#!/usr/bin/env python3 import json, re, os BASE="/home/dev/.claude/projects/-opt-projects--dashboard-cinik-worktrees-tom-auto-populate/5c79ced8-8c91-47a0-a9c5-54e5c5f0ee15/tool-results" FILES_WRAPPED={ 'dr-cinik': f"{BASE}/mcp-claude_ai_Ahref-site-explorer-all-backlinks-1780853366112.txt", 'dr-serkan-aygin': f"{BASE}/mcp-claude_ai_Ahref-site-explorer-all-backlinks-1780853366079.txt", 'cosmedica': f"{BASE}/mcp-claude_ai_Ahref-site-explorer-all-backlinks-1780853369017.txt", 'vera-clinic': f"{BASE}/mcp-claude_ai_Ahref-site-explorer-all-backlinks-1780853371962.txt", } FILES_PLAIN={ 'clinicana': "/tmp/cinik-pull/clinicana-raw.json", 'smile-hair-clinic': "/tmp/cinik-pull/smile-raw.json", } def load_wrapped(p): raw=json.load(open(p)); return json.loads(raw[0]['text'])['backlinks'] def load_plain(p): return json.load(open(p))['backlinks'] # ---- Editorial filter ---- SPAM_TLD_TOKENS=re.compile(r'(seo|link|backlink|rank|authority|dofollow)', re.I) REVIEW_TPL=re.compile(r'there was a time last year', re.I) # Known junk domains (directories, similarity, redirect/login/ugc wrappers) JUNK_DOMAINS={ 'sitelike.org','siteprice.org','technofizi.net','publicwww.com','sitewideanalytics.com', 'freelistingusa.com','askmap.net','fyple.co.uk','place123.net','todosbiz.es','sektor.gen.tr', 'yandex.az','yandex.com','yandex.ru','yango.com','safariclub.org','zdn.vn','mints.ne.jp', 'boardhost.com','pastelink.net','diigo.com','brunch.co.kr','clan.su','luma.com', 'thesocietypages.org','lonestar.edu','usac.edu.gt','plesk.page','co.com','opovo.com.br', 'neuthema.de','wbez.org','leasefintech.com','moneyfx.boardhost.com','finishedtask.com', 'thefrisky.com','hairlinetransplantation.com', # similarity / aggregator / freehost / paste / bookmark / generic junk seen in wrapped files 'web.app','page.tl','figma.site','ghost.io','figma.com','glarity.app','parse.gl', 'video-bookmark.com','joinrunway.io','vuink.com','2ch.org','d-h.st','xpr.media', 'yandex.md','yandex.uz','yandex.com.tr','moviezone.cz','stadt1.de','abc-directory.com', 'insightmag.com','epubzone.org','dowjones.io','minyanville.com','sourcefed.com', 'flokii.com','flokib.at','bizlistusa.com','bizratings.com','yplocal.com','citysquares.com', 'connectively.us','synara.ar','am-news.com','marketminute.com', } # Syndicated-PR mirror network: td==0 (or tiny) news-mirror domains republishing one wire headline. # These are PBN-like syndication mirrors, not genuine editorial. PR_MIRROR_TITLE=re.compile( r'(Vera Clinic and Appsilon|Bio-Authenticity Premium|Becomes Official Global Partner|' r'Official Partner of the Houston Rockets|Unveil Vector 10|Fixed-Price Hair Transplant Packages|' r'Top 5 Hair Transplant Clinics in Turkey: Leading Providers|' r'Cliniques de Greffe de Cheveux en Turquie . Guide Sp)', re.I) # Mass guest-post farm: traffic_domain 0, DR ~68-76, generic invented brand + spun title GUESTPOST_TITLE=re.compile(r'(hair transplant turkey|hair transplant istanbul|best hair transplant|hair transplant turkey)', re.I) def domain_is_pbn_shop(dom): if dom.endswith('.shop') or dom.endswith('.store'): return True return False def is_excluded(b): dom=b.get('root_name_source','') or '' title=b.get('title','') or '' td=b.get('traffic_domain',0) or 0 dr=b.get('domain_rating_source',0) or 0 if b.get('is_ugc'): return True,'ugc' if domain_is_pbn_shop(dom): return True,'pbn-shop' if REVIEW_TPL.search(title): return