sources: авто-расширение из публичных API без ключей (46 живых), universal generic_json, bsky/mastodon, --only по префиксу, auto_cycle

README.md

@@ -198,7 +198,7 @@ python -m omni export --format jsonl --out data/dump.jsonl
 
 ## Реестр источников
 
-`omni/sources.py` + `omni/sources_ext.py` — **589 источников в 9 категориях**, у каждого свой адаптер и флаг вкл/выкл:
+`omni/sources.py` + `omni/sources_ext.py` + `omni/sources_auto.py` — **1072 источника в 11 категориях**, у каждого свой адаптер и флаг вкл/выкл. Из них **467 abuse** (включая 46 авто-подключённых `aa_`) и **33 в активной авто-группе `aa_`** (остальные 13 авто-кандидатов совпали по url с уже существующими источниками и были схлопнуты дедупом):
 
 - **papers 49** — arxiv по 25+ разделам (cs.AI/LG/CL/CV/RO/IR/CR/SE/NE/MA/DC/DB/HC/SD/CY, stat.ML, eess.AS/SP/IV, math.OC, q-bio.NC, cs.GR, cs.AR, cs.PF), HF daily papers, OpenAlex, Semantic Scholar, Papers with Code, Connected Papers
 - **company 95** — ленты OpenAI, Google AI/Research/DeepMind, Microsoft Research, NVIDIA, Apple ML, Amazon Science, HF, Meta AI, Mistral, Cohere, Together, Perplexity, Groq, Cerebras, Stability, Databricks, Anyscale, Modal, W&B, LangChain, LlamaIndex, vLLM, Replicate, Fireworks, SambaNova, Scale, xAI, Salesforce, AWS ML, Azure AI, Anthropic, AI2, EleutherAI, LAION, Runway, ElevenLabs, Deepgram, AssemblyAI, Pinecone, Weaviate, Qdrant, Chroma, Unsloth, Axolotl
@@ -212,6 +212,37 @@ python -m omni export --format jsonl --out data/dump.jsonl
 
 Адаптеры: `rss`, `generic_json`, `or_models`, `cisa_kev`, `nvd`, `json` (HF, HN, OpenAlex, S2, Discourse), `html` (список + детальная), `github_org`. Медленные хосты (reddit, github, HF, arxiv, openrouter) обходятся последовательно с паузой, остальные — в 8 потоков. Для arxiv RSS, который часто отдаёт пустое, есть fallback на официальный API (`export.arxiv.org/api/query`).
 
+## Авто-расширение источников (публичные API без ключей)
+
+`omni/sources_auto.py` **генерируется** `tools/auto_expand.py` — руками не править, правки перезапишутся. Пул кандидатов лежит в `data/source_candidates.json`, здоровье — в `data/source_health.json`.
+
+```
+python tools/auto_expand.py --probe          # проверить всех: имя | LIVE/DEAD | items | заголовки
+python tools/auto_expand.py --write          # probe + пересобрать omni/sources_auto.py (живые, items>0)
+python tools/auto_expand.py --write-only     # пересобрать из последнего health (без сети)
+python tools/auto_expand.py --discover       # добрать бесплатные (Auth=No) эндпоинты из public-apis
+python tools/auto_expand.py --add NAME URL --adapter rss   # новый кандидат одной строкой
+python tools/auto_cycle.py                   # полный цикл: probe → write → feed → индекс → коммит/пуш
+```
+
+Правила: кандидат попадает в реестр только если адаптер реально вернул `items > 0`; после 3 прогонов подряд с нулём источник выкидывается из генерации автоматически. Группы в сгенерированном файле: `json`, `reddit`, `social`, `discourse`, `rss`, `tg`.
+
+Универсальный `generic_json` понимает описание в `extra`: `rows` (путь к списку через точку), `title_key`, `url_key`, `text_key`, `date_key`, `base` + `slug_key` (собрать ссылку из слага), `deref` (спуститься в подобъект). Поэтому большинство новых JSON-источников — одна строка в `data/source_candidates.json`, без нового кода. Отдельные адаптеры только там, где схема нестандартная: `bsky_search` (Bluesky public API, без токена), `mastodon_tag`, `tg_web` (`t.me/s/<channel>`), `discourse` (`latest.json`).
+
+Подключено в этом апдейте (46 живых из 52 проверенных, `пусто 0/33 | собрано 458, новых 408`):
+
+- **JSON без ключа** — dev.to, Lobsters, npm search (llm / mcp / free api keys), crates.io, Docker Hub, Stack Overflow (openai / prompt-injection), GitLab (new/llm), Hugging Face spaces + models trending
+- **Reddit** — через `api.pullpush.io` (архив Reddit, без OAuth): LocalLLaMA, FreeAI, OpenAIAPI, ChatGPT, selfhosted
+- **Bluesky** — `public.api.bsky.app` searchPosts: free tokens api, openrouter, free api, ai keys, llm api
+- **Mastodon** — теги ai / openai / llm
+- **Discourse** — community.openai, discuss.huggingface, forum.cursor, discuss.python, discuss.fly, forum.obsidian
+- **RSS** — Habr, The Decoder, MarkTechPost, OpenAI news, Simon Willison, HF blog, TechCrunch AI, 52pojie
+- **Telegram-превью** — ai_newz, data_secrets, neurohive, gpt4telegram, dataleak, denissexy
+
+`--only` в `omni feed` понимает префикс: `python -m omni feed --only aa_` берёт всю авто-группу. Для `github_search` поднимается лимит, если в `.env` есть `GITHUB_TOKEN` (анонимно 10 запросов/мин, с токеном 30).
