopping mihomo")
+        _mihomo_proc.send_signal(signal.SIGTERM)
+        try:
+            _mihomo_proc.wait(timeout=10)
+        except subprocess.TimeoutExpired:
+            _mihomo_proc.kill()
+    _mihomo_proc = None
+
+
+def reload_mihomo() -> None:
+    """Force reload via external controller; fall back to process restart."""
+    body = json.dumps({"path": str(CONFIG_PATH)}).encode("utf-8")
+    req = Request(
+        f"http://{CONTROLLER}/configs?force=true",
+        data=body,
+        headers={"Content-Type": "application/json"},
+        method="PUT",
+    )
+    try:
+        with urlopen(req, timeout=10) as resp:
+            resp.read()
+        logger.info("Mihomo config reloaded")
+        return
+    except (HTTPError, URLError, OSError) as exc:
+        logger.warning("Reload via API failed (%s); restarting mihomo", exc)
+    stop_mihomo()
+    start_mihomo()
+
+
+def apply_config(cfg: dict) -> None:
+    global _last_fingerprint
+    fp = fingerprint(cfg)
+    if fp == _last_fingerprint and _mihomo_proc is not None and _mihomo_proc.poll() is None:
+        return
+
+    enabled = bool(cfg.get("enabled"))
+    mode = (cfg.get("mode") or "").strip()
+    url = (cfg.get("subscription_url") or "").strip()
+    interval = int(cfg.get("subscription_interval_seconds") or 3600)
+
+    if enabled and mode == "subscription" and url:
+        logger.info("Applying subscription config (interval=%ss)", interval)
+        write_subscription_config(url, interval)
+    else:
+        logger.info("Applying DIRECT config (enabled=%s mode=%s)", enabled, mode)
+        write_direct_config()
+
+    already_running = _mihomo_proc is not None and _mihomo_proc.poll() is None
+    start_mihomo()
+    if already_running or _last_fingerprint is not None:
+        time.sleep(1)
+        reload_mihomo()
+    _last_fingerprint = fp
+
+
+def main() -> None:
+    global _mihomo_proc
+    logger.info(
+        "cp-vpn controller started (poll=%ss api=%s)",
+        POLL_SECONDS,
+        CA_API_URL,
+    )
+    write_direct_config()
+    start_mihomo()
+
+    while True:
+        try:
+            cfg = fetch_vpn()
+            apply_config(cfg)
+        except Exception:
+            logger.exception("Failed to sync VPN config")
+        if _mihomo_proc is not None and _mihomo_proc.poll() is not None:
+            logger.warning("mihomo exited with code %s; restarting", _mihomo_proc.returncode)
+            _mihomo_proc = None
+            start_mihomo()
+        time.sleep(POLL_SECONDS)
+
+
+if __name__ == "__main__":
+    main()

centers/parsing/workers/worker.py

@@ -179,10 +179,31 @@ async def run_with_listener() -> None:
         await worker_loop(ctx=WorkerContext())
         return
 
+    from workers.sources.telegram_client import TelegramAuthError, TelegramConfigError
     from workers.sources.telegram_listener import TelegramListener
     from workers.sources.telegram_session import close_shared_client, get_shared_client
 
-    client = await get_shared_client()
+    while True:
+        try:
+            client = await get_shared_client()
+            break
+        except (TelegramAuthError, TelegramConfigError, ConnectionError, OSError) as exc:
+            logger.error(
+                "Telegram connect failed (%s); retry in 60s (batch without TG)",
+                exc,
+            )
+            try:
+                await asyncio.wait_for(
+                    worker_loop(ctx=WorkerContext(tg_client=None)),
+                    timeout=60,
+                )
+            except asyncio.TimeoutError:
+                pass
+            continue
+        except Exception:
+            logger.exception("Unexpected Telegram connect error; retry in 60s")
+            await asyncio.sleep(60)
+
     listener = TelegramListener(client)
     ctx = WorkerContext(tg_client=client)
     worker_task = asyncio.create_task(worker_loop(ctx=ctx))

centers/parsing/workers/workers/adapters/crawl4ai_adapter.py

@@ -57,7 +57,25 @@ class Crawl4AIAdapter:
         events: list[dict] = []
         errors: list[str] = []
 
-        async with AsyncWebCrawler(verbose=False) as crawler:
+        crawler_kwargs: dict[str, Any] = {"verbose": False}
+        try:
+            from workers.sources.vpn_config import http_proxy_url
+
+            proxy = http_proxy_url("crawl4ai")
+            if proxy:
+                try:
+                    from crawl4ai import BrowserConfig  # type: ignore
+
+                    crawler_kwargs["config"] = BrowserConfig(proxy=proxy)
+                    logger.info("Crawl4AI via proxy %s", proxy.split("@")[-1])
+                except Exception:
+                    logger.exception(
+                        "Could not apply VPN proxy to Crawl4AI; continuing without"
+                    )
+        except Exception:
+            logger.exception("VPN config lookup failed for crawl4ai")
+
+        async with AsyncWebCrawler(**crawler_kwargs) as crawler:
             for url in cfg.urls:
                 try:
                     if cfg.extract_mode == "llm":

centers/parsing/workers/workers/adapters/telegram.py

@@ -7,12 +7,15 @@ import logging
 from contracts.sources import TelegramSourceConfig
 from workers.adapters.base import WorkerContext
 from workers.converter import event_record_to_ingest
+from workers.heuristic_profile import extract_with_profile
 from workers.llm_extract import (
     DEFAULT_EXTRACT_SCHEMA,
     DEFAULT_INSTRUCTION,
-    extract_event_fields,
+    event_source_url,
+    extract_event_list,
     fields_to_ingest,
     llm_enabled,
+    match_llm_required,
 )
 from workers.parsers.telegram_events import parse_event_posts
 from workers.sources.telegram_client import (
@@ -57,43 +60,87 @@ class TelegramAdapter:
                 return [], "extract_mode=llm requires DEEPSEEK_API_KEY in worker env"
             return await _extract_posts_with_llm(posts, cfg)
 
+        if cfg.extract_mode == "profile":
+            return _extract_posts_with_profile(posts, cfg)
+
         records = parse_event_posts(posts)
         events = [event_record_to_ingest(r) for r in records]
         return events, None
 
 
+def _extract_posts_with_profile(posts, cfg: TelegramSourceConfig) -> tuple[list[dict], str | None]:
+    profile = cfg.heuristic_profile or {}
+    events: list[dict] = []
+    for post in posts:
+        text = (post.text or "").strip()
+        if not text:
+            continue
+        event = extract_with_profile(
+            text,
+            profile,
+            source_url=post.url,
+            source_type="telegram",
+            extra_metadata={
+                "channel": post.channel,
+                "message_id": post.id,
+                "post_date": post.date.isoformat() if post.date else None,
+            },
+        )
+        if event is None:
+            logger.debug("Profile skip %s (required fields missing)", post.url)
+            continue
+        events.append(event)
+    return events, None
+
+
 async def _extract_posts_with_llm(posts, cfg: TelegramSourceConfig) -> tuple[list[dict], str | None]:
     schema = cfg.extract_schema or DEFAULT_EXTRACT_SCHEMA
     instruction = cfg.instruction or DEFAULT_INSTRUCTION
+    multi_event = bool(cfg.multi_event)
     events: list[dict] = []
     errors: list[str] = []
+    required = list(cfg.required_fields or [])
 
     for post in posts:
         text = (post.text or "").strip()
         if not text:
             continue
         try:
-            fields = await extract_event_fields(
+            field_list = await extract_event_list(
                 text,
                 extract_schema=schema,
                 instruction=instruction,
+                multi_event=multi_event,
             )
-            if not fields.get("is_event", True):
+            if not field_list:
                 continue
-            events.append(
-                fields_to_ingest(
-                    source_type="telegram",
-                    source_url=post.url,
-                    raw_text=text,
-                    fields=fields,
-                    domain_profile="telegram_llm",
-                    extra_metadata={
-                        "channel": post.channel,
-                        "message_id": post.id,
-                        "post_date": post.date.isoformat() if post.date else None,
-                    },
+            total = len(field_list)
+            for idx, fields in enumerate(field_list):
+                if not match_llm_required(fields, required):
+                    logger.debug(
+                        "LLM skip %s event %s/%s (required fields missing)",
+                        post.url,
+                        idx + 1,
+                        total,
+                    )
+                    continue
+                events.append(
+                    fields_to_ingest(
+                        source_type="telegram",
+                        source_url=event_source_url(post.url, idx, total),
+                        raw_text=text,
+                        fields=fields,
+                        domain_profile="telegram_llm",
+                        extra_metadata={
+                            "channel": post.channel,
+                            "message_id": post.id,
+                            "post_date": post.date.isoformat() if post.date else None,
+                            "event_index": idx + 1,
+                            "event_count": total,
+                            "multi_event": multi_event and total > 1,
+                        },
+                    )
                 )
-            )
         except Exception as exc:
             logger.exception("LLM extract failed for %s", post.url)
             errors.append(f"{post.url}: {exc}")

centers/parsing/workers/workers/adapters/viina.py

@@ -70,7 +70,14 @@ class ViinaAdapter:
 
 
 async def _fetch_article_text(url: str) -> str:
-    async with httpx.AsyncClient(timeout=60.0, follow_redirects=True) as client:
+    from workers.sources.vpn_config import http_proxy_url
+
+    proxy = http_proxy_url("viina")
+    kwargs: dict = {"timeout": 60.0, "follow_redirects": True}
+    if proxy:
+        kwargs["proxy"] = proxy
+        logger.info("VIINA fetch via proxy %s", proxy.split("@")[-1])
+    async with httpx.AsyncClient(**kwargs) as client:
         response = await client.get(
             url,
             headers={"User-Agent": "MapMil-CP-Viina/1.0"},