True,'review-tpl' if dom in JUNK_DOMAINS: return True,'junk-domain' # syndication "User |" profile/author junk pages if title.startswith('User |') or title.startswith('User ') and '|' in title[:8]: return True,'user-profile-junk' # syndicated-PR mirror network: recognizable wire headline republished across mirror domains. # Keep only genuine PR-wire originators (einnews, ipsnews, newsblaze, issuewire); drop the rest. PR_ORIGINATORS={'einnews.com','ipsnews.net','newsblaze.com','issuewire.com', 'globenewswire.com','24-7pressrelease.com','natlawreview.com', 'presseportal.de','presseportal.ch','ots.at'} if PR_MIRROR_TITLE.search(title) and dom not in PR_ORIGINATORS: return True,'pr-mirror' # Off-topic incidental links on real outlets (manually reviewed): page not about the brand, # hair transplant, or aesthetic medicine -> not brand editorial. OFFTOPIC_DOMAINS={'tastingtable.com','ladbible.com','ctinsider.com','gathered.how', 'yardbarker.com','pastemagazine.com','patient.info','audioboom.com', 'factually.co','morecore.de','nona.my','lifecoachcode.com','lifepositive.com', 'sport.fr','darelhilal.com','2ch.org','moviezone.cz'} if dom in OFFTOPIC_DOMAINS: return True,'offtopic-incidental' # generic news-mirror clone domains (newswatch/newsonline/gazette/worldreport/times/daily etc.) # republishing the same story with traffic_domain 0 -> syndication PBN if td==0 and re.search(r'(newswatch|newsonline|newsnetwork|nouvelles|gazette|worldreport|' r'industrytoday|bizsense|dailynews|daily-news|times-online|phenomena|' r'pluralist|mmminimal|sausalito|fictiontalk|thedishh|small-biz|' r'newscenter|travelindustry)', dom, re.I): return True,'news-mirror' # edu.pl / .id / .nl / .org clones with spun titles & traffic_domain 0 if td==0 and GUESTPOST_TITLE.search(title): return True,'guestpost-farm' # generic invented-brand .com with traffic_domain 0 and spun hair-transplant title if td==0 and re.search(r'\b(2026|2027)\b', title) and GUESTPOST_TITLE.search(title): return True,'guestpost-farm' return False,None # ---- Classify ---- PR_DOMAINS={'globenewswire.com','presseportal.de','presseportal.ch','ots.at','natlawreview.com', 'einnews.com','abnewswire.com','prnewswire.com','businesswire.com','ipsnews.net', 'newsblaze.com','24-7pressrelease.com','issuewire.com','financialcontent.com', 'marketminute.com','ftnnews.com'} REAL_MEDIA={'ouest-france.fr','ledauphine.com','scotsman.com','yorkshirepost.co.uk', 'derwesten.de','ga.de','volksfreund.de','oe24.at','gulfnews.com','targatocn.it', 'iprima.cz','andaluciainformacion.es','nyweekly.com','intermediatv.ro', 'businessinsider.com','20minutes.fr','westword.com','muscleandfitness.com', 'baltimoresun.com','ctinsider.com','ladbible.com','vanguardia.com','epoznan.pl', 'newcastleherald.com.au','corkbeo.ie','cultmtl.com','sunraysiadaily.com.au', 'observer-reporter.com','tastingtable.com','yardbarker.com','pastemagazine.com', 'index.hr','sport.fr','traveldailynews.com','podkarpacielive.pl','gigwise.com', 'lifeandstylemag.com','varsity.co.uk','oberberg-aktuell.de','buzzwebzine.fr', 'gathered.how','ma-grande-taille.com','viepratique.fr','les-docus.com', 'copenhagenfashionsummit.com','yourhealthmagazine.net','primalamartesana.it', 'radiortm.it','termometropolitico.it','cicloweb.it','forschung-und-wissen.de', 'theexeterdaily.co.uk','emilyluxton.co.uk','barbedudaron.fr','vocal.media', 