+
+Крон (каждые 6 часов, `tools/auto_cycle.py`): проверка кандидатов → пересборка `sources_auto.py` → докачка только при изменениях → пересборка abuse-индекса → коммит `auto: источники N live` и пуш в gitea/github. Лог — `data/auto_cycle.log`.
+
 ## Панель и команды
 
 ```

omni/feeds.py

@@ -484,41 +484,123 @@ def _json_of(url, timeout=35):
 
 
 def _generic_json(src, limit):
-    """Любой JSON-эндпоинт: ищем список объектов и вытаскиваем title/url по частым ключам."""
+    """Любой JSON-эндпоинт: список объектов (+extra: rows/title_key/url_key/text_key/date_key/base/slug_key/deref)."""
     d = _json_of(src.url)
+    ex = src.extra or {}
     rows = []
     if isinstance(d, list):
         rows = d
     elif isinstance(d, dict):
-        for key in ("data", "items", "results", "papers", "models", "hits", "notes", "posts", "records"):
+        path = ex.get("rows")
-            v = d.get(key)
+        if path:
-            if isinstance(v, list):
+            cur = d
-                rows = v
+            for part in str(path).split("."):
-                break
+                cur = cur.get(part) if isinstance(cur, dict) else None
+            if isinstance(cur, list):
+                rows = cur
+        if not rows:
+            for key in ("data", "items", "results", "papers", "models", "hits", "notes",
+                        "posts", "records", "objects", "crates", "hackathons"):
+                v = d.get(key)
+                if isinstance(v, list):
+                    rows = v
+                    break
         if not rows:
             for v in d.values():
                 if isinstance(v, list) and v and isinstance(v[0], dict):
                     rows = v
                     break
     out = []
+    deref = ex.get("deref")
+    base = str(ex.get("base") or "")
+    title_keys = ([ex["title_key"]] if ex.get("title_key") else []) + \
+                 ["title", "name", "paper_title", "full_name", "id"]
+    url_keys = ([ex["url_key"]] if ex.get("url_key") else []) + \
+               ["url", "link", "html_url", "paper_url", "web_url"]
+    text_keys = ([ex["text_key"]] if ex.get("text_key") else []) + \
+                ["abstract", "summary", "description", "body", "content"]
+    date_keys = ([ex["date_key"]] if ex.get("date_key") else []) + \
+                ["published", "created", "created_at", "date", "published_at"]
     for r in rows[:limit]:
         if not isinstance(r, dict):
             continue
-        title = r.get("title") or r.get("name") or r.get("paper_title") or r.get("full_name") or r.get("id")
+        if deref and isinstance(r.get(deref), dict):
-        url = r.get("url") or r.get("link") or r.get("html_url") or r.get("paper_url") or r.get("id")
+            r = r[deref]
+        title = ""
+        for k in title_keys:
+            v = r.get(k)
+            if isinstance(v, str) and v.strip():
+                title = v.strip()
+                break
+        url = ""
+        for k in url_keys:
+            v = r.get(k)
+            if isinstance(v, str) and v.strip():
+                url = v.strip()
+                break
+        if base and (not url or not url.startswith("http")):
+            slug = ""
+            if ex.get("slug_key"):
+                slug = str(r.get(ex["slug_key"]) or "")
+            url = base.rstrip("/") + "/" + (url or slug).lstrip("/")
         if isinstance(url, str) and url and not url.startswith("http"):
             if "github.com" in src.url:
                 url = "https://github.com/" + url.lstrip("/")
             elif "huggingface.co" in src.url:
                 url = "https://huggingface.co/" + url.lstrip("/")
-        text = r.get("abstract") or r.get("summary") or r.get("description") or r.get("body") or r.get("content") or ""
+        text = ""
-        if isinstance(text, dict):
+        for k in text_keys:
-            text = json.dumps(text, ensure_ascii=False)[:400]
+            v = r.get(k)
-        published = r.get("published") or r.get("created") or r.get("created_at") or r.get("date") or ""
+            if isinstance(v, str) and v.strip():
+                text = v
+                break
+            if isinstance(v, (dict, list)):
+                text = json.dumps(v, ensure_ascii=False)[:400]
+                break
+        published = ""
+        for k in date_keys:
+            v = r.get(k)
+            if isinstance(v, str) and v.strip():
+                published = v.strip()
+                break
         if not title:
             continue
-        out.append(_item(src, "entry", str(url or ""), str(title), str(text)[:800],
+        out.append(_item(src, "entry", str(url or ""), _txt(title)[:180], _txt(text)[:800],
-                         str(r.get("author") or r.get("owner") or ""), str(published)[:24]))
+                         str(r.get("author") or r.get("owner") or r.get("by") or ""),
+                         str(published)[:24]))
+    return out
+
+
+def _bsky_search(src, limit):
+    """Bluesky: публичный поиск постов без ключа (api.bsky.app)."""
+    d = _json_of(src.url)
+    out = []
+    for p in (d.get("posts") or [])[:limit]:
+        if not isinstance(p, dict):
+            continue
+        rec = p.get("record") or {}
+        txt = _txt(str(rec.get("text") or ""))
+        author = str((p.get("author") or {}).get("handle") or "")
+        uri = str(p.get("uri") or "")
+        rkey = uri.rsplit("/", 1)[-1] if uri else ""
+        url = "https://bsky.app/profile/%s/post/%s" % (author, rkey) if author and rkey else uri
+        out.append(_item(src, "post", url, txt[:150], txt[:800], author,
+                         str(rec.get("createdAt") or "")[:24]))
+    return out
+
+
+def _mastodon_tag(src, limit):
+    """Mastodon: публичная лента по тегу, без токена."""
