analyzthis_design 2.0.0 → 2.1.0

package/dist/lib/collect.js

@@ -1,2 +1,859 @@
 #!/usr/bin/env node
-'use strict';const fs=require('fs'),path=require('path'),os=require('os'),crypto=require('crypto'),session=require('./session'),knowledge=require('./knowledge'),cache=require('./cache'),research=require('./research'),{discoverKnowledgeSources,writeSourceManifestNotes,connectDiscoveredSources}=require('./source-discovery'),{resolvePackageRoot}=require('./platforms'),PACKAGE_ROOT=resolvePackageRoot(__dirname),HARD_EXCLUDES=new Set(['node_modules','dist','build','.next','.nuxt','coverage','.git','.turbo','.vercel','.cache','out','.analyzthis_design','vendor','__pycache__','.venv','venv','target']),MAX_FILE_BYTES=0x30d40,MAX_EXTRACT_LINES=0x28,BATCH_SIZE=0x6,DEFAULT_WEB_LIMIT=0xa,WEB_SOURCE_FOLDERS=new Set(['PRDs','Product','Research','Design']),URL_RE=/https?:\/\/[^\s)'">\]`]+/gi;function loadCollectConfig(){const a=path['join'](os['homedir'](),'.analyzthis_design','config.json');if(!fs['existsSync'](a))return{};try{return JSON['parse'](fs['readFileSync'](a,'utf8'));}catch{return{};}}function normalizeUrl(a){return String(a)['replace'](/^[`'"]+|[`'"]+$/g,'')['replace'](/[.,;:!?)]+$/g,'');}function isResearchableUrl(a){if(!a||a['length']<0xc)return![];if(/[{}]/['test'](a))return![];try{const b=new URL(a);if(!['http:','https:']['includes'](b['protocol']))return![];const c=b['hostname']['toLowerCase']();if(!c['includes']('.'))return![];if(c==='localhost'||c['startsWith']('127.')||c==='0.0.0.0')return![];const d=b['pathname']['toLowerCase']();if(/\.(png|jpe?g|gif|svg|ico|webp|woff2?|ttf|eot|css|js|map|zip|tar|gz|pdf)$/i['test'](d))return![];if(c==='www.npmjs.com'||c==='npmjs.com')return![];if(c==='example.com'||c==='www.example.com')return![];if(c==='github.com'&&/\/(blob|raw|tree)\//['test'](b['pathname']))return![];return!![];}catch{return![];}}function extractUrlsFromText(a){const b=String(a)['match'](URL_RE)||[];return b['map'](normalizeUrl)['filter'](isResearchableUrl);}function discoverWebUrls({cwd:a,notes:b,config:c,webLimit:d}){const e=c['collect']||{},f=d??e['web_limit']??DEFAULT_WEB_LIMIT,g=new Set(),h=[],i=(k,l)=>{const m=normalizeUrl(k);if(!isResearchableUrl(m)||g['has'](m))return;g['add'](m),h['push']({'url':m,'source':l});};for(const k of[...e['web_urls']||[],...c['research']?.['urls']||[]]){i(k,'config');}if(e['web_from_repo']!==![]){const l=path['join'](a,'README.md');if(fs['existsSync'](l))try{for(const m of extractUrlsFromText(fs['readFileSync'](l,'utf8')))i(m,'README.md');}catch{}for(const n of b){if(!WEB_SOURCE_FOLDERS['has'](n['classifier']['folder']))continue;const o=path['join'](a,n['sourceRel']);try{for(const p of extractUrlsFromText(fs['readFileSync'](o,'utf8')))i(p,n['sourceRel']);}catch{}}}const j=e['web_queries']||c['research']?.['queries']||[];return{'urls':h['slice'](0x0,f),'queries':j['slice'](0x0,f),'limit':f,'totalDiscovered':h['length']};}async function collectWebResearch({urls:a,queries:b,projectId:c,dryRun:d}){const e={'fetched':[],'failed':[],'queries':[],'dryRun':!!d};if(d)return e['urls']=a,e;for(const {url:f,source:g}of a){try{console['log']('\x20\x20\x20•\x20'+f+'\x20\x20(from\x20'+g+')');const h=await research['researchUrl']({'url':f,'project':c});e['fetched']['push']({'url':f,'source':g,'chars':h['chars']});}catch(i){e['failed']['push']({'url':f,'source':g,'error':i['message']}),console['log']('\x20\x20\x20\x20\x20⚠\x20\x20'+i['message']);}}for(const j of