centers/parsing/workers/workers/heuristic_profile.py

@@ -0,0 +1,75 @@
+"""Apply HeuristicProfile to Telegram posts → IngestEventItem-shaped dicts."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from contracts.heuristic_profile import HeuristicProfile, match_profile
+from workers.llm_extract import parse_coords, parse_date
+
+
+def profile_fields_to_ingest(
+    *,
+    source_type: str,
+    source_url: str,
+    raw_text: str,
+    fields: dict[str, str],
+    extra_metadata: dict | None = None,
+) -> dict:
+    lat, lng = parse_coords(str(fields.get("coords") or ""))
+    description = str(fields.get("description") or raw_text)[:8000]
+    locality = str(fields.get("locality") or "")
+    region = str(fields.get("region") or locality or "") or None
+    if fields.get("title"):
+        title = str(fields["title"])
+    elif locality:
+        title = locality
+    elif description:
+        title = description.splitlines()[0][:120]
+    else:
+        title = ""
+    topic = str(fields.get("topic") or "telegram_profile")
+    event_date = parse_date(str(fields.get("event_date") or ""))
+
+    meta: dict[str, Any] = {
+        "extract_mode": "profile",
+        "extracted": dict(fields),
+    }
+    if extra_metadata:
+        meta.update(extra_metadata)
+
+    return {
+        "source_type": source_type,
+        "source_url": source_url,
+        "raw_text": raw_text[:20000],
+        "title": title[:255],
+        "description": description,
+        "locality": locality,
+        "latitude": lat,
+        "longitude": lng,
+        "event_date": event_date.isoformat() if event_date else None,
+        "region": region,
+        "topic": topic,
+        "tags": [source_type, "profile"],
+        "metadata": meta,
+    }
+
+
+def extract_with_profile(
+    text: str,
+    profile: HeuristicProfile | dict[str, Any],
+    *,
+    source_url: str,
+    source_type: str = "telegram",
+    extra_metadata: dict | None = None,
+) -> dict | None:
+    matched, fields, _missing = match_profile(text, profile)
+    if not matched:
+        return None
+    return profile_fields_to_ingest(
+        source_type=source_type,
+        source_url=source_url,
+        raw_text=text,
+        fields=fields,
+        extra_metadata=extra_metadata,
+    )

centers/parsing/workers/workers/llm_extract.py

@@ -5,30 +5,31 @@ from __future__ import annotations
 import json
 import logging
 import os
-import re
 from datetime import datetime, timezone
 from typing import Any
 
 import httpx
 
+from contracts.heuristic_profile import parse_coords
+from contracts.llm_profile import (
+    DEFAULT_EXTRACT_SCHEMA,
+    DEFAULT_INSTRUCTION,
+    match_llm_required,
+)
+
 logger = logging.getLogger("cp-worker.llm")
 
-COORDS_RE = re.compile(r"(-?\d{1,3}\.\d+)\s*,\s*(-?\d{1,3}\.\d+)")
-
-DEFAULT_EXTRACT_SCHEMA: dict[str, str] = {
-    "title": "string — short event title",
-    "locality": "string — place / settlement name",
-    "event_date": "string — date as DD.MM.YYYY or YYYY-MM-DD if known",
-    "description": "string — concise event summary",
-    "coords": "string — latitude, longitude if present else empty",
-    "topic": "string — short topic tag",
-}
-
-DEFAULT_INSTRUCTION = (
-    "Extract structured military/news event fields from the text. "
-    "If the text is not an event, return is_event=false. "
-    "Respond with a single JSON object only."
-)
+# Re-export for adapters that import from this module
+__all__ = [
+    "DEFAULT_EXTRACT_SCHEMA",
+    "DEFAULT_INSTRUCTION",
+    "event_source_url",
+    "extract_event_fields",
+    "extract_event_list",
+    "fields_to_ingest",
+    "llm_enabled",
+    "match_llm_required",
+]
 
 
 def llm_enabled() -> bool:
@@ -43,29 +44,57 @@ def llm_settings() -> dict[str, str]:
     }
 
 
-async def extract_event_fields(
-    text: str,
+def event_source_url(post_url: str, index: int, total: int) -> str:
+    """Stable per-event URL for CA dedup. Single event keeps bare post URL."""
+    base = (post_url or "").strip()
+    if total <= 1:
+        return base
+    # index is 0-based; fragment uses 1-based #eN
+    return f"{base}#e{index + 1}"
+
+
+def _normalize_fields(raw: dict[str, Any], schema: dict[str, str]) -> dict[str, str]:
+    return {key: str(raw.get(key) or "").strip() for key in schema}
+
+
+def _parse_llm_content(
+    content: str,
+    schema: dict[str, str],
     *,
-    extract_schema: dict[str, str] | None = None,
-    instruction: str | None = None,
-) -> dict[str, Any]:
-    """Ask DeepSeek to fill schema fields from free text. Returns dict (+ is_event)."""
+    multi_event: bool,
+) -> tuple[bool, list[dict[str, str]]]:
+    """Return (is_event, list of field dicts)."""
+    parsed = json.loads(content)
+    if not isinstance(parsed, dict):
+        return False, []
+
+    is_event = bool(parsed.get("is_event", True))
+    if not is_event:
+        return False, []
+
+    if multi_event and isinstance(parsed.get("events"), list):
+        events: list[dict[str, str]] = []
+        for item in parsed["events"]:
+            if isinstance(item, dict):
+                events.append(_normalize_fields(item, schema))
+        return True, events
+
+    # Single-event fallback (legacy fields / top-level keys)
+    fields_raw = parsed.get("fields") if isinstance(parsed.get("fields"), dict) else parsed
+    if not isinstance(fields_raw, dict):
+        fields_raw = {}
+    # Drop non-schema keys that confuse normalize when falling back to top-level
+    cleaned = {k: fields_raw.get(k) for k in schema}
+    return True, [_normalize_fields(cleaned, schema)]
+
+
+async def _call_deepseek(user_prompt: str) -> str:
     settings = llm_settings()
     if not settings["api_key"]:
         raise RuntimeError(
             "DEEPSEEK_API_KEY is not set. Add it to .env for LLM extract_mode."
         )
 
-    schema = extract_schema or DEFAULT_EXTRACT_SCHEMA
-    instr = instruction or DEFAULT_INSTRUCTION
-    schema_lines = "\n".join(f"- {k}: {v}" for k, v in schema.items())
-    user_prompt = (
-        f"{instr}\n\n"
-        f"Fields to extract:\n{schema_lines}\n\n"
-        'Return JSON: {"is_event": true|false, "fields": {<field>: <string>}}\n\n'
-        f"Text:\n{text[:12000]}"
-    )
-
     payload = {
         "model": settings["model"],
         "messages": [
@@ -95,14 +124,72 @@ async def extract_event_fields(
         response.raise_for_status()
         data = response.json()
 
-    content = data["choices"][0]["message"]["content"]
-    parsed = json.loads(content)
-    fields = parsed.get("fields") if isinstance(parsed.get("fields"), dict) else parsed
-    if not isinstance(fields, dict):
-        fields = {}
-    # Normalize to strings for known keys
-    result = {key: str(fields.get(key) or "").strip() for key in schema}
-    result["is_event"] = bool(parsed.get("is_event", True))
+    return data["choices"][0]["message"]["content"]
+
+
+async def extract_event_list(
+    text: str,
+    *,
+    extract_schema: dict[str, str] | None = None,
+    instruction: str | None = None,
+    multi_event: bool = False,
+) -> list[dict[str, Any]]:
+    """Extract zero or more event field dicts from free text.
+
+    Each dict has schema keys as strings. Empty list if not an event / no items.
+    """
+    schema = extract_schema or DEFAULT_EXTRACT_SCHEMA
+    instr = instruction or DEFAULT_INSTRUCTION
+    schema_lines = "\n".join(f"- {k}: {v}" for k, v in schema.items())
+
+    if multi_event:
+        user_prompt = (
+            f"{instr}\n\n"
+            "If the text describes multiple distinct events (different places, "
+            "coords, or dates), return one object per event in \"events\".\n"
+            f"Fields per event:\n{schema_lines}\n\n"
+            'Return JSON: {"is_event": true|false, "events": [{<field>: <string>}, ...]}\n'
+            "If there is no event, return is_event=false and events=[].\n\n"
+            f"Text:\n{text[:12000]}"
+        )
+    else:
+        user_prompt = (
+            f"{instr}\n\n"
+            f"Fields to extract:\n{schema_lines}\n\n"
+            'Return JSON: {"is_event": true|false, "fields": {<field>: <string>}}\n\n'
+            f"Text:\n{text[:12000]}"
+        )
+
+    content = await _call_deepseek(user_prompt)
+    is_event, events = _parse_llm_content(content, schema, multi_event=multi_event)
+    if not is_event:
+        return []
+    return events
+
+
+async def extract_event_fields(
+    text: str,
+    *,
+    extract_schema: dict[str, str] | None = None,
+    instruction: str | None = None,
+) -> dict[str, Any]:
+    """Ask DeepSeek to fill schema fields from free text. Returns dict (+ is_event).
+
+    Single-event API kept for crawl4ai and legacy callers.
+    """
+    events = await extract_event_list(
+        text,
+        extract_schema=extract_schema,
+        instruction=instruction,
+        multi_event=False,
+    )
+    if not events:
+        schema = extract_schema or DEFAULT_EXTRACT_SCHEMA
+        empty = {key: "" for key in schema}
+        empty["is_event"] = False
+        return empty
+    result = dict(events[0])
+    result["is_event"] = True
     return result
 
 
@@ -146,13 +233,6 @@ def fields_to_ingest(
     }
 
 
-def parse_coords(raw: str) -> tuple[float | None, float | None]:
-    match = COORDS_RE.search(raw or "")
-    if not match:
-        return None, None
-    return float(match.group(1)), float(match.group(2))
-
-
 def parse_date(raw: str) -> datetime | None:
     if not raw:
         return None

centers/parsing/workers/workers/sources/telegram_listener.py

@@ -9,6 +9,7 @@ import httpx
 from telethon import TelegramClient, events
 
 from workers.converter import event_record_to_ingest
+from workers.heuristic_profile import extract_with_profile
 from workers.parsers.telegram_events import parse_event_post
 from workers.sources.telegram_client import message_to_post, normalize_channel
 
@@ -22,11 +23,12 @@ REFRESH_SECONDS = int(os.getenv("TELEGRAM_LISTENER_REFRESH_SECONDS", "60"))
 class TelegramListener:
     def __init__(self, client: TelegramClient) -> None:
         self.client = client
-        self._channels: dict[str, int] = {}
+        # channel -> {job_id, source_config}
+        self._channels: dict[str, dict[str, Any]] = {}
         self._chat_ids: set[int] = set()
         self._handlers_registered = False
 
-    async def fetch_subscriptions(self) -> dict[str, int]:
+    async def fetch_subscriptions(self) -> dict[str, dict[str, Any]]:
         async with httpx.AsyncClient(timeout=30.0) as http:
             response = await http.get(
                 f"{CA_API_URL}/internal/listener/subscriptions",
@@ -35,7 +37,7 @@ class TelegramListener:
             response.raise_for_status()
             data = response.json()
 