+    d = _json_of(src.url)
+    rows = d if isinstance(d, list) else []
+    out = []
+    for r in rows[:limit]:
+        if not isinstance(r, dict):
+            continue
+        txt = _txt(str(r.get("content") or ""))
+        out.append(_item(src, "post", str(r.get("url") or ""), txt[:150], txt[:800],
+                         str((r.get("account") or {}).get("acct") or ""),
+                         str(r.get("created_at") or "")[:24]))
     return out
 
 
@@ -840,6 +922,9 @@ def _github_search(src, limit):
            + "&sort=" + str(sort) + "&order=desc&per_page=" + str(max(1, min(int(limit), 50))))
     hdr = {"Accept": "application/vnd.github+json", "User-Agent": "omni-parse"}
     tok = ((config.CONFIG.get("keys") or {}).get("github") or [])
+    if not tok:
+        single = config.KEYS.get("GITHUB_TOKEN")
+        tok = [single] if single else []
     if isinstance(tok, list) and tok:
         hdr["Authorization"] = "Bearer " + str(tok[0])
     r = net.get(url, headers=hdr)
@@ -917,6 +1002,8 @@ ADAPTERS = {
     "newapi_pricing": _newapi_pricing,
     "github_search": _github_search,
     "hn_search": _hn_search,
+    "bsky_search": _bsky_search,
+    "mastodon_tag": _mastodon_tag,
 
 }
 

omni/sources.py

@@ -21,6 +21,7 @@ from .sources_ext5 import EXTRA6
 from .sources_fresh import fresh as _fresh_sources
 from .sources_quality import quality_sources as _quality_sources
 from .sources_x import X_SOURCES
+from .sources_auto import AUTO as AUTO_SOURCES
 
 
 @dataclass
@@ -204,7 +205,7 @@ def _dedup(rows):
     return out
 
 
-SOURCES = _dedup(SOURCES_BASE + [Source(**d) for d in EXTRA] + [Source(**d) for d in EXTRA2] + [Source(**d) for d in EXTRA3] + [Source(**d) for d in ABUSE_LIVE] + [Source(**d) for d in EXTRA4] + [Source(**d) for d in EXTRA5] + [Source(**d) for d in EXTRA6] + [Source(**d) for d in ABUSE_DEEP] + [Source(**d) for d in TG_ABUSE] + [Source(**d) for d in X_SOURCES])
+SOURCES = _dedup(SOURCES_BASE + [Source(**d) for d in EXTRA] + [Source(**d) for d in EXTRA2] + [Source(**d) for d in EXTRA3] + [Source(**d) for d in ABUSE_LIVE] + [Source(**d) for d in EXTRA4] + [Source(**d) for d in EXTRA5] + [Source(**d) for d in EXTRA6] + [Source(**d) for d in ABUSE_DEEP] + [Source(**d) for d in TG_ABUSE] + [Source(**d) for d in X_SOURCES] + [Source(**d) for d in AUTO_SOURCES])
 
 BY_NAME = {s.name: s for s in SOURCES}
 CATEGORIES = sorted({s.category for s in SOURCES})
@@ -365,7 +366,18 @@ def pick(category=None, only=None, tags=None, respect_enabled=True):
         out = [s for s in out if s.category in cats]
     if only:
         names = {n.strip() for n in only.split(",") if n.strip()}
-        out = [s for s in out if s.name in names]
+
+        def _hit(src):
+            if src.name in names:
+                return True
+            for n in names:
+                if n.endswith("*") and src.name.startswith(n[:-1]):
+                    return True
+                if n.endswith("_") and src.name.startswith(n):
+                    return True
+            return False
+
+        out = [x for x in out if _hit(x)]
     if tags:
         wanted = {t.strip() for t in tags.split(",") if t.strip()}
         out = [s for s in out if wanted & set(s.tags)]

omni/sources_auto.py

@@ -0,0 +1,75 @@
+"""Авто-расширение: источники на публичных API без ключа.
+
+Файл генерируется `tools/auto_expand.py` из `data/source_candidates.json`
+(живые кандидаты) — правки руками перезапишутся, добавляй кандидатов в JSON.
