liart.py: OPAC-Global protocol probed (direct.exe FindView/FindSize, guest session); SearXNG record-URL fallback works; direct mode blocked on search labels+iddb (admin-only InfoDB) — data/liart-params.json stub for browser read

AGENTS.md

@@ -110,9 +110,10 @@ sections/0N-<name>/  # one .md per BOOK (chapter-level list items are merged int
   `| RU title | EN original | publisher, year | ISBN |` → `## Homonym warnings` (ЕДИНОЕ название,
   legacy-варианты: «Homonym warning (date)», «Homonyms / notes», «Омонимы») → `## Notes` (датированные
   append; legacy «Round N» логи сюда). Запрещено: `**EN identity:**` отдельно, `**Dossier status:**`.
-- **TODO (later, after 4 sections in order):** `tools/liart.py` — search for opac.liart.ru (РГБИ):
+- **TODO (later, after 4 sections in order):** `tools/liart.py` — **HALF DONE 2026-09-21:**
-  platform «Библика» (api.bibliosearch.ru/bs.min.js), search goes via JS `simpleSearch()`,
+  OPAC-Global protocol reverse-engineered (direct.exe/FindView, guest session), SearXNG
-  endpoint not yet probed; until then — SearXNG `site:opac.liart.ru <author>` only.
+  record-URL fallback works; BLOCKED on search labels + iddb (server InfoDB, admin-only for
+  GUEST) — needs a one-time browser read (see data/liart-params.json).
 - **Author dossiers** (`authors/<lastname>-<firstname>.md`) are the **CANONICAL** home for
   identity, shared across ALL sections (same author can appear in 01, 04, …). One dossier per
   PERSON, slug `<lastname>-<firstname>` (lowercase, hyphens). Hold: dates, RU name candidates
@@ -385,8 +386,19 @@ sections/0N-<name>/  # one .md per BOOK (chapter-level list items are merged int
 
 ### 26. book.ivran.ru + opac.liart.ru — small RU library OPACs, curl-able GOST records (2026-09-20)
 - `book.ivran.ru/book?id=<N>` — «Книги сотрудников ИВ РАН» DB: full GOST rows (title, editor,
-  city/year/pages, ISBN). `opac.liart.ru/record/<db>/<hex>` — РГБИ (State Art Library) OPAC.
+  city/year/pages, ISBN). Find via SearXNG `site:book.ivran.ru`.
-  Find via SearXNG `site:book.ivran.ru` / `site:opac.liart.ru <author>`.
+- **opac.liart.ru (РГБИ) — platform switched to OPAC-Global (DITM-Global)** (2026-09-21, probed):
+  - `/opacg/` guest session: POST `arg0=GUEST&arg1=GUESTE&TypeAccess=PayAccess` to
+    `/cgiopac/opacg/opac.exe`; session `9096453`/`GUEST` (hardcoded in JS).
+  - Search = POST `/cgiopac/opacg/direct.exe` FLAT fields: `_service=opacfindd.FindView`
+    (or `FindSize`), `query/body="<label> <value>"`, `iddb=<N>`, `userId`, `session`,
+    `_xsl=_version=2.7.0/1.1.0` — full protocol in `tools/liart.py` docstring.
+  - **BLOCKER:** valid search labels + iddb come from server InfoDB (admin-only for GUEST;
+    iddb 1–4 absent, 7+ exist). Record pages `/record/<db>/<hex>` ARE curl-able (GOST in <title>).
+    SearXNG `site:opac.liart.ru` index is sparse (author queries 0, `site:.../record` 20).
+  - **Tool: `tools/liart.py`** — default mode = SearXNG record-URL fallback (works today);
+    `--direct` mode ready, needs `data/liart-params.json` filled from the browser
+    (Поиск → Базовый → F12: `input id=basic_<label>` / `select id=field_1`). TODO remains.
 