-        channels: dict[str, int] = {}
+        channels: dict[str, dict[str, Any]] = {}
         for item in data:
             raw = item.get("channel")
             job_id = item.get("job_id")
@@ -46,7 +48,13 @@ class TelegramListener:
             except ValueError:
                 logger.warning("Skip invalid channel in subscription: %r", raw)
                 continue
-            channels.setdefault(key, int(job_id))
+            channels.setdefault(
+                key,
+                {
+                    "job_id": int(job_id),
+                    "source_config": dict(item.get("source_config") or {}),
+                },
+            )
         return channels
 
     async def refresh_subscriptions(self) -> None:
@@ -73,10 +81,37 @@ class TelegramListener:
             ", ".join(sorted(channels)) or "(none)",
         )
 
-    async def _ingest_post(self, channel: str, post) -> None:
+    def _build_event(self, channel: str, post) -> dict | None:
+        sub = self._channels.get(normalize_channel(channel)) or {}
+        cfg = sub.get("source_config") or {}
+        extract_mode = cfg.get("extract_mode") or "heuristic"
+        text = (post.text or "").strip()
+
+        if extract_mode == "profile" and cfg.get("heuristic_profile"):
+            return extract_with_profile(
+                text,
+                cfg["heuristic_profile"],
+                source_url=post.url,
+                source_type="telegram",
+                extra_metadata={
+                    "channel": post.channel,
+                    "message_id": post.id,
+                    "post_date": post.date.isoformat() if post.date else None,
+                    "listener": True,
+                },
+            )
+
+        # llm jobs fall back to heuristic in listener (LLM is batch-only by design)
         record = parse_event_post(post)
-        event = event_record_to_ingest(record)
-        job_id = self._channels.get(normalize_channel(channel))
+        return event_record_to_ingest(record)
+
+    async def _ingest_post(self, channel: str, post) -> None:
+        event = self._build_event(channel, post)
+        if event is None:
+            logger.debug("Listener skip %s (profile required fields missing)", post.url)
+            return
+        sub = self._channels.get(normalize_channel(channel)) or {}
+        job_id = sub.get("job_id")
 
         async with httpx.AsyncClient(timeout=60.0) as http:
             response = await http.post(

centers/parsing/workers/workers/sources/telegram_session.py

@@ -1,20 +1,39 @@
 """Единое подключение Telethon для listener и batch-заданий."""
 
+import logging
 from pathlib import Path
 
 from telethon import TelegramClient
 from telethon.errors import AuthKeyUnregisteredError, SessionPasswordNeededError
 
-from workers.sources.telegram_settings import create_client, get_api_credentials, get_session_path
+from workers.sources.telegram_settings import (
+    create_client,
+    describe_connection,
+    get_api_credentials,
+    get_session_path,
+)
 from workers.sources.telegram_client import TelegramAuthError, TelegramConfigError
+from workers.sources.vpn_config import proxy_fingerprint
+
+logger = logging.getLogger("cp-worker.telegram-session")
 
 _shared_client: TelegramClient | None = None
+_proxy_fingerprint: str | None = None
 
 
 async def get_shared_client() -> TelegramClient:
-    global _shared_client
+    global _shared_client, _proxy_fingerprint
+
+    fp = proxy_fingerprint()
     if _shared_client is not None and _shared_client.is_connected():
-        return _shared_client
+        if fp == _proxy_fingerprint:
+            return _shared_client
+        logger.info(
+            "VPN proxy changed (%s → %s); reconnecting Telethon",
+            _proxy_fingerprint,
+            fp,
+        )
+        await close_shared_client()
 
     try:
         api_id, api_hash = get_api_credentials()
@@ -29,6 +48,7 @@ async def get_shared_client() -> TelegramClient:
         )
 
     client = create_client(session_path, api_id, api_hash)
+    logger.info("Connecting Telethon %s", describe_connection())
     try:
         await client.connect()
         if not await client.is_user_authorized():
@@ -45,11 +65,13 @@ async def get_shared_client() -> TelegramClient:
         ) from exc
 
     _shared_client = client
+    _proxy_fingerprint = fp
     return client
 
 
 async def close_shared_client() -> None:
-    global _shared_client
+    global _shared_client, _proxy_fingerprint
     if _shared_client is not None:
         await _shared_client.disconnect()
         _shared_client = None
+    _proxy_fingerprint = None

centers/parsing/workers/workers/sources/telegram_settings.py

@@ -28,7 +28,8 @@ def get_api_credentials() -> tuple[int, str]:
     return api_id, api_hash
 
 
-def get_proxy() -> tuple | None:
+def get_env_proxy() -> tuple | None:
+    """Legacy TELEGRAM_PROXY_* env fallback (used when CA VPN is off)."""
     proxy_type = os.environ.get("TELEGRAM_PROXY_TYPE", "").strip().lower()
     if not proxy_type or proxy_type == "none":
         return None
@@ -49,6 +50,15 @@ def get_proxy() -> tuple | None:
     return proxy_type, host, port
 
 
+def get_proxy() -> tuple | None:
+    try:
+        from workers.sources.vpn_config import telethon_proxy_tuple
+
+        return telethon_proxy_tuple()
+    except Exception:
+        return get_env_proxy()
+
+
 def describe_connection() -> str:
     proxy = get_proxy()
     if not proxy:

centers/parsing/workers/workers/sources/vpn_config.py

@@ -0,0 +1,124 @@
+"""Fetch VPN settings from CA /internal/vpn (cached)."""
+
+from __future__ import annotations
+
+import logging
+import os
+import time
+from typing import Any
+
+import httpx
+
+from contracts.vpn import VpnSettings
+
+logger = logging.getLogger("cp-worker.vpn")
+
+CA_API_URL = os.getenv("CA_API_URL", "http://ca-api:8000").rstrip("/")
+INTERNAL_TOKEN = os.getenv("INTERNAL_TOKEN", "dev-internal-token")
+CACHE_TTL = float(os.getenv("VPN_CONFIG_CACHE_SECONDS", "30"))
+CP_VPN_HOST = os.getenv("CP_VPN_HOST", "cp-vpn")
+CP_VPN_PORT = int(os.getenv("CP_VPN_PORT", "1080"))
+
+_cache: VpnSettings | None = None
+_cache_at: float = 0.0
+_cache_failed: bool = False
+
+
+def invalidate_vpn_cache() -> None:
+    global _cache, _cache_at, _cache_failed
+    _cache = None
+    _cache_at = 0.0
+    _cache_failed = False
+
+
+def fetch_vpn_settings(*, force: bool = False) -> VpnSettings | None:
+    """Return VPN settings from CA, or None if unreachable / unset."""
+    global _cache, _cache_at, _cache_failed
+
+    now = time.monotonic()
+    if (
+        not force
+        and _cache is not None
+        and (now - _cache_at) < CACHE_TTL
+        and not _cache_failed
+    ):
+        return _cache
+
+    try:
+        with httpx.Client(timeout=10.0) as client:
+            response = client.get(
+                f"{CA_API_URL}/internal/vpn",
+                headers={"X-Internal-Token": INTERNAL_TOKEN},
+            )
+            response.raise_for_status()
+            raw: dict[str, Any] = response.json()
+        settings = VpnSettings.model_validate(raw)
+        _cache = settings
+        _cache_at = now
+        _cache_failed = False
+        return settings
+    except Exception:
+        logger.exception("Failed to load VPN settings from CA")
+        _cache_failed = True
+        _cache_at = now
+        # Keep stale cache if present
+        return _cache
+
+
+def proxy_fingerprint(settings: VpnSettings | None = None) -> str:
+    cfg = settings if settings is not None else fetch_vpn_settings()
+    if cfg is None or not cfg.proxies_source("telegram"):
+        # Env fallback fingerprint
+        from workers.sources.telegram_settings import get_env_proxy
+
+        env = get_env_proxy()
+        return f"env:{env!r}"
+    if cfg.mode == "subscription":
+        return f"sub:{CP_VPN_HOST}:{CP_VPN_PORT}:{cfg.subscription_url}"
+    return (
+        f"{cfg.mode}:{cfg.host}:{cfg.port}:"
+        f"{cfg.username or ''}:{'*' if cfg.password else ''}"
+    )
+
+
+def telethon_proxy_tuple(settings: VpnSettings | None = None) -> tuple | None:
+    """Telethon-compatible proxy tuple for telegram traffic, or None."""
+    cfg = settings if settings is not None else fetch_vpn_settings()
+    if cfg is not None and cfg.proxies_source("telegram"):
+        if cfg.mode == "subscription":
+            return ("socks5", CP_VPN_HOST, CP_VPN_PORT)
+        if cfg.mode in ("socks5", "http") and cfg.host and cfg.port:
+            if cfg.username:
+                return (
+                    cfg.mode,
+                    cfg.host,
+                    int(cfg.port),
+                    True,
+                    cfg.username,
+                    cfg.password or "",
+                )
+            return (cfg.mode, cfg.host, int(cfg.port))
+        return None
+
+    from workers.sources.telegram_settings import get_env_proxy
+
+    return get_env_proxy()
+
+
+def http_proxy_url(source_type: str, settings: VpnSettings | None = None) -> str | None:
+    """HTTP(S)/SOCKS proxy URL for httpx / browsers, or None."""
+    cfg = settings if settings is not None else fetch_vpn_settings()
+    if cfg is None or not cfg.proxies_source(source_type):
+        return None
+    if cfg.mode == "subscription":
+        return f"socks5://{CP_VPN_HOST}:{CP_VPN_PORT}"
+    if cfg.mode in ("socks5", "http") and cfg.host and cfg.port:
+        auth = ""
+        if cfg.username:
+            from urllib.parse import quote
+
+            user = quote(cfg.username, safe="")
+            password = quote(cfg.password or "", safe="")
+            auth = f"{user}:{password}@"
+        return f"{cfg.mode}://{auth}{cfg.host}:{int(cfg.port)}"
+    return None