'paperblog.com'} BRAND_PROFILE={'qanomed.com','aslitarcanglobal.com','sapphirehairclinic.com','provenexpert.com', 'alopezie.de','drcinik.com','charlesmedicalgroup.com','shapiromedical.com', 'bestclinic.co.uk','traya.health','blue-print.co','myhair.ai','verasmile.com', 'bizzr.uk','hashtagbeauty.de','meinhaushalt.at'} def classify(b): dom=b.get('root_name_source','') or '' url=b.get('url_from','') or '' if dom in PR_DOMAINS: return 'press_release' if b.get('is_sponsored') or '/sponsored/' in url or '/paroles-de-partenaires/' in url or '/friday-partner/' in url: return 'sponsored_article' if dom in REAL_MEDIA: return 'advertorial' if dom in BRAND_PROFILE: return 'other' return 'guest_post' # ---- Country mapping ---- TLD_COUNTRY={'fr':'FR','de':'DE','it':'IT','es':'ES','cz':'CZ','at':'AT','ch':'CH','co.uk':'GB', 'uk':'GB','ro':'RO','com.br':'BR','ne.jp':'JP','jp':'JP','kr':'KR','co.kr':'KR', 'gen.tr':'TR','tr':'TR','az':'AZ','su':'RU','nl':'NL','gt':'GT','pl':'PL','id':'ID', 'vn':'VN'} LANG_COUNTRY={'fr':'FR','de':'DE','it':'IT','es':'ES','cs':'CZ','pt':'BR','ko':'KR','ja':'JP', 'tr':'TR','ru':'RU','ar':None,'id':'ID','ro':'RO','nl':'NL','sv':'SE','en':None} def country_of(b): dom=b.get('root_name_source','') or '' parts=dom.split('.') # try two-level tld then one-level for n in (2,1): suf='.'.join(parts[-n:]) if suf in TLD_COUNTRY: return TLD_COUNTRY[suf] if parts and parts[-1] in TLD_COUNTRY: return TLD_COUNTRY[parts[-1]] langs=b.get('languages') or [] if langs: c=LANG_COUNTRY.get(langs[0]) if c: return c if langs[0]=='en': return 'GB' return None INT32_MAX=2147483647 def transform(b): traffic=b.get('traffic',0) or 0 td=b.get('traffic_domain',0) or 0 langs=b.get('languages') or [] eff_traffic=traffic if traffic>0 else td if eff_traffic>INT32_MAX: eff_traffic=INT32_MAX # DB column is 32-bit Int return { 'url': b['url_from'], 'domain': b['root_name_source'], 'domain_rating': int(round(b.get('domain_rating_source') or 0)), 'traffic': eff_traffic, 'country': country_of(b), 'language': langs[0] if langs else None, 'category': classify(b), 'is_dofollow': bool(b.get('is_dofollow')), 'first_seen': (b.get('first_seen_link') or '')[:10], 'comment': "", } TARGETS={'dr-cinik':'emrahcinik.com','dr-serkan-aygin':'drserkanaygin.com', 'cosmedica':'cosmedica.com','vera-clinic':'veraclinic.net', 'clinicana':'clinicana.com','smile-hair-clinic':'smilehairclinic.com'} summary={} for slug,loader,path in ( [(s,load_wrapped,p) for s,p in FILES_WRAPPED.items()] + [(s,load_plain,p) for s,p in FILES_PLAIN.items()]): bl=loader(path) raw_count=len(bl) kept=[] seen_urls=set(); seen_domains=set() excl=0 for b in bl: ex,reason=is_excluded(b) if ex: excl+=1; continue url=b['url_from']; dom=b['root_name_source'] if url in seen_urls: continue if dom in seen_domains: continue seen_urls.add(url); seen_domains.add(dom) kept.append(transform(b)) # category breakdown cats={} for k in kept: cats[k['category']]=cats.get(k['category'],0)+1 # top placements top=sorted(kept,key=lambda x:-x['domain_rating'])[:3] out={'meta':{'target':TARGETS[slug],'period':'2026-05','total':len(kept)},'backlinks':kept} with open(f"/tmp/cinik-bl-{slug}.json",'w') as f: json.dump(out,f,ensure_ascii=False,indent=1) summary[slug]={'raw':raw_count,'excluded':excl,'kept':len(kept),'cats':cats, 'top':[(t['domain'],t['domain_rating']) for t in top]} print(json.dumps(summary,ensure_ascii=False,indent=2))