 ### 27. maap.pro «Библиотека» — RU article translations of Jungian authors (2026-09-20, user-provided)
 - `https://www.maap.pro/biblioteka/stati/` — МОСКОВСКАЯ АССОЦИАЦИЯ АНАЛИТИЧЕСКОЙ ПСИХОЛОГИИ

data/liart-params.json

@@ -0,0 +1,6 @@
+{
+  "_comment": "FILL FROM BROWSER: open https://opac.liart.ru/opacg/ then Поиск then Базовый; F12 Elements: input id=basic_<label> and select id=field_1 options. iddb: try 7..20 (1-4 not for GUEST, 7+ exist).",
+  "iddb": null,
+  "labels": {"author": "TBD", "title": "TBD", "free": "TBD"},
+  "probe_labels": []
+}

tools/liart.py

@@ -0,0 +1,164 @@
+#!/usr/bin/env python3
+"""РГБИ (State Art Library of Russia, opac.liart.ru) search.
+
+Platform: OPAC-Global (DITM-Global), switched from «Библика» — probed 2026-09-21.
+
+Protocol (reverse-engineered from /opacg/js):
+  1. GET  /opacg/                                  → frameset (login as guest auto)
+  2. POST /cgiopac/opacg/opac.exe  arg0=GUEST&arg1=GUESTE&TypeAccess=PayAccess
+     → guest session (numsean=9096453 hardcoded in JS, identif=GUEST)
+  3. POST /cgiopac/opacg/direct.exe  (FLAT form fields, not querylist):
+        _service   = opacfindd.FindView     (results) | opacfindd.FindSize (count)
+        _version   = 2.7.0 (FindView) | 1.1.0 (FindSize)
+        _xsl       = /opacg/html/search/xsl/search_results.xsl | .../history_find_size.xsl
+        _errorXsl  = /opacg/html/common/xsl/error.xsl
+        _wait      = 6M
+        query/body = "<label> <value>"      (FindView) | query=... (FindSize)
+        length=10  start=0
+        userId=GUEST  session=9096453
+        iddb=<logical DB number>            (1..4 not found for GUEST; 7+ exist)
+        level[i]  = SearchLevels (server-provided, unknown)
+  Query syntax (search_term_handlers.js): label + ' ' + value, e.g. "AUT Нойманн".
+  KNOWN BLOCKER: valid search LABELS come from server InfoDB (code XML), which is
+  admin-only for GUEST (opacsettingd.* on opac.exe → error page). Labels are rendered
+  into the search form by JS in a real browser. To configure: open
+  https://opac.liart.ru/opacg/ → «Поиск» → «Базовый» and fill data/liart-params.json:
+      {"iddb": N, "labels": {"author": "...", "title": "...", "free": "..."}}
+  (values = the exact label tokens; in the browser: F12 → Elements → the
+   <input id="basic_<label>"> / <select id="field_1"> <option value="...">).
+
+Modes:
+  liart.py --direct "Нойманн" [FIELD]   direct.exe FindView (needs data/liart-params.json)
+  liart.py "Нойманн"                    SearXNG site:opac.liart.ru fallback (works today)
+  liart.py --label-probe "Нойманн"      brute a label list against FindSize (debug)
+
+Polite: ~3s between direct.exe calls. Output: parsed records (title/author/year) or raw XML.
+"""
+import json, os, re, subprocess, sys, time, html
+
+BASE = 'https://opac.liart.ru'
+CFG = os.path.join(os.path.dirname(__file__), '..', 'data', 'liart-params.json')
+UA = 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0 Safari/537.36'
+
+def curl(url, data=None, cookie=None):
+    cmd = ['curl', '-s', '--max-time', '45', '-A', UA]
+    if cookie:
+        cmd += ['-b', cookie, '-c', cookie]
+    if data:
+        cmd += ['-X', 'POST', url]
+        for k, v in data.items():
+            cmd += ['--data-urlencode', f'{k}={v}']
+    else:
+        cmd.append(url)
+    r = subprocess.run(cmd, capture_output=True, text=False)
+    return r.stdout.decode('utf-8', 'replace')
+
+def session(cook='/tmp/liart_cookies.txt'):
+    open(cook, 'w').close()
+    curl(BASE + '/opacg/', cookie=cook)
+    r = curl(BASE + '/cgiopac/opacg/opac.exe',
+             {'arg0': 'GUEST', 'arg1': 'GUESTE', 'TypeAccess': 'PayAccess'}, cookie=cook)
+    return cook
+
+def load_cfg():
+    if not os.path.exists(CFG):
+        sys.exit(f'no {CFG} — fill it from the browser form (see module docstring)')
+    return json.load(open(CFG))
+