contracts/heuristic_profile.py

@@ -0,0 +1,218 @@
+"""Declarative heuristic profile: schema + static rule interpreter (CA preview + CP runtime)."""
+
+from __future__ import annotations
+
+import re
+from typing import Any, Literal
+
+from pydantic import BaseModel, Field, field_validator, model_validator
+
+TARGET_FIELDS: tuple[str, ...] = (
+    "title",
+    "description",
+    "locality",
+    "event_date",
+    "coords",
+    "topic",
+    "region",
+)
+
+COORDS_RE = re.compile(r"(-?\d{1,3}\.\d+)\s*,\s*(-?\d{1,3}\.\d+)")
+
+Strategy = Literal["regex", "line", "after_marker", "between", "full_text", "literal"]
+
+
+class FieldRule(BaseModel):
+    strategy: Strategy
+    # regex: named or group(1); line: 0-based index; after_marker/between: markers
+    pattern: str | None = None
+    group: int = 1
+    line_index: int | None = None
+    marker: str | None = None
+    end_marker: str | None = None
+    value: str | None = None  # literal
+    flags: str = ""  # e.g. "im" → re.I|re.M
+    strip: bool = True
+
+    @field_validator("pattern", "marker", "end_marker", "value", mode="before")
+    @classmethod
+    def empty_to_none(cls, value: Any) -> Any:
+        if value is None:
+            return None
+        if isinstance(value, str) and not value.strip():
+            return None
+        return value
+
+
+class HeuristicProfile(BaseModel):
+    version: Literal[1] = 1
+    fields: dict[str, FieldRule] = Field(default_factory=dict)
+    required_fields: list[str] = Field(default_factory=list)
+    notes: str = ""
+
+    @field_validator("required_fields")
+    @classmethod
+    def known_required(cls, value: list[str]) -> list[str]:
+        seen: list[str] = []
+        unknown: list[str] = []
+        for name in value:
+            if name not in TARGET_FIELDS:
+                unknown.append(name)
+                continue
+            if name not in seen:
+                seen.append(name)
+        if unknown:
+            raise ValueError(f"Unknown required_fields: {sorted(set(unknown))}")
+        return seen
+
+    @model_validator(mode="after")
+    def known_fields_only(self) -> "HeuristicProfile":
+        unknown = set(self.fields) - set(TARGET_FIELDS)
+        if unknown:
+            raise ValueError(f"Unknown profile fields: {sorted(unknown)}")
+        return self
+
+
+def target_field_specs() -> list[dict[str, str]]:
+    """Fixed target table for profile UI (roadmap: custom tables later)."""
+    return [
+        {"name": "title", "type": "string", "description": "Short event title"},
+        {"name": "description", "type": "string", "description": "Event summary / body"},
+        {"name": "locality", "type": "string", "description": "Place / settlement name"},
+        {
+            "name": "event_date",
+            "type": "string",
+            "description": "Date as DD.MM.YYYY or YYYY-MM-DD",
+        },
+        {
+            "name": "coords",
+            "type": "string",
+            "description": "Latitude, longitude if present",
+        },
+        {"name": "topic", "type": "string", "description": "Short topic tag"},
+        {"name": "region", "type": "string", "description": "Region (optional)"},
+    ]
+
+
+def _compile_flags(flags: str) -> int:
+    mapping = {
+        "i": re.IGNORECASE,
+        "m": re.MULTILINE,
+        "s": re.DOTALL,
+    }
+    result = 0
+    for ch in (flags or "").lower():
+        result |= mapping.get(ch, 0)
+    return result
+
+
+def _apply_rule(text: str, rule: FieldRule) -> str:
+    raw = text or ""
+    value = ""
+
+    if rule.strategy == "literal":
+        value = rule.value or ""
+    elif rule.strategy == "full_text":
+        value = raw
+    elif rule.strategy == "line":
+        lines = raw.splitlines()
+        idx = 0 if rule.line_index is None else rule.line_index
+        if 0 <= idx < len(lines):
+            value = lines[idx]
+    elif rule.strategy == "regex":
+        if not rule.pattern:
+            return ""
+        match = re.search(rule.pattern, raw, _compile_flags(rule.flags))
+        if match:
+            try:
+                value = match.group(rule.group)
+            except IndexError:
+                value = match.group(0)
+    elif rule.strategy == "after_marker":
+        marker = rule.marker or ""
+        if not marker:
+            return ""
+        pos = raw.find(marker)
+        if pos < 0:
+            return ""
+        start = pos + len(marker)
+        rest = raw[start:]
+        if rule.end_marker:
+            end = rest.find(rule.end_marker)
+            value = rest[:end] if end >= 0 else rest
+        elif rule.pattern:
+            match = re.search(rule.pattern, rest, _compile_flags(rule.flags))
+            if match:
+                try:
+                    value = match.group(rule.group)
+                except IndexError:
+                    value = match.group(0)
+        else:
+            # first non-empty line after marker
+            for line in rest.splitlines():
+                if line.strip():
+                    value = line
+                    break
+    elif rule.strategy == "between":
+        start_m = rule.marker or ""
+        end_m = rule.end_marker or ""
+        if not start_m or not end_m:
+            return ""
+        start = raw.find(start_m)
+        if start < 0:
+            return ""
+        start += len(start_m)
+        end = raw.find(end_m, start)
+        if end < 0:
+            return ""
+        value = raw[start:end]
+
+    if rule.strip:
+        value = value.strip()
+    return value
+
+
+def apply_profile(text: str, profile: HeuristicProfile | dict[str, Any]) -> dict[str, str]:
+    """Apply static rules to post text → string field map (no LLM)."""
+    if isinstance(profile, dict):
+        profile = HeuristicProfile.model_validate(profile)
+
+    result: dict[str, str] = {name: "" for name in TARGET_FIELDS}
+    for name, rule in profile.fields.items():
+        result[name] = _apply_rule(text, rule)
+    return result
+
+
+def parse_coords(raw: str) -> tuple[float | None, float | None]:
+    match = COORDS_RE.search(raw or "")
+    if not match:
+        return None, None
+    return float(match.group(1)), float(match.group(2))
+
+
+def _field_is_filled(name: str, value: str) -> bool:
+    if not (value or "").strip():
+        return False
+    if name == "coords":
+        lat, lng = parse_coords(value)
+        return lat is not None and lng is not None
+    return True
+
+
+def match_profile(
+    text: str,
+    profile: HeuristicProfile | dict[str, Any],
+) -> tuple[bool, dict[str, str], list[str]]:
+    """Apply profile and report whether required_fields are filled.
+
+    Empty required_fields means no gate (legacy profiles ingest every post).
+    """
+    if isinstance(profile, dict):
+        profile = HeuristicProfile.model_validate(profile)
+    fields = apply_profile(text, profile)
+    missing = [
+        name
+        for name in profile.required_fields
+        if not _field_is_filled(name, fields.get(name, ""))
+    ]
+    return (not missing, fields, missing)

contracts/llm_profile.py

@@ -0,0 +1,86 @@
+"""Reusable LLM extract profile (CA storage + flattened into TelegramSourceConfig)."""
+
+from __future__ import annotations
+
+from typing import Any
+
+from pydantic import BaseModel, Field, field_validator, model_validator
+
+from contracts.heuristic_profile import TARGET_FIELDS
+
+# LLM defaults omit region (same as workers/llm_extract.DEFAULT_EXTRACT_SCHEMA).
+LLM_SCHEMA_FIELDS: tuple[str, ...] = (
+    "title",
+    "locality",
+    "event_date",
+    "description",
+    "coords",
+    "topic",
+)
+
+DEFAULT_EXTRACT_SCHEMA: dict[str, str] = {
+    "title": "string — short event title",
+    "locality": "string — place / settlement name",
+    "event_date": "string — date as DD.MM.YYYY or YYYY-MM-DD if known",
+    "description": "string — concise event summary",
+    "coords": "string — latitude, longitude if present else empty",
+    "topic": "string — short topic tag",
+}
+
+DEFAULT_INSTRUCTION = (
+    "Extract structured military/news event fields from the text. "
+    "If the text is not an event, return is_event=false. "
+    "Respond with a single JSON object only."
+)
+
+
+class LlmProfile(BaseModel):
+    instruction: str | None = None
+    extract_schema: dict[str, str] = Field(default_factory=lambda: dict(DEFAULT_EXTRACT_SCHEMA))
+    required_fields: list[str] = Field(default_factory=list)
+    # When true, LLM may return multiple events per post (array "events")
+    multi_event: bool = False
+
+    @field_validator("instruction", mode="before")
+    @classmethod
+    def empty_instruction_to_none(cls, value: Any) -> Any:
+        if value is None:
+            return None
+        if isinstance(value, str) and not value.strip():
+            return None
+        return value
+
+    @field_validator("required_fields")
+    @classmethod
+    def known_required(cls, value: list[str]) -> list[str]:
+        seen: list[str] = []
+        unknown: list[str] = []
+        for name in value:
+            if name not in TARGET_FIELDS:
+                unknown.append(name)
+                continue
+            if name not in seen:
+                seen.append(name)
+        if unknown:
+            raise ValueError(f"Unknown required_fields: {sorted(set(unknown))}")
+        return seen
+
+    @model_validator(mode="after")
+    def known_schema_keys(self) -> "LlmProfile":
+        if not self.extract_schema:
+            raise ValueError("extract_schema must not be empty")
+        unknown = set(self.extract_schema) - set(TARGET_FIELDS)
+        if unknown:
+            raise ValueError(f"Unknown extract_schema fields: {sorted(unknown)}")
+        return self
+
+
+def match_llm_required(fields: dict[str, Any], required_fields: list[str]) -> bool:
+    """True if all required_fields are non-empty strings (empty required = no gate)."""
+    if not required_fields:
+        return True
+    for name in required_fields:
+        val = fields.get(name)
+        if val is None or not str(val).strip():
+            return False
+    return True

contracts/sources.py

@@ -10,16 +10,55 @@ from pydantic import BaseModel, Field, field_validator, model_validator
 class TelegramSourceConfig(BaseModel):
     channel: str = Field(min_length=1)
     limit: int = Field(default=100, ge=1, le=1000)
-    # heuristic = legacy telegram_events parser; llm = DeepSeek structured extract
-    extract_mode: Literal["heuristic", "llm"] = "heuristic"
+    # heuristic = legacy telegram_events; llm = DeepSeek per post; profile = static rules
+    extract_mode: Literal["heuristic", "llm", "profile"] = "heuristic"
     extract_schema: dict[str, str] | None = None
     instruction: str | None = None
+    heuristic_profile: dict | None = None
+    # Post-extract gate for extract_mode=llm (empty = only is_event filter)
+    required_fields: list[str] = Field(default_factory=list)
+    # Split one post into N events when extract_mode=llm (opt-in)
+    multi_event: bool = False
+    sample_post: str | None = None  # audit / re-generate; not required at runtime
 
     @field_validator("channel")
     @classmethod
     def strip_channel(cls, value: str) -> str:
         return value.strip()
 