+
+Обновлено: 2026-10-27 16:41 UTC · живых 46 из 52
+"""
+
+
+def A(name, kind, url, adapter, tags=("abuse",), limit=20, extra=None):
+    return {"name": name, "category": "abuse", "kind": kind, "url": url,
+            "adapter": adapter, "tags": tuple(tags), "limit": limit,
+            "detail": False, "extra": dict(extra or {})}
+
+
+AUTO = [
+    # ---------- открытые JSON API без ключа ----------
+    A("aa_devto_ai", "json", "https://dev.to/api/articles?per_page=25&top=7&tag=ai", "generic_json", ("abuse", "devto"), extra={"date_key": "published_at"}),
+    A("aa_lobsters", "json", "https://lobste.rs/hottest.json", "generic_json", ("abuse")),
+    A("aa_npm_llm", "json", "https://registry.npmjs.org/-/v1/search?text=llm%20api&size=25", "generic_json", ("abuse", "npm"), extra={"rows": "objects", "deref": "package", "title_key": "name", "text_key": "description", "base": "https://www.npmjs.com/package/", "slug_key": "name"}),
+    A("aa_npm_mcp", "json", "https://registry.npmjs.org/-/v1/search?text=mcp%20server&size=25", "generic_json", ("abuse", "npm"), extra={"rows": "objects", "deref": "package", "title_key": "name", "text_key": "description", "base": "https://www.npmjs.com/package/", "slug_key": "name"}),
+    A("aa_crates_llm", "json", "https://crates.io/api/v1/crates?q=llm&per_page=25", "generic_json", ("abuse", "crates"), extra={"rows": "crates", "base": "https://crates.io/crates/", "slug_key": "name"}),
+    A("aa_dockerhub_llm", "json", "https://hub.docker.com/v2/repositories/library/?page_size=25", "generic_json", ("abuse", "docker"), extra={"rows": "results", "base": "https://hub.docker.com/r/library/", "slug_key": "name"}),
+    A("aa_stack_openai", "json", "https://api.stackexchange.com/2.3/questions?order=desc&sort=activity&tagged=openai&site=stackoverflow&pagesize=25", "generic_json", ("abuse", "stack"), extra={"rows": "items", "url_key": "link", "date_key": "creation_date"}),
+    A("aa_stack_llm", "json", "https://api.stackexchange.com/2.3/questions?order=desc&sort=activity&tagged=llm&site=stackoverflow&pagesize=25", "generic_json", ("abuse", "stack"), extra={"rows": "items", "url_key": "link", "date_key": "creation_date"}),
+    A("aa_gitlab_new", "json", "https://gitlab.com/api/v4/projects?sort=desc&order_by=created_at&per_page=25", "generic_json", ("abuse", "gitlab"), extra={"url_key": "web_url"}),
+    A("aa_devpost", "json", "https://devpost.com/api/hackathons?status[]=open", "generic_json", ("abuse", "hackathon"), extra={"rows": "hackathons", "url_key": "url"}),
+    A("aa_hf_spaces_likes", "json", "https://huggingface.co/api/spaces?sort=likes&limit=25", "generic_json", ("abuse", "hf"), extra={"base": "https://huggingface.co/spaces/", "slug_key": "id", "title_key": "id"}),
+    A("aa_hf_spaces_trending", "json", "https://huggingface.co/api/spaces?sort=trendingScore&limit=25", "generic_json", ("abuse", "hf"), extra={"base": "https://huggingface.co/spaces/", "slug_key": "id", "title_key": "id"}),
+    A("aa_hf_models_trending", "json", "https://huggingface.co/api/models?sort=trendingScore&limit=25", "generic_json", ("abuse"), extra={"base": "https://huggingface.co/", "slug_key": "id", "title_key": "id"}),
+    A("aa_gitlab_llm", "json", "https://gitlab.com/api/v4/projects?search=llm&sort=desc&order_by=created_at&per_page=25", "generic_json", ("abuse"), extra={"url_key": "web_url"}),
+    A("aa_npm_free_api", "json", "https://registry.npmjs.org/-/v1/search?text=free%20api%20keys&size=25", "generic_json", ("abuse"), extra={"rows": "objects", "deref": "package", "title_key": "name", "text_key": "description", "base": "https://www.npmjs.com/package/", "slug_key": "name"}),
+    A("aa_crates_proxy", "json", "https://crates.io/api/v1/crates?q=openai%20proxy&per_page=20", "generic_json", ("abuse"), extra={"rows": "crates", "base": "https://crates.io/crates/", "slug_key": "name"}),
+
+    # ---------- Reddit через pullpush.io (без ключа) ----------
+    A("aa_pp_freeai", "json", "https://api.pullpush.io/reddit/search/submission/?subreddit=FreeAI&size=25", "generic_json", ("abuse"), extra={"rows": "data", "url_key": "url", "title_key": "title", "text_key": "selftext", "date_key": "created_utc"}),
+    A("aa_pp_openaiapi", "json", "https://api.pullpush.io/reddit/search/submission/?subreddit=OpenAIAPI&size=25", "generic_json", ("abuse"), extra={"rows": "data", "url_key": "url", "title_key": "title", "text_key": "selftext", "date_key": "created_utc"}),
+
+    # ---------- Bluesky / Mastodon ----------
+    A("aa_bsky_free_api", "json", "https://api.bsky.app/xrpc/app.bsky.feed.searchPosts?q=free%20api%20key&limit=25", "bsky_search", ("abuse", "bsky")),
+    A("aa_bsky_claude_api", "json", "https://api.bsky.app/xrpc/app.bsky.feed.searchPosts?q=claude%20api&limit=25", "bsky_search", ("abuse", "bsky")),
+    A("aa_bsky_jailbreak", "json", "https://api.bsky.app/xrpc/app.bsky.feed.searchPosts?q=jailbreak%20llm&limit=25", "bsky_search", ("abuse", "bsky", "jb")),
+    A("aa_masto_llm", "json", "https://mastodon.social/api/v1/timelines/tag/llm?limit=25", "mastodon_tag", ("abuse", "mastodon")),
+    A("aa_masto_jailbreak", "json", "https://mastodon.social/api/v1/timelines/tag/jailbreak?limit=25", "mastodon_tag", ("abuse", "mastodon", "jb")),
+    A("aa_masto_ai", "json", "https://mastodon.social/api/v1/timelines/tag/openai?limit=25", "mastodon_tag", ("abuse")),
+    A("aa_bsky_free_tokens", "json", "https://api.bsky.app/xrpc/app.bsky.feed.searchPosts?q=free%20tokens%20api&limit=25", "bsky_search", ("abuse")),
+    A("aa_bsky_openrouter", "json", "https://api.bsky.app/xrpc/app.bsky.feed.searchPosts?q=openrouter&limit=25", "bsky_search", ("abuse")),
+