+def findview(cook, iddb, query, length=20, start=0, raw=False):
+    r = curl(BASE + '/cgiopac/opacg/direct.exe', {
+        '_xsl': '/opacg/html/search/xsl/search_results.xsl',
+        '_errorXsl': '/opacg/html/common/xsl/error.xsl',
+        '_version': '2.7.0',
+        '_service': 'opacfindd.FindView',
+        'length': str(length), 'start': str(start),
+        'query/body': query,
+        'userId': 'GUEST', 'session': '9096453',
+        'iddb': str(iddb),
+    }, cookie=cook)
+    if raw:
+        return r
+    # parse: results are <document>... records with MARC-ish fields; try common shapes
+    texts = re.findall(r'<text[^>]*>(.*?)</text>', r, re.S)
+    titles = re.findall(r'<title[^>]*>(.*?)</title>', r, re.S)
+    if '<error' in r:
+        err = ' | '.join(re.findall(r'<entry>([^<]*)</entry>', r)[:3])
+        return {'error': err.strip()}
+    recs = re.findall(r'<record.*?>(.*?)</record>', r, re.S)
+    out = []
+    for rec in recs:
+        row = {}
+        for m in re.finditer(r'<field[^>]*name="([^"]*)"[^>]*>(.*?)</field>', rec, re.S):
+            name, val = m.group(1), re.sub(r'<[^>]+>', ' ', m.group(2)).strip()
+            row.setdefault(name, val)
+        out.append(row)
+    return {'count': len(out), 'records': out, 'titles': [html.unescape(t) for t in titles[:30]]}
+
+def findsize(cook, iddb, query):
+    r = curl(BASE + '/cgiopac/opacg/direct.exe', {
+        '_xsl': '/opacg/html/search/xsl/history_find_size.xsl',
+        '_version': '1.1.0',
+        '_service': 'opacfindd.FindSize',
+        'query': query,
+        'userId': 'GUEST', 'session': '9096453',
+        'iddb': str(iddb),
+    }, cookie=cook)
+    m = re.search(r'<total[^>]*>(\d+)</total>|<findsize[^>]*>\s*(\d+)', r)
+    if m:
+        return int(m.group(1) or m.group(2))
+    err = ' | '.join(re.findall(r'<entry>([^<]*)</entry>', r)[:2])
+    return {'error': err.strip()} if err else 0
+
+def searxng(q):
+    import urllib.parse
+    url = 'http://localhost:8888/search?' + urllib.parse.urlencode(
+        {'q': q, 'format': 'json', 'language': 'ru'})
+    r = subprocess.run(['curl', '-s', '--max-time', '30', url], capture_output=True, text=True)
+    return json.loads(r.stdout or '{}').get('results', [])
+
+def searxng_fallback(q):
+    # record pages are curl-able: /record/<db>/<hex> with GOST line in <title>.
+    # Google index of the site is sparse — try variants.
+    seen = set()
+    for query in (f'site:opac.liart.ru/record {q}', f'site:opac.liart.ru {q}'):
+        for r in searxng(query):
+            u = r['url']
+            if u in seen:
+                continue
+            seen.add(u)
+            print(f"{r['title']}")
+            print(f"   {u}")
+            c = (r.get('content') or '').replace('\n', ' ')
+            if c:
+                print(f'   {c[:220]}')
+            print()
+    if not seen:
+        print('(0 results — site: index is sparse; try the OPAC in browser, or another catalog)')
+
+def main():
+    args = sys.argv[1:]
+    if not args:
+        sys.exit(__doc__)
+    if args[0] == '--direct':
+        cfg = load_cfg()
+        cook = session()
+        q = args[1]
+        field = args[2] if len(args) > 2 else 'free'
+        label = cfg['labels'][field]
+        res = findview(cook, cfg['iddb'], f'{label} {q}', raw='--raw' in args)
+        if isinstance(res, dict) and 'error' in res:
+            print('ERROR:', res['error'])
+        else:
+            print(json.dumps(res, ensure_ascii=False, indent=1)[:6000])
+    elif args[0] == '--label-probe':
+        cfg = load_cfg()
+        cook = session()
+        for lab in cfg.get('probe_labels', []):
+            n = findsize(cook, cfg['iddb'], f'{lab} {args[1]}')
+            print(f'{lab:20} -> {n}')
+            time.sleep(3)
+    else:
+        searxng_fallback(' '.join(args))
+
+if __name__ == '__main__':
+    main()