+    @field_validator("required_fields")
+    @classmethod
+    def known_required(cls, value: list[str]) -> list[str]:
+        from contracts.heuristic_profile import TARGET_FIELDS
+
+        seen: list[str] = []
+        unknown: list[str] = []
+        for name in value:
+            if name not in TARGET_FIELDS:
+                unknown.append(name)
+                continue
+            if name not in seen:
+                seen.append(name)
+        if unknown:
+            raise ValueError(f"Unknown required_fields: {sorted(set(unknown))}")
+        return seen
+
+    @model_validator(mode="after")
+    def validate_extract_mode(self) -> "TelegramSourceConfig":
+        if self.extract_mode == "profile":
+            if not self.heuristic_profile:
+                raise ValueError("heuristic_profile required when extract_mode=profile")
+            from contracts.heuristic_profile import HeuristicProfile
+
+            HeuristicProfile.model_validate(self.heuristic_profile)
+        elif self.extract_mode == "llm" and self.extract_schema is not None:
+            from contracts.heuristic_profile import TARGET_FIELDS
+
+            unknown = set(self.extract_schema) - set(TARGET_FIELDS)
+            if unknown:
+                raise ValueError(f"Unknown extract_schema fields: {sorted(unknown)}")
+        return self
+
 
 class Crawl4AISourceConfig(BaseModel):
     urls: list[str] = Field(min_length=1)

contracts/vpn.py

@@ -0,0 +1,71 @@
+"""VPN / proxy settings shared by CA admin UI and CP workers."""
+
+from __future__ import annotations
+
+from typing import Literal
+
+from pydantic import BaseModel, Field, field_validator, model_validator
+
+from contracts.queues import known_source_types
+
+VpnMode = Literal["subscription", "socks5", "http"]
+
+VPN_MODES: tuple[str, ...] = ("subscription", "socks5", "http")
+
+
+class VpnSettings(BaseModel):
+    enabled: bool = False
+    mode: VpnMode = "subscription"
+    subscription_url: str | None = None
+    subscription_interval_seconds: int = Field(default=3600, ge=60, le=86400)
+    host: str | None = None
+    port: int | None = Field(default=None, ge=1, le=65535)
+    username: str | None = None
+    password: str | None = None
+    proxied_source_types: list[str] = Field(default_factory=list)
+
+    @field_validator("subscription_url", "host", "username", "password", mode="before")
+    @classmethod
+    def empty_to_none(cls, value: object) -> object:
+        if value is None:
+            return None
+        if isinstance(value, str):
+            stripped = value.strip()
+            return stripped or None
+        return value
+
+    @field_validator("proxied_source_types")
+    @classmethod
+    def validate_source_types(cls, value: list[str]) -> list[str]:
+        known = set(known_source_types())
+        cleaned: list[str] = []
+        seen: set[str] = set()
+        for raw in value or []:
+            item = str(raw).strip()
+            if not item or item in seen:
+                continue
+            if item not in known:
+                raise ValueError(
+                    f"Unknown source_type {item!r}. "
+                    f"Allowed: {', '.join(sorted(known))}"
+                )
+            seen.add(item)
+            cleaned.append(item)
+        return cleaned
+
+    @model_validator(mode="after")
+    def require_fields_when_enabled(self) -> "VpnSettings":
+        if not self.enabled:
+            return self
+        if self.mode == "subscription":
+            if not self.subscription_url:
+                raise ValueError("subscription_url required when mode=subscription")
+        elif self.mode in ("socks5", "http"):
+            if not self.host:
+                raise ValueError(f"host required when mode={self.mode}")
+            if self.port is None:
+                raise ValueError(f"port required when mode={self.mode}")
+        return self
+
+    def proxies_source(self, source_type: str) -> bool:
+        return self.enabled and source_type in self.proxied_source_types

docker-compose.yml

@@ -22,11 +22,16 @@ services:
     build:
       context: .
       dockerfile: centers/analytics/api/Dockerfile
+    env_file:
+      - .env
     environment:
       DATABASE_URL: postgresql://mapmil:mapmil@ca-db:5432/mapmil
       REDIS_URL: redis://redis:6379/0
       INTERNAL_TOKEN: dev-internal-token
       TEST_PI_API_KEY: test-pi-api-key-change-me
+      ADMIN_USER: ${ADMIN_USER:-admin}
+      ADMIN_PASSWORD: ${ADMIN_PASSWORD:-change-me}
+      ADMIN_JWT_SECRET: ${ADMIN_JWT_SECRET:-change-me-jwt-secret}
     volumes:
       - ca_uploads:/data
     depends_on:
@@ -44,6 +49,17 @@ services:
       - ca-api
     restart: unless-stopped
 
+  cp-vpn:
+    build: ./centers/parsing/vpn
+    environment:
+      CA_API_URL: http://ca-api:8000
+      INTERNAL_TOKEN: ${INTERNAL_TOKEN:-dev-internal-token}
+      VPN_POLL_SECONDS: "15"
+      VPN_SOCKS_PORT: "1080"
+    depends_on:
+      - ca-api
+    restart: unless-stopped
+
   cp-workers:
     build:
       context: .
@@ -57,16 +73,25 @@ services:
       TELEGRAM_SESSION_PATH: /data/telegram.session
       TELEGRAM_LISTENER_ENABLED: "true"
       TELEGRAM_LISTENER_REFRESH_SECONDS: "60"
+      # Env fallback when VPN tab is off (host Happ/xray; socat 11080→10808).
+      # Overrides .env 127.0.0.1 which is wrong inside the container.
+      TELEGRAM_PROXY_HOST: host.docker.internal
+      TELEGRAM_PROXY_PORT: "11080"
+      CP_VPN_HOST: cp-vpn
+      CP_VPN_PORT: "1080"
       ENABLED_ADAPTERS: telegram
       WORKER_FAMILIES: telegram
       CA_API_URL: http://ca-api:8000
       REDIS_URL: redis://redis:6379/0
-      INTERNAL_TOKEN: dev-internal-token
+      INTERNAL_TOKEN: ${INTERNAL_TOKEN:-dev-internal-token}
+    extra_hosts:
+      - "host.docker.internal:host-gateway"
     volumes:
       - ./data:/data
     depends_on:
       - redis
       - ca-api
+      - cp-vpn
     restart: unless-stopped
 
   cp-workers-web:
@@ -82,12 +107,15 @@ services:
       TELEGRAM_LISTENER_ENABLED: "false"
       ENABLED_ADAPTERS: crawl4ai
       WORKER_FAMILIES: web
+      CP_VPN_HOST: cp-vpn
+      CP_VPN_PORT: "1080"
       CA_API_URL: http://ca-api:8000
       REDIS_URL: redis://redis:6379/0
-      INTERNAL_TOKEN: dev-internal-token
+      INTERNAL_TOKEN: ${INTERNAL_TOKEN:-dev-internal-token}
     depends_on:
       - redis
       - ca-api
+      - cp-vpn
     restart: unless-stopped
 
   cp-workers-nlp:
@@ -101,12 +129,15 @@ services:
       TELEGRAM_LISTENER_ENABLED: "false"
       ENABLED_ADAPTERS: viina
       WORKER_FAMILIES: nlp
+      CP_VPN_HOST: cp-vpn
+      CP_VPN_PORT: "1080"
       CA_API_URL: http://ca-api:8000
       REDIS_URL: redis://redis:6379/0
-      INTERNAL_TOKEN: dev-internal-token
+      INTERNAL_TOKEN: ${INTERNAL_TOKEN:-dev-internal-token}
     depends_on:
       - redis
       - ca-api
+      - cp-vpn
     restart: unless-stopped
 
 volumes:

docs/README.md

@@ -0,0 +1,35 @@
+# Документация MapMil для разработчиков
+
+Карта документов. Начните с обзора, затем углубитесь в нужный центр.
+
+## С чего начать
+
+1. [Обзор архитектуры](architecture-overview.md) — общая картина, затем детали
+2. [Поток данных](data-flow.md) — от парсера до карты и ПИ
+3. [Локальный запуск](local-dev.md) — `.env`, сессия, Docker
+
+## По зонам
+
+| Документ | Содержание |
+|----------|------------|
+| [architecture-overview.md](architecture-overview.md) | Платформа целиком: ЦА, ЦП, ПИ, границы, сервисы |
+| [data-flow.md](data-flow.md) | E2E: admin → Redis → адаптер → ingest → карта → ПИ |
+| [contracts.md](contracts.md) | Shared-схемы `contracts/` |
+| [local-dev.md](local-dev.md) | Разработка и отладка локально |
+| [ЦА: ARCHITECTURE](../centers/analytics/ARCHITECTURE.md) | API, модели, UI, scheduler |
+| [ЦА: DISTRIBUTION (ПИ)](../centers/analytics/DISTRIBUTION.md) | External API для потребителей |
+| [ЦП: ARCHITECTURE](../centers/parsing/ARCHITECTURE.md) | Адаптеры, очереди, Telegram listener |
+
+## Корень репозитория
+
+| Файл | Для кого |
+|------|----------|
+| [`README.md`](../README.md) | Быстрый старт и краткий API |
+| [`AGENTS.md`](../AGENTS.md) | Ориентация для AI-агентов |
+| [`.cursor/skills/add-parser-adapter/`](../.cursor/skills/add-parser-adapter/) | Чеклист нового `source_type` |
+
+## Принцип
+
+- **Contracts first** — изменение формы данных или маршрутизации начинается в `contracts/`.
+- **Одна зона за раз** — не смешивать контракты + UI + инфра в одном несвязанном изменении.
+- **ЦП не трогает БД ЦА** — только Redis и `POST/PATCH /internal/*`.