+    # ---------- Discourse-сообщества ----------
+    A("aa_dis_openai", "json", "https://community.openai.com/latest.json", "discourse", ("abuse")),
+    A("aa_dis_hf", "json", "https://discuss.huggingface.co/latest.json", "discourse", ("abuse")),
+    A("aa_dis_cursor", "json", "https://forum.cursor.com/latest.json", "discourse", ("abuse")),
+    A("aa_dis_python", "json", "https://discuss.python.org/latest.json", "discourse", ("abuse")),
+    A("aa_dis_fly", "json", "https://community.fly.io/latest.json", "discourse", ("abuse")),
+    A("aa_dis_obsidian", "json", "https://forum.obsidian.md/latest.json", "discourse", ("abuse")),
+
+    # ---------- RSS/Atom ----------
+    A("aa_rss_habr", "rss", "https://habr.com/ru/rss/best/daily/?fl=ru", "rss", ("abuse")),
+    A("aa_rss_decoder", "rss", "https://the-decoder.com/feed/", "rss", ("abuse")),
+    A("aa_rss_marktechpost", "rss", "https://www.marktechpost.com/feed/", "rss", ("abuse")),
+    A("aa_rss_openai_news", "rss", "https://openai.com/news/rss.xml", "rss", ("abuse")),
+    A("aa_rss_simonw", "rss", "https://simonwillison.net/atom/everything/", "rss", ("abuse")),
+    A("aa_rss_hf_blog", "rss", "https://huggingface.co/blog/feed.xml", "rss", ("abuse")),
+    A("aa_rss_tc_ai", "rss", "https://techcrunch.com/category/artificial-intelligence/feed/", "rss", ("abuse")),
+    A("aa_rss_52pojie", "rss", "https://www.52pojie.cn/forum.php?mod=rss", "rss", ("abuse")),
+
+    # ---------- Telegram-превью ----------
+    A("aa_tg_ai_newz", "html", "https://t.me/s/ai_newz", "tg_web", ("abuse", "tg")),
+    A("aa_tg_data_secrets", "html", "https://t.me/s/data_secrets", "tg_web", ("abuse", "tg")),
+    A("aa_tg_denissexy", "html", "https://t.me/s/denissexy", "tg_web", ("abuse", "tg")),
+    A("aa_tg_neurohive", "json", "https://t.me/s/neurohive", "tg_web", ("abuse")),
+    A("aa_tg_gpt4telegram", "json", "https://t.me/s/gpt4telegram", "tg_web", ("abuse")),
+    A("aa_tg_dataleak", "json", "https://t.me/s/dataleak", "tg_web", ("abuse")),
+
+]

tools/auto_cycle.py

@@ -0,0 +1,90 @@
+"""Один цикл автоулучшения: проверить кандидатов → пересобрать sources_auto.py →
+докачать новое → пересобрать abuse-индекс → коммит/пуш. Запускать по крону (6ч).
+
+    python tools/auto_cycle.py            # полный цикл
+    python tools/auto_cycle.py --dry      # без коммита и без сети (только отчёт)
+
+Лог: data/auto_cycle.log (append).
+"""
+import argparse
+import os
+import subprocess
+import sys
+import time
+from datetime import datetime, timezone
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+PY = sys.executable
+LOG = ROOT / "data" / "auto_cycle.log"
+ENV = dict(os.environ, PYTHONPATH="", PYTHONIOENCODING="utf-8", PYTHONUTF8="1")
+
+
+def log(msg):
+    line = "%s %s" % (datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M"), msg)
+    print(line)
+    with LOG.open("a", encoding="utf-8") as fh:
+        fh.write(line + "\n")
+
+
+def run(args, timeout=1800):
+    t = time.time()
+    p = subprocess.run([PY] + args, cwd=str(ROOT), env=ENV, capture_output=True, text=True,
+                       encoding="utf-8", errors="replace", timeout=timeout)
+    out = (p.stdout or "") + (p.stderr or "")
+    return p.returncode, out.strip(), round(time.time() - t, 1)
+
+
+def git(*args, timeout=180):
+    p = subprocess.run(["git"] + list(args), cwd=str(ROOT), capture_output=True, text=True,
+                       encoding="utf-8", errors="replace", timeout=timeout)
+    return p.returncode, ((p.stdout or "") + (p.stderr or "")).strip()
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--dry", action="store_true")
+    ap.add_argument("--pause", type=float, default=6.0)
+    a = ap.parse_args()
+    log("=== cycle start ===")
+
+    if not a.dry:
+        code, out, sec = run(["tools/auto_expand.py", "--probe", "--write", "--pause", str(a.pause)])
+        tail = [l for l in out.splitlines() if "LIVE" in l or "wrote" in l or "skip (" in l]
+        log("probe/write rc=%s %ss | %s" % (code, sec, " / ".join(tail[-3:]) or out[-200:]))
+
+    changed = git("status", "--porcelain", "--", "omni/sources_auto.py")[1]
+    live = 0
+    try:
+        import re
+        txt = (ROOT / "omni" / "sources_auto.py").read_text(encoding="utf-8")
+        live = len(re.findall(r"^    A\(", txt, re.M))
+    except Exception:
+        pass
+    log("sources_auto: live=%d changed=%s" % (live, bool(changed)))
+
+    if changed and not a.dry:
+        code, out, sec = run(["-m", "omni", "feed", "--only", "aa_", "--limit", "15"])
+        log("feed rc=%s %ss | %s" % (code, sec, out.splitlines()[-1] if out else ""))
+
+    if not a.dry:
+        code, out, sec = run(["-c", "import sys;sys.path.insert(0,'.');"
+                              "from omni import abuse;print('rebuilt', abuse.rebuild())"], timeout=2400)
+        log("abuse rc=%s %ss | %s" % (code, sec, (out.splitlines() or [""])[-1]))
+
+    if changed and not a.dry:
+        git("add", "--", "omni/sources_auto.py")
+        code, out = git("commit", "-m", "auto: источники %d live (%s)" %
+                        (live, datetime.now(timezone.utc).strftime("%d.%m %H:%M UTC")))
+        log("commit rc=%s | %s" % (code, out.splitlines()[0] if out else ""))
+        for remote in ("gitea", "github"):
+            if not git("remote")[1].find(remote) >= 0:
+                continue
+            rc, o = git("push", remote, "main")
+            log("push %s rc=%s | %s" % (remote, rc, o.splitlines()[-1] if o else ""))
+    log("=== cycle end ===")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

tools/auto_expand.py

@@ -0,0 +1,218 @@
+"""Авто-расширение источников: тест кандидатов → генерация omni/sources_auto.py.