docs/architecture-overview.md

@@ -0,0 +1,297 @@
+# Обзор архитектуры MapMil
+
+Документ для разработчиков: сначала общая картина, затем устройство платформы подробнее.
+
+Связанные материалы: [поток данных](data-flow.md), [контракты](contracts.md), [ЦА](../centers/analytics/ARCHITECTURE.md), [ЦП](../centers/parsing/ARCHITECTURE.md), [ПИ](../centers/analytics/DISTRIBUTION.md).
+
+---
+
+## 1. Общими чертами
+
+**MapMil** — monorepo платформы сбора и отображения событий:
+
+```text
+Источники  →  ЦП (парсинг)  →  ЦА (хранение + карта + admin)  →  ПИ (внешние клиенты)
+```
+
+| Роль | Что это | Где код |
+|------|---------|---------|
+| **ЦА** (аналитика) | Единственный UI, PostgreSQL, ingest, admin, карта, distribution API | `centers/analytics/` |
+| **ЦП** (парсинг) | Адаптеры источников без своей БД и без UI | `centers/parsing/workers/` |
+| **ПИ** (потребители) | Не отдельный сервис: `GET /api/v1/events` в ЦА + ключи consumers | `centers/analytics/api/.../v1.py` |
+| **Contracts** | Общие Pydantic-схемы очередей и payload | `contracts/` |
+
+### Главный поток
+
+```mermaid
+flowchart LR
+  Admin[Admin UI / API]
+  Redis[(Redis queues)]
+  CP[CP adapters]
+  Src[Telegram / Web / NLP]
+  CA[(PostgreSQL)]
+  Map[Map UI]
+  PI[External API]
+
+  Admin -->|enqueue job| Redis
+  Redis -->|BLPOP| CP
+  CP --> Src
+  CP -->|POST /internal/ingest| CA
+  CA --> Map
+  CA --> PI
+```
+
+### Сценарий: создание парсера и взаимодействие
+
+Объекты: **ЦА** (UI + API), **ЦП**, **ПИ**, **PostgreSQL**, **Redis**. Кто кому что передаёт при создании парсера и дальнейшем чтении данных.
+
+```mermaid
+sequenceDiagram
+  participant UI as ЦА UI
+  participant CA as ЦА API
+  participant PG as PostgreSQL
+  participant R as Redis
+  participant CP as ЦП
+  participant PI as ПИ клиент
+
+  Note over UI,CP: Создание парсера и batch-прогон
+  UI->>CA: POST /admin/jobs<br/>source_type + source_config
+  CA->>PG: INSERT ParseJob status=queued
+  CA->>R: RPUSH cp:jobs:family<br/>JobPayload job_id type config
+  R->>CP: BLPOP задание
+  CP->>CA: PATCH /internal/jobs/id<br/>status=running
+  CA->>PG: UPDATE ParseJob status
+  CP->>CP: fetch источника<br/>адаптер → события
+  CP->>CA: POST /internal/ingest<br/>events IngestEventItem
+  CA->>PG: INSERT Event<br/>дедуп по source_url
+  CA->>PG: sync MapObject<br/>если есть coords
+  CP->>CA: PATCH job completed/failed
+  CA->>PG: UPDATE ParseJob
+
+  Note over UI,PI: Чтение результата
+  UI->>CA: GET /api/map/objects
+  CA->>PG: SELECT map/events
+  CA-->>UI: точки на карте
+
+  PI->>CA: GET /api/v1/events<br/>Bearer API key
+  CA->>PG: SELECT events<br/>фильтры Consumer
+  CA-->>PI: срез событий
+```
+
+| От | Кому | Что передаётся |
+|----|------|----------------|
+| ЦА UI → ЦА API | Конфиг парсера (`source_type`, `source_config`) |
+| ЦА API → PostgreSQL | `ParseJob`, затем `Event` / `MapObject` |
+| ЦА API → Redis | `JobPayload` (`job_id`, type, config) |
+| Redis → ЦП | То же задание (`BLPOP`) |
+| ЦП → ЦА API | Статус job + пакет событий (ingest) |
+| ЦА API → ЦА UI | Объекты карты |
+| ЦА API → ПИ | Отфильтрованный срез событий |
+
+### Жёсткие границы
+
+1. ЦП **никогда** не подключается к PostgreSQL ЦА.
+2. Связь ЦА ↔ ЦП: Redis (`cp:jobs:{family}`) + HTTP `/internal/*` с `X-Internal-Token`.
+3. Весь фронтенд — только ЦА (`ca-frontend` на порту **8080**).
+4. Дедуп событий в ЦА по стабильному `source_url`.
+5. Секреты (`.env`, `data/telegram.session`) не коммитятся.
+
+### Docker-сервисы одной командой
+
+| Сервис | Назначение |
+|--------|------------|
+| `ca-db` | PostgreSQL |
+| `redis` | Очереди заданий |
+| `ca-api` | FastAPI ЦА |
+| `ca-frontend` | Vue admin + карта |
+| `cp-workers` | Telegram (batch + listener) |
+| `cp-workers-web` | Crawl4AI |
+| `cp-workers-nlp` | VIINA |
+
+Запуск: `docker compose up --build` → http://localhost:8080
+
+---
+
+## 2. Подробнее
+
+### 2.1. Структура monorepo
+
+```text
+MapMil/
+├── centers/
+│   ├── analytics/          # ЦА
+│   │   ├── api/            # FastAPI
+│   │   ├── frontend/       # Vue 3 + Leaflet
+│   │   ├── ARCHITECTURE.md
+│   │   └── DISTRIBUTION.md
+│   └── parsing/            # ЦП
+│       ├── ARCHITECTURE.md
+│       └── workers/        # adapters + Telethon
+├── contracts/              # ingest, jobs, sources, queues
+├── docs/                   # эта документация
+├── data/                   # telegram.session (локально)
+├── docker-compose.yml
+├── .env.example
+├── README.md
+└── AGENTS.md
+```
+
+### 2.2. Центр аналитики (ЦА)
+
+**Ответственность:** владение данными, UI, постановка jobs, приём ingest, отдача срезов ПИ.
+
+**Backend** (`centers/analytics/api/`):
+
+| Слой | Содержание |
+|------|------------|
+| Routers | `/api/*` (карта, объекты), `/admin/*`, `/internal/*`, `/api/v1/*` |
+| Models | `Event`, `ParseJob`, `MapObject`, `ObjectMedia`, `Consumer`, `ConsumerFilter` |
+| Services | enqueue → Redis, scheduler, ingest + map sync, analytics queries |
+| Auth internal | `X-Internal-Token` |
+| Auth ПИ | Bearer API key (SHA-256 hash в БД) |
+
+**Frontend** (`centers/analytics/frontend/`):
+
+| Маршрут | Экран |
+|---------|-------|
+| `/` | Карта событий и ручных объектов |
+| `/parsers` | CRUD парсеров (`telegram` / `crawl4ai` / `viina`) |
+| `/events` | Список событий |
+| `/analytics` | KPI и таймлайны |
+| `/consumers` | Подписчики ПИ |
+
+Nginx проксирует `/api/`, `/admin/`, `/internal/` на `ca-api:8000`.
+
+Подробности: [centers/analytics/ARCHITECTURE.md](../centers/analytics/ARCHITECTURE.md).
+
+### 2.3. Центр парсинга (ЦП)
+
+**Ответственность:** забрать данные из внешнего источника и вернуть список `IngestEventItem`.
+
+Адаптерный контракт:
+
+```python
+async def run(job_id, source_config, *, ctx) -> tuple[list[dict], str | None]
+```
+
+| source_type | Очередь | Контейнер |
+|-------------|---------|-----------|
+| `telegram` | `cp:jobs:telegram` | `cp-workers` |
+| `crawl4ai` | `cp:jobs:web` | `cp-workers-web` |
+| `viina` | `cp:jobs:nlp` | `cp-workers-nlp` |
+
+Дополнительно для Telegram: **real-time listener** (Telethon `NewMessage` / `Album`) параллельно с batch, если `TELEGRAM_LISTENER_ENABLED=true`.
+
+Опционально: `extract_mode: llm` (DeepSeek) для telegram/crawl4ai в batch.
+
+**Профиль + канал:** UI `/parser-profiles` (CRUD, Generate/Preview для heuristic; instruction/schema/Preview для llm) и `/channels`; связка на `/parsers` → `ParseJob` с FK. При enqueue ЦА flatten по `ParserProfile.kind`:
+
+- `heuristic` → `extract_mode=profile` + `heuristic_profile` (runtime без LLM; listener поддерживает);
+- `llm` → `extract_mode=llm` + `extract_schema` / `instruction` / `required_fields` (DeepSeek на каждый пост в batch; listener пока fallback на heuristic).
+
+Целевые поля — фиксированная схема Event/`IngestEventItem`. Кастомные пользовательские таблицы — roadmap.
+
+Подробности: [centers/parsing/ARCHITECTURE.md](../centers/parsing/ARCHITECTURE.md), [centers/analytics/ARCHITECTURE.md](../centers/analytics/ARCHITECTURE.md).
+
+### 2.4. Distribution (ПИ)
+
+Отдельного центра нет. Внешний клиент:
+
+1. Получает API-ключ в UI «ПИ» / admin API.
+2. Вызывает `GET /api/v1/events` с `Authorization: Bearer <key>`.
+3. Получает события с учётом фильтров consumer (регионы, темы, `date_from`).
+
+Подробности: [centers/analytics/DISTRIBUTION.md](../centers/analytics/DISTRIBUTION.md).
+
+### 2.5. Contracts
+
+Единый источник правды для формы обмена:
+
+| Модуль | Назначение |
+|--------|------------|
+| `contracts/jobs.py` | `JobPayload` в Redis |
+| `contracts/ingest.py` | `IngestEventItem` / payload ingest |
+| `contracts/sources.py` | Валидация `source_config` по `source_type` |
+| `contracts/queues.py` | `SOURCE_FAMILY` → ключ очереди |
+| `contracts/heuristic_profile.py` | Статичный профиль конструктора + `apply_profile` |
+| `contracts/llm_profile.py` | Reusable LLM instruction/schema + `match_llm_required` |
+
+Правило: меняете форму события / конфиг источника / очередь — сначала `contracts/`, потом CA/CP/UI.
+
+Подробности: [contracts.md](contracts.md).
+
+### 2.6. Данные в PostgreSQL (ядро)
+
+```mermaid
+erDiagram
+  ParseJob ||--o{ Event : produces
+  Event ||--o| MapObject : may_create
+  MapObject ||--o{ ObjectMedia : has
+  Consumer ||--o| ConsumerFilter : has
+
+  ParseJob {
+    int id
+    string source_type
+    json source_config
+    int interval_seconds
+    bool is_active
+    string status
+  }
+  Event {
+    int id
+    string source_url UK
+    string source_type
+    float latitude
+    float longitude
+    string region
+    string topic
+  }
+  MapObject {
+    int id
+    int event_id FK
+    float latitude
+    float longitude
+  }
+  Consumer {
+    int id
+    string name
+    string api_key_hash
+  }
+```
+
+- **ParseJob** — конфигурация парсера и статус batch-прогона.
+- **Event** — нормализованное событие; уникальность по `source_url`.
+- **MapObject** — точка на карте (из события с координатами или ручная).
+- **Consumer** — внешний подписчик ПИ.
+
+### 2.7. Планировщик
+
+В `ca-api` фоновый scheduler (~каждые 30 с):
+
+- берёт активные `ParseJob` со статусом `completed` / `failed`;
+- если прошло ≥ `interval_seconds` с `last_run_at` — снова ставит в Redis (`queued`).
+
+Так работают периодические парсеры без cron снаружи.
+
+### 2.8. Секреты и тома
+
+| Артефакт | Где | Заметка |
+|----------|-----|---------|
+| `.env` | корень | `TELEGRAM_*`, опционально `DEEPSEEK_*` |
+| `data/telegram.session` | volume `./data` → `/data` в `cp-workers` | Telethon SQLite |
+| `pgdata` | Docker volume | PostgreSQL |
+| `ca_uploads` | Docker volume | медиа объектов карты |
+
+---
+
+## 3. Куда смотреть при типовых задачах
+
+| Задача | Документ / код |
+|--------|----------------|
+| Новый источник парсинга | skill `add-parser-adapter` + `contracts/sources.py` + `queues.py` |
+| Изменить поля события | `contracts/ingest.py` → ingest CA → map/ПИ |
+| Карта / UI | `centers/analytics/frontend/src/` |
+| Admin API | `centers/analytics/api/app/routers/admin.py` |
+| Ingest / дедуп | `centers/analytics/api/app/services/ingest.py` |
+| Telegram listener | `centers/parsing/.../telegram_listener.py` |
+| Ключи для внешних систем | [DISTRIBUTION.md](../centers/analytics/DISTRIBUTION.md) |