+
+    python tools/auto_expand.py --probe         # только проверка, без записи
+    python tools/auto_expand.py --write         # живые кандидаты → sources_auto.py
+    python tools/auto_expand.py --discover      # добрать кандидатов из public-apis
+    python tools/auto_expand.py --add NAME URL --adapter rss
+
+Кандидаты: data/source_candidates.json (list of {name,url,adapter,kind,extra,tags,limit}).
+Здоровье: data/source_health.json — сколько раз источник отдал items и когда.
+Мёртвым считается кандидат с 0 items в 3 прогонах подряд → исключается из генерации.
+"""
+import argparse
+import json
+import sys
+import time
+import urllib.request
+from datetime import datetime, timezone
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+from omni import feeds  # noqa: E402
+from omni.sources import Source  # noqa: E402
+
+CAND = ROOT / "data" / "source_candidates.json"
+HEALTH = ROOT / "data" / "source_health.json"
+OUT = ROOT / "omni" / "sources_auto.py"
+PUBLIC_APIS = "https://raw.githubusercontent.com/public-apis/public-apis/master/README.md"
+
+HEADER = '''"""Авто-расширение: источники на публичных API без ключа.
+
+Файл генерируется `tools/auto_expand.py` из `data/source_candidates.json`
+(живые кандидаты) — правки руками перезапишутся, добавляй кандидатов в JSON.
+
+Обновлено: {ts} · живых {live} из {total}
+"""
+
+
+def A(name, kind, url, adapter, tags=("abuse",), limit=20, extra=None):
+    return {{"name": name, "category": "abuse", "kind": kind, "url": url,
+            "adapter": adapter, "tags": tuple(tags), "limit": limit,
+            "detail": False, "extra": dict(extra or {{}})}}
+
+
+AUTO = [
+{body}]
+'''
+
+GROUP_RU = {
+    "json": "# ---------- открытые JSON API без ключа ----------",
+    "reddit": "# ---------- Reddit через pullpush.io (без ключа) ----------",
+    "social": "# ---------- Bluesky / Mastodon ----------",
+    "discourse": "# ---------- Discourse-сообщества ----------",
+    "rss": "# ---------- RSS/Atom ----------",
+    "tg": "# ---------- Telegram-превью ----------",
+}
+
+
+def load(path, default):
+    try:
+        return json.loads(path.read_text(encoding="utf-8"))
+    except Exception:
+        return default
+
+
+def save(path, data):
+    path.write_text(json.dumps(data, ensure_ascii=False, indent=1), encoding="utf-8")
+
+
+def to_source(c):
+    return Source(name=c["name"], category=c.get("category", "abuse"), kind=c.get("kind", "json"),
+                  url=c["url"], adapter=c["adapter"], tags=tuple(c.get("tags") or ("abuse",)),
+                  limit=int(c.get("limit", 20)), detail=False, extra=dict(c.get("extra") or {}))
+
+
+def probe(cands, pause=2.0):
+    health = load(HEALTH, {})
+    res = []
+    last_host = None
+    for c in cands:
+        host = c["url"].split("/")[2] if "//" in c["url"] else ""
+        if host == last_host and pause:
+            time.sleep(pause)
+        last_host = host
+        try:
+            items = feeds.fetch(to_source(c), limit=min(20, int(c.get("limit", 20))))
+            n = len(items)
+            sample = [i.title[:70] for i in items[:2]]
+            err = ""
+        except Exception as exc:
+            n, sample, err = 0, [], "%s: %s" % (type(exc).__name__, str(exc)[:90])
+        h = health.setdefault(c["name"], {"ok": 0, "fail": 0, "items": 0})
+        h["items"] = n
+        h["ok"] = h.get("ok", 0) + 1 if n else 0
+        h["fail"] = 0 if n else h.get("fail", 0) + 1
+        h["last"] = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M")
+        h["url"] = c["url"]
+        if err:
+            h["error"] = err
+        else:
+            h.pop("error", None)
+        res.append((c, n, sample, err))
+        print("%-26s %-4s %3d  %s" % (c["name"], "LIVE" if n else "DEAD", n, err or " | ".join(sample)))
+    save(HEALTH, health)
+    return res
+
+
+def write_module(cands, results):
+    health = load(HEALTH, {})
+    order = ["json", "reddit", "social", "discourse", "rss", "tg"]
+    buckets = {k: [] for k in order}
+    for c, n, _, _ in results:
+        if not n:
+            continue
+        if (health.get(c["name"]) or {}).get("fail", 0) >= 3:
+            print("skip (3 fails подряд):", c["name"])
+            continue
+        g = c.get("group") or ("tg" if c["adapter"] == "tg_web" else
+                               "discourse" if c["adapter"] == "discourse" else
+                               "reddit" if "pullpush" in c["url"] or "reddit" in c["url"] else
+                               "social" if c["adapter"] in ("bsky_search", "mastodon_tag") else
+                               "rss" if c["adapter"] == "rss" else "json")
+        buckets.setdefault(g, []).append(c)
+    body = []
+    live = 0
+    for g in order + [k for k in buckets if k not in order]:
+        rows = buckets.get(g) or []
+        if not rows:
+            continue
+        body.append("    " + GROUP_RU.get(g, "# ---------- прочее ----------"))
+        for c in rows:
+            live += 1
+            extra = ""
+            if c.get("extra"):
+                extra = ", extra=%s" % json.dumps(c["extra"], ensure_ascii=False)
+            tags = ", ".join('"%s"' % t for t in (c.get("tags") or ("abuse",)))
+            body.append('    A("%s", "%s", "%s", "%s", (%s)%s),' % (
+                c["name"], c.get("kind", "json"), c["url"], c["adapter"], tags, extra))
+        body.append("")
+    ts = datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M UTC")
+    OUT.write_text(HEADER.format(ts=ts, live=live, total=len(cands), body="\n".join(body) + "\n"),
+                   encoding="utf-8")
+    print("\nwrote %s: %d live of %d" % (OUT.name, live, len(cands)))
+
+
+def discover(limit=40):
+    """Добрать бесплатные (Auth=No) API из public-apis."""
+    req = urllib.request.Request(PUBLIC_APIS, headers={"User-Agent": "omni-parse/1.0"})
+    md = urllib.request.urlopen(req, timeout=40).read().decode("utf-8", "replace")
+    rows = []
+    for line in md.splitlines():
+        if not line.startswith("|") or "http" not in line:
+            continue
+        cells = [c.strip() for c in line.strip("|").split("|")]
+        if len(cells) < 6:
+            continue
+        url, auth, https = cells[2], cells[4], cells[5]
+        if "No" not in auth or "http" not in url:
+            continue
+        if any(b in url for b in ("github.com/public-apis", "localhost")):
+            continue
+        rows.append({"name": "dis_" + url.split("//")[-1].split("/")[0].replace("www.", "").replace(".", "_")[:24],
+                     "url": url, "adapter": "generic_json", "kind": "json",
+                     "tags": ["abuse", "public_api"], "group": "json", "limit": 15})
+    uniq = {}
+    for r in rows:
+        uniq.setdefault(r["name"], r)
+    cands = load(CAND, [])
+    have = {c["name"] for c in cands} | {c["url"] for c in cands}
+    new = [r for r in uniq.values() if r["name"] not in have and r["url"] not in have][:limit]
+    cands.extend(new)
+    save(CAND, cands)
+    print("discover: +%d кандидатов (всего %d)" % (len(new), len(cands)))
+    return new
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--probe", action="store_true")
+    ap.add_argument("--write", action="store_true")
+    ap.add_argument("--write-only", action="store_true", help="сгенерировать из data/source_health.json без сети")
+    ap.add_argument("--discover", action="store_true")
+    ap.add_argument("--add", nargs=2, metavar=("NAME", "URL"))
+    ap.add_argument("--adapter", default="generic_json")
+    ap.add_argument("--pause", type=float, default=2.0)
+    a = ap.parse_args()
+
+    if a.add:
+        cands = load(CAND, [])
+        cands.append({"name": a.add[0], "url": a.add[1], "adapter": a.adapter,
+                      "kind": "json", "tags": ["abuse"], "limit": 20})
+        save(CAND, cands)
+        print("added", a.add[0])
+        return 0
+    if a.discover:
+        discover()
+        return 0
+
+    cands = load(CAND, [])
+    if a.write_only:
+        health = load(HEALTH, {})
+        results = [(c, (health.get(c["name"]) or {}).get("items", 0), [], "") for c in cands]
+        write_module(cands, results)
+        return 0
+    if not cands:
+        print("нет кандидатов в", CAND)
+        return 1
+    results = probe(cands, pause=a.pause)
+    live = sum(1 for _, n, _, _ in results if n)
+    print("\nLIVE %d / %d, items %d" % (live, len(results), sum(n for _, n, _, _ in results)))
+    if a.write:
+        write_module(cands, results)
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

tools/check_yield.py

@@ -0,0 +1,31 @@
+"""Проверка yield каждого источника: имя | items | первых 2 заголовка."""
+import sys
+import traceback
+from concurrent.futures import ThreadPoolExecutor
+
+sys.path.insert(0, ".")
+
+from omni import feeds, sources  # noqa: E402
+
+prefix = sys.argv[1] if len(sys.argv) > 1 else "aa_"
+picked = [s for s in sources.SOURCES if s.name.startswith(prefix)]
+print("sources:", len(picked))
+
+
+def run(src):
+    try:
+        items = feeds.fetch(src, limit=min(20, src.limit or 20))
+        return src.name, len(items), [i.title[:60] for i in items[:2]], ""
+    except Exception as exc:
+        return src.name, 0, [], "%s: %s" % (type(exc).__name__, str(exc)[:70])
+
+
+with ThreadPoolExecutor(max_workers=10) as ex:
+    rows = list(ex.map(run, picked))
+
+live = 0
+for name, n, titles, err in sorted(rows, key=lambda r: -r[1]):
+    flag = "LIVE" if n else "DEAD"
+    live += bool(n)
+    print("%-26s %-4s %3d  %s" % (name, flag, n, err or " | ".join(titles)))
+print("LIVE %d / %d, items %d" % (live, len(rows), sum(r[1] for r in rows)))

tools/probe_candidates.py

@@ -0,0 +1,89 @@
+"""Живость кандидатов: статус, размер, число элементов в JSON/ленте."""