docs/contracts.md

@@ -0,0 +1,99 @@
+# Contracts — общие схемы
+
+Каталог [`contracts/`](../contracts/) — shared truth для формы данных и маршрутизации между ЦА и ЦП.
+
+Правило платформы: **сначала contracts**, потом реализация в CA/CP/UI.
+
+---
+
+## Модули
+
+| Файл | Назначение |
+|------|------------|
+| [`jobs.py`](../contracts/jobs.py) | Контракт задания. Payload задания в Redis |
+| [`ingest.py`](../contracts/ingest.py) |Контракт результата. Событие и пакет ingest |
+| [`sources.py`](../contracts/sources.py) | Контракт настроек парсера.  Схемы `source_config` по `source_type` |
+| [`queues.py`](../contracts/queues.py) | Контракт доставки задания нужному воркеру. `source_type` → family → Redis key  |
+| [`heuristic_profile.py`](../contracts/heuristic_profile.py) | Статичные правила extract_mode=profile |
+| [`llm_profile.py`](../contracts/llm_profile.py) | Reusable LLM instruction/schema + required_fields + `multi_event` |
+
+ЦА при admin CRUD валидирует конфиг через `parse_source_config`.  
+ЦП адаптеры должны отдавать dict, совместимые с `IngestEventItem`.
+
+> На практике CA дублирует часть DTO в `app/schemas.py` для FastAPI; при изменении формы сверяйте оба места и UI.
+
+---
+
+## JobPayload (`jobs.py`)
+
+```python
+class JobPayload(BaseModel):
+    job_id: int
+    source_type: str
+    source_config: dict = {}
+```
+
+Уходит в Redis как JSON. Семейство очереди выбирается по `source_type`, не по содержимому config.
+
+---
+
+## Ingest (`ingest.py`)
+
+Ключевые поля `IngestEventItem`:
+
+| Поле | Обязательность | Заметка |
+|------|----------------|---------|
+| `source_url` | да | Стабильный URL; дедуп в ЦА |
+| `source_type` | да (часто default) | Должен соответствовать адаптеру |
+| `raw_text` / `title` / `description` | нет | Текст события |
+| `latitude` / `longitude` | нет | Без них точка на карту не создаётся |
+| `event_date`, `locality`, `region`, `topic` | нет | Фильтры карты и ПИ |
+| `tags`, `metadata` | нет | Расширения |
+
+Пакет: `{ job_id?, events: [...] }`.  
+Listener добавляет флаг `listener: true` на стороне CA API (см. internal schemas).
+
+---
+
+## Source config (`sources.py`)
+
+| source_type | Модель | Главные поля |
+|-------------|--------|--------------|
+| `telegram` | `TelegramSourceConfig` | `channel`, `limit`, `extract_mode` (`heuristic`\|`llm`\|`profile`), `heuristic_profile`, `extract_schema`, `instruction`, `required_fields` (llm gate) |
+| `crawl4ai` | `Crawl4AISourceConfig` | `urls`, `extract_mode`, `extract_schema`, `domain_profile` |
+| `viina` | `ViinaSourceConfig` | `urls` / `texts`, `input_mode` |
+
+Реестр: `CONFIG_MODELS` + `parse_source_config(source_type, raw)`.
+
+Профили в БД ЦА (`ParserProfile`):
+
+| kind | Хранилище | Flatten в Redis |
+|------|-----------|-----------------|
+| `heuristic` | `heuristic_profile` (`contracts/heuristic_profile.py`) | `extract_mode=profile` |
+| `llm` | `llm_profile` (`contracts/llm_profile.py`: instruction, extract_schema, required_fields, multi_event) | `extract_mode=llm` (+ `#eN` URLs if multi) |
+
+`required_fields` — обязательные поля; пустой список = без фильтра (для llm остаётся только `is_event`). Канал + профиль живут в ЦА; CP получает только плоский `source_config`.
+
+---
+
+## Очереди (`queues.py`)
+
+```text
+SOURCE_FAMILY = {
+  "telegram": "telegram",  # cp:jobs:telegram
+  "crawl4ai": "web",       # cp:jobs:web
+  "viina": "nlp",          # cp:jobs:nlp
+}
+```
+
+- `queue_key_for_source(source_type)` — куда enqueue из ЦА.
+- Legacy-ключ `cp:jobs` ещё может дренироваться telegram-воркерами (совместимость).
+
+При добавлении нового `source_type` обязательно:
+
+1. схема в `sources.py`;
+2. запись в `SOURCE_FAMILY`;
+3. адаптер + registry + (при необходимости) новый Docker-сервис;
+4. поля формы в `ParsersView.vue`.
+
+Чеклист: [`.cursor/skills/add-parser-adapter/`](../.cursor/skills/add-parser-adapter/).

docs/data-flow.md

@@ -0,0 +1,134 @@
+# Поток данных MapMil
+
+Пошаговый путь события от настройки парсера до карты и внешнего API.
+
+См. также: [обзор архитектуры](architecture-overview.md), [ЦП](../centers/parsing/ARCHITECTURE.md).
+
+---
+
+## Схема end-to-end
+
+```mermaid
+sequenceDiagram
+  participant UI as Admin UI
+  participant CA as ca-api
+  participant DB as PostgreSQL
+  participant R as Redis
+  participant CP as cp-workers
+  participant Src as Source
+  participant Map as Map UI
+  participant PI as External client
+
+  UI->>CA: POST /admin/jobs
+  CA->>DB: INSERT ParseJob queued
+  CA->>R: RPUSH cp:jobs:family
+  R->>CP: BLPOP
+  CP->>CA: PATCH /internal/jobs/id running
+  CP->>Src: fetch / listen
+  Src-->>CP: raw posts/pages
+  CP->>CA: POST /internal/ingest
+  CA->>DB: INSERT Event if new source_url
+  CA->>DB: sync MapObject if coords
+  CA->>CP: 200 ingested/skipped
+  CP->>CA: PATCH job completed or failed
+  Map->>CA: GET /api/map/objects
+  PI->>CA: GET /api/v1/events Bearer
+```
+
+---
+
+## 1. Создание парсера (ЦА)
+
+1. Пользователь в UI `/parsers` или `POST /admin/jobs` передаёт:
+   - `source_type` (`telegram` | `crawl4ai` | `viina`);
+   - `source_config` (валидируется `contracts/sources.py`);
+   - опционально `interval_seconds`, `is_active`.
+2. ЦА пишет строку `ParseJob` со статусом `queued`.
+3. ЦА делает `RPUSH` в Redis-ключ `cp:jobs:{family}` (`contracts/queues.py`).
+
+Примеры `source_config`:
+
+```json
+{ "channel": "example", "limit": 50, "extract_mode": "heuristic" }
+```
+
+```json
+{ "urls": ["https://news.example/"], "extract_mode": "llm" }
+```
+
+---
+
+## 2. Batch-обработка (ЦП)
+
+1. Воркер нужного семейства делает `BLPOP` своей очереди.
+2. `PATCH /internal/jobs/{id}?status=running`.
+3. Registry выбирает адаптер по `source_type` → `adapter.run(...)`.
+4. Адаптер возвращает список dict в форме `IngestEventItem` (+ опциональный warning).
+5. ЦП шлёт `POST /internal/ingest` с `{ job_id, events }`.
+6. При успехе/ошибке — `PATCH` статуса `completed` / `failed`.
+
+Если job попал не в ту очередь (чужой family), воркер **перекладывает** его в правильную очередь.
+
+---
+
+## 3. Real-time Telegram (параллельно)
+
+Только `cp-workers` при `TELEGRAM_LISTENER_ENABLED=true`:
+
+1. Listener опрашивает `GET /internal/listener/subscriptions` (активные telegram-jobs).
+2. Telethon получает `NewMessage` / `Album`.
+3. Пост парсится (эвристика) → один event → `POST /internal/ingest` с `listener: true`.
+4. Статус `ParseJob` от listener **не** переводится в `completed` (в отличие от batch).
+
+---
+
+## 4. Ingest в ЦА
+
+Сервис `centers/analytics/api/app/services/ingest.py`:
+
+1. Для каждого item проверяет уникальность `source_url`.
+2. Уже есть → skip (дедуп).
+3. Нет → INSERT `Event`.
+4. Если есть `latitude` / `longitude` → создать/обновить связанный `MapObject`.
+5. Batch (`listener` не true) обновляет статус job и `last_run_at`.
+
+---
+
+## 5. Карта
+
+1. UI периодически (около 30 с) и по фильтрам вызывает `GET /api/map/objects`.
+2. Ответ — точки с данными события (даты, регион, тема, источник).
+3. Ручные объекты карты живут в `map_objects` без обязательного `event_id` (CRUD через UI / `/api/objects`).
+
+---
+
+## 6. Внешний потребитель (ПИ)
+
+1. В UI `/consumers` создаётся consumer + API-ключ (в БД хранится hash).
+2. Клиент: `GET /api/v1/events` + `Authorization: Bearer <key>`.
+3. ЦА применяет `ConsumerFilter` (регионы, темы, `date_from`) и отдаёт срез событий.
+
+---
+
+## 7. Периодический перезапуск
+
+Scheduler в `ca-api` (~30 с):
+
+- активный job;
+- статус `completed` или `failed`;
+- прошло ≥ `interval_seconds` с последнего запуска  
+
+→ снова `queued` + enqueue в Redis.
+
+---
+
+## Точки отказа (для отладки)
+
+| Симптом | Куда смотреть |
+|---------|----------------|
+| Job висит в `queued` | Redis, нужный `cp-workers*`, `WORKER_FAMILIES` |
+| Job `failed` | `last_error` в admin, логи контейнера адаптера |
+| Событий 0 при успехе | Парсер/LLM не извлёк поля; дедуп по `source_url` |
+| Нет точек на карте | Нет координат у Event; фильтры даты/региона в UI |
+| 401 на `/api/v1/events` | Ключ, `is_active` consumer |
+| Telegram auth error | `data/telegram.session`, `TELEGRAM_API_ID/HASH` |