+import json
+import sys
+import urllib.request
+from concurrent.futures import ThreadPoolExecutor
+
+UA = ("Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+      "(KHTML, like Gecko) Chrome/131.0 Safari/537.36")
+
+CAND = [
+    ("devto", "https://dev.to/api/articles?per_page=25&top=7"),
+    ("lobsters", "https://lobste.rs/hottest.json"),
+    ("npm", "https://registry.npmjs.org/-/v1/search?text=llm&size=25"),
+    ("crates", "https://crates.io/api/v1/crates?q=llm&per_page=25"),
+    ("dockerhub", "https://hub.docker.com/v2/repositories/library/?page_size=25"),
+    ("stackoverflow", "https://api.stackexchange.com/2.3/questions?order=desc&sort=activity&tagged=openai&site=stackoverflow&pagesize=25"),
+    ("gitlab", "https://gitlab.com/api/v4/projects?sort=desc&order_by=created_at&per_page=25"),
+    ("devpost", "https://devpost.com/api/hackathons?status[]=open"),
+    ("hf_spaces", "https://huggingface.co/api/spaces?sort=likes&limit=25"),
+    ("mastodon", "https://mastodon.social/api/v1/timelines/tag/llm?limit=25"),
+    ("bsky", "https://public.api.bsky.app/xrpc/app.bsky.feed.searchPosts?q=free%20api&limit=25"),
+    ("reddit_localllama", "https://www.reddit.com/r/LocalLLaMA/.json?limit=25"),
+    ("reddit_chatgptproxies", "https://www.reddit.com/r/ChatGPTProxies/.json?limit=25"),
+    ("reddit_jailbreak", "https://www.reddit.com/r/jailbreak/.json?limit=25"),
+    ("reddit_cybersecurity", "https://www.reddit.com/r/cybersecurity/.json?limit=25"),
+    ("reddit_selfhosted", "https://www.reddit.com/r/selfhosted/.json?limit=25"),
+    ("reddit_piracy", "https://www.reddit.com/r/Piracy/.json?limit=25"),
+    ("dis_openai", "https://community.openai.com/latest.json"),
+    ("dis_hf", "https://discuss.huggingface.co/latest.json"),
+    ("dis_cursor", "https://forum.cursor.com/latest.json"),
+    ("dis_python", "https://discuss.python.org/latest.json"),
+    ("dis_fly", "https://community.fly.io/latest.json"),
+    ("dis_obsidian", "https://forum.obsidian.md/latest.json"),
+    ("rss_habr", "https://habr.com/ru/rss/best/daily/?fl=ru"),
+    ("rss_decoder", "https://the-decoder.com/feed/"),
+    ("rss_marktechpost", "https://www.marktechpost.com/feed/"),
+    ("rss_openai", "https://openai.com/news/rss.xml"),
+    ("rss_simonw", "https://simonwillison.net/atom/everything/"),
+    ("rss_hf", "https://huggingface.co/blog/feed.xml"),
+    ("rss_tc_ai", "https://techcrunch.com/category/artificial-intelligence/feed/"),
+    ("tg_ai_newz", "https://t.me/s/ai_newz"),
+    ("tg_seeallo", "https://t.me/s/seeallo"),
+    ("tg_data_secrets", "https://t.me/s/data_secrets"),
+    ("tg_gptblog", "https://t.me/s/gptblog"),
+    ("tg_denissexy", "https://t.me/s/denissexy"),
+    ("gh_trending", "https://github.com/trending?since=daily"),
+]
+
+
+def probe(item):
+    name, url = item
+    req = urllib.request.Request(url, headers={"User-Agent": UA,
+                                               "Accept": "application/json, text/html, application/xml;q=0.9"})
+    try:
+        with urllib.request.urlopen(req, timeout=20) as r:
+            raw = r.read(400000)
+            code = r.status
+    except Exception as exc:
+        return name, str(exc)[:60], 0, 0
+    n = 0
+    try:
+        d = json.loads(raw.decode("utf-8", "replace"))
+        if isinstance(d, list):
+            n = len(d)
+        elif isinstance(d, dict):
+            for k in ("items", "posts", "results", "objects", "crates", "hackathons", "data", "articles", "children"):
+                v = d.get(k)
+                if isinstance(v, list):
+                    n = len(v)
+                    break
+            if not n:
+                n = len([v for v in d.values() if isinstance(v, list)])
+    except Exception:
+        try:
+            text = raw.decode("utf-8", "replace")
+            n = text.count("<item") or text.count("<entry") or text.count("message-text")
+            if not n and "t.me" in url:
+                n = text.count("tgme_widget_message_text")
+        except Exception:
+            n = 0
+    return name, "ok", code, n
+
+
+with ThreadPoolExecutor(max_workers=12) as ex:
+    res = list(ex.map(probe, CAND))
+
+for name, status, code, n in res:
+    print("%-24s %-6s %s n=%s" % (name, code, status, n))
+print("LIVE:", sum(1 for _, s, _, n in res if s == "ok" and n), "of", len(res))