docs/local-dev.md

@@ -0,0 +1,133 @@
+# Локальная разработка
+
+## Требования
+
+- Docker + Docker Compose
+- Файл `.env` (из `.env.example`)
+- Для Telegram: `data/telegram.session` + `TELEGRAM_API_ID` / `TELEGRAM_API_HASH`
+- Опционально: `DEEPSEEK_API_KEY` для `extract_mode: llm` (воркеры) и Generate профиля на `ca-api`
+
+## Быстрый старт
+
+```bash
+cp .env.example .env
+# заполнить TELEGRAM_* (и при необходимости DEEPSEEK_*)
+
+mkdir -p data
+# положить telegram.session в data/
+
+docker compose up --build
+```
+
+UI: http://localhost:8080  
+Health: `curl http://localhost:8080/api/health`
+
+## Сервисы
+
+| Сервис | Порт снаружи | Когда нужен |
+|--------|--------------|-------------|
+| `ca-frontend` | 8080 | всегда (UI) |
+| `ca-api` | внутренний 8000 | всегда |
+| `ca-db` / `redis` | — | всегда |
+| `cp-workers` | — | Telegram |
+| `cp-workers-web` | — | Crawl4AI |
+| `cp-workers-nlp` | — | VIINA |
+| `cp-vpn` | внутренний 1080 | SOCKS из subscription (вкладка VPN) |
+
+Пересборка одного воркера:
+
+```bash
+docker compose up --build cp-workers-web
+```
+
+Проверка compose-файла:
+
+```bash
+docker compose config
+```
+
+## Telegram-сессия
+
+- Путь в контейнере: `/data/telegram.session` (`TELEGRAM_SESSION_PATH`)
+- Локально: `./data` монтируется в `cp-workers`
+- Файл **не** коммитить (`.gitignore`)
+- API id/hash в `.env` должны совпадать с теми, под которыми создавалась сессия
+
+## VPN / прокси (вкладка VPN)
+
+Основной путь: UI → **VPN** (`/vpn`).
+
+1. Включите «Прокси включён».
+2. Способ **Subscription URL** — вставьте ссылку подписки (не коммитьте её в git).
+3. Отметьте источники (`telegram`, при необходимости `crawl4ai` / `viina`).
+4. Сохраните. Сервис `cp-vpn` (mihomo) подтянет подписку и отдаст SOCKS на `cp-vpn:1080`; воркеры читают `/internal/vpn`.
+
+Режимы **SOCKS5** / **HTTP** задают host:port напрямую (без `cp-vpn` для Telegram).
+
+### Fallback: локальный Happ/xray (без вкладки VPN)
+
+Если VPN в UI выключен, воркеры используют `TELEGRAM_PROXY_*` из `.env`:
+
+1. Остановите Telegram-воркер, чтобы не делить session-файл:
+   `docker compose stop cp-workers`
+2. В `.env`: `TELEGRAM_PROXY_TYPE=socks5`, `TELEGRAM_PROXY_HOST=127.0.0.1`, `TELEGRAM_PROXY_PORT=10808` (Happ/xray).
+3. Проброс SOCKS в Docker (xray слушает только `127.0.0.1:10808`):
+   ```bash
+   socat TCP-LISTEN:11080,bind=0.0.0.0,fork,reuseaddr TCP:127.0.0.1:10808 &
+   ```
+   Для auth из контейнера с host-сетью: `TELEGRAM_PROXY_HOST=127.0.0.1`.
+4. Авторизация (интерактивно, код из Telegram):
+   ```bash
+   docker run --rm -it --network host \
+     --env-file .env \
+     -e TELEGRAM_SESSION_PATH=/data/telegram.session \
+     -e TELEGRAM_PROXY_HOST=127.0.0.1 \
+     -v "$PWD/data:/data" \
+     -v "$PWD/scripts:/scripts:ro" \
+     mapmil-cp-workers \
+     python /scripts/telegram_auth.py
+   ```
+5. `docker compose up -d cp-workers`
+
+Без сессии batch/listener Telegram не авторизуются; остальные адаптеры могут работать.
+
+## Полезные curl
+
+Создать telegram-парсер:
+
+```bash
+curl -X POST http://localhost:8080/admin/jobs \
+  -H 'Content-Type: application/json' \
+  -d '{"source_type":"telegram","source_config":{"channel":"example","limit":50}}'
+```
+
+Срез ПИ (подставьте ключ из seed / UI):
+
+```bash
+curl -H "Authorization: Bearer test-pi-api-key-change-me" \
+  "http://localhost:8080/api/v1/events"
+```
+
+## Логи
+
+```bash
+docker compose logs -f ca-api
+docker compose logs -f cp-workers
+docker compose logs -f cp-workers-web
+```
+
+## Типичные проблемы
+
+| Проблема | Действие |
+|----------|----------|
+| Frontend 000 / нет контейнеров | `docker compose up -d` (без `--build`, если registry недоступен, но образы уже есть) |
+| Job не берётся | Смотреть `ENABLED_ADAPTERS` / `WORKER_FAMILIES` нужного сервиса |
+| LLM не работает | `DEEPSEEK_API_KEY` в `.env`, перезапуск `cp-workers` / `cp-workers-web` |
+| Generate профиля 503 | `DEEPSEEK_API_KEY` в `.env` + `env_file` у `ca-api`, перезапуск `ca-api` |
+| Изменения UI не видны | Пересобрать `ca-frontend` |
+
+## Границы при разработке
+
+- Не добавлять прямой доступ к БД из CP workers.
+- Не коммитить `.env`, `data/`, API-ключи.
+- Изменение формы данных — через `contracts/` (см. [contracts.md](contracts.md)).

scripts/telegram_auth.py

@@ -0,0 +1,115 @@
+#!/usr/bin/env python3
+"""Интерактивная авторизация Telethon → data/telegram.session.
+
+Запуск через образ cp-workers и host-сеть (локальный xray/Happ на 127.0.0.1:10808):
+
+  docker compose stop cp-workers
+  docker run --rm -it --network host \\
+    --env-file .env \\
+    -e TELEGRAM_SESSION_PATH=/data/telegram.session \\
+    -e TELEGRAM_PROXY_HOST=127.0.0.1 \\
+    -v \"$PWD/data:/data\" \\
+    -v \"$PWD/scripts:/scripts:ro\" \\
+    mapmil-cp-workers \\
+    python /scripts/telegram_auth.py
+  docker compose up -d cp-workers
+"""
+
+from __future__ import annotations
+
+import asyncio
+import os
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+WORKERS = ROOT / "centers" / "parsing" / "workers"
+if WORKERS.exists() and str(WORKERS) not in sys.path:
+    sys.path.insert(0, str(WORKERS))
+
+
+def _load_dotenv(path: Path) -> None:
+    if not path.is_file():
+        return
+    for line in path.read_text(encoding="utf-8").splitlines():
+        line = line.strip()
+        if not line or line.startswith("#") or "=" not in line:
+            continue
+        key, _, value = line.partition("=")
+        key = key.strip()
+        value = value.strip().strip("'").strip('"')
+        os.environ.setdefault(key, value)
+
+
+async def main() -> int:
+    _load_dotenv(ROOT / ".env")
+
+    # Prefer repo-local session when running outside Docker /data mount.
+    if not os.environ.get("TELEGRAM_SESSION_PATH"):
+        local = ROOT / "data" / "telegram.session"
+        os.environ["TELEGRAM_SESSION_PATH"] = str(local)
+
+    from workers.sources.telegram_settings import (
+        create_client,
+        describe_connection,
+        get_api_credentials,
+        get_session_path,
+    )
+
+    api_id, api_hash = get_api_credentials()
+    session_path = get_session_path()
+    Path(session_path).parent.mkdir(parents=True, exist_ok=True)
+
+    print(f"Сессия: {session_path}")
+    print(f"Подключение: {describe_connection()}")
+
+    client = create_client(session_path, api_id, api_hash)
+    await client.connect()
+
+    if await client.is_user_authorized():
+        me = await client.get_me()
+        print(
+            f"Уже авторизовано: id={me.id} "
+            f"username={getattr(me, 'username', None) or '—'} "
+            f"phone={getattr(me, 'phone', None) or '—'}"
+        )
+        await client.disconnect()
+        return 0
+
+    print("Сессия не авторизована — вход в Telegram.")
+    phone = input("Номер телефона (+7...): ").strip()
+    if not phone:
+        print("Номер не указан.", file=sys.stderr)
+        await client.disconnect()
+        return 1
+
+    await client.send_code_request(phone)
+    code = input("Код из Telegram/SMS: ").strip()
+    try:
+        await client.sign_in(phone=phone, code=code)
+    except Exception as exc:
+        # 2FA
+        from telethon.errors import SessionPasswordNeededError
+
+        if not isinstance(exc, SessionPasswordNeededError):
+            raise
+        password = input("Пароль 2FA: ").strip()
+        await client.sign_in(password=password)
+
+    me = await client.get_me()
+    print(
+        f"Готово. Авторизован: id={me.id} "
+        f"username={getattr(me, 'username', None) or '—'} "
+        f"phone={getattr(me, 'phone', None) or '—'}"
+    )
+    print(f"Файл сессии: {session_path}")
+    await client.disconnect()
+    return 0
+
+
+if __name__ == "__main__":
+    try:
+        raise SystemExit(asyncio.run(main()))
+    except KeyboardInterrupt:
+        print("\nОтменено.", file=sys.stderr)
+        raise SystemExit(130)