b){try{console['log']('\x20\x20\x20•\x20query:\x20\x22'+j+'\x22');const k=await research['researchQuery']({'query':j,'project':c});e['queries']['push']({'query':j,'mode':k['mode'],'chars':k['chars']});}catch(l){e['failed']['push']({'query':j,'error':l['message']}),console['log']('\x20\x20\x20\x20\x20⚠\x20\x20'+l['message']);}}return e;}const CLASSIFIERS=[{'folder':'PRDs','category':'prd','tags':['prd'],'match':(a,b)=>/^(readme|contributing|changelog)\.md$/i['test'](b)||/\b(prd|requirements|user[-_]?stor|acceptance|epic|jtbd)\b/i['test'](a)||/^docs\//i['test'](a)&&/\.md$/i['test'](b)},{'folder':'Brand','category':'brand','tags':['brand','design'],'match':(a,b)=>/tailwind\.config\./i['test'](b)||/^(globals|global|tokens|theme|variables)\.(css|scss|ts|js)$/i['test'](b)||/design[-_]?system|brand|tokens?\./i['test'](a)||/\.css$/i['test'](b)&&/(theme|token|color|brand)/i['test'](a)},{'folder':'Product','category':'product','tags':['product'],'match':a=>/\b(roadmap|vision|product|okr|kpi|north[-_]?star)\b/i['test'](a)&&/\.md$/i['test'](a)},{'folder':'Research','category':'research','tags':['research'],'match':a=>/\b(research|interview|survey|insight|persona|usability)\b/i['test'](a)&&/\.md$/i['test'](a)},{'folder':'Pages','category':'design','tags':['design','page'],'match':(a,b)=>/(^|\/)(app|pages|src\/pages)\//i['test'](a)&&(/page\.(tsx|jsx|ts|js|vue|svelte)$/i['test'](b)||/^index\.(tsx|jsx|ts|js)$/i['test'](b))},{'folder':'Components','category':'design','tags':['design','component'],'match':(a,b)=>/(^|\/)(components|src\/ui|src\/components|ui)\//i['test'](a)&&/\.(tsx|jsx|vue|svelte)$/i['test'](b)&&!/\.(test|spec|stories)\./i['test'](b)},{'folder':'Design','category':'design','tags':['design'],'match':(a,b)=>/\.md$/i['test'](b)&&/\b(design|figma|wireframe|layout|ux|ui)\b/i['test'](a)},{'folder':'Tech','category':'tech','tags':['tech'],'match':(a,b)=>b==='package.json'||/^(tsconfig|next\.config|vite\.config|astro\.config|nuxt\.config)/i['test'](b)||/(^|\/)(api|app\/api|pages\/api|server)\//i['test'](a)&&/\.(ts|js)$/i['test'](b)}];function loadGitignore(a){const b=[],c=path['join'](a,'.gitignore');if(!fs['existsSync'](c))return b;for(const d of fs['readFileSync'](c,'utf8')['split']('\x0a')){const e=d['trim']();if(!e||e['startsWith']('#')||e['startsWith']('!'))continue;b['push'](e['replace'](/\/$/,''));}return b;}function isIgnored(a,b){const c=a['split'](path['sep']);if(c['some'](d=>HARD_EXCLUDES['has'](d)||d['startsWith']('.'))){if(c['some'](d=>HARD_EXCLUDES['has'](d)))return!![];if(c['some'](d=>d['startsWith']('.')&&d!=='.'))return!![];}for(const d of b){const e=d['replace'](/^\*\*\//,'')['replace'](/\*\*/g,'');if(!e)continue;if(a===d||a['startsWith'](d+path['sep'])||a['includes'](path['sep']+d+path['sep']))return!![];if(d['endsWith']('*')&&a['startsWith'](d['slice'](0x0,-0x1)))return!![];if(d['startsWith']('*.')&&a['endsWith'](d['slice'](0x1)))return!![];}return![];}function walkFiles(a,b,c=[]){const d=e=>{let f;try{f=fs['readdirSync'](e,{'withFileTypes':!![]});}catch{return;}for(const g of f){const h=path['join'](e,g['name']),i=path['relative'](a,h);if(isIgnored(i,b))continue;if(g['isDirectory']())d(h);else{if(g['isFile']())c['push']({'full':h,'rel':i,'base':g['name']});}}};return d(a),c;}function sha1(a){return crypto['createHash']('sha1')['update'](a)['digest']('hex')['slice'](0x0,0xc);}function slugify(a){return String(a)['replace'](/\.[^.]+$/,'')['replace'](/[^a-zA-Z0-9-_]+/g,'-')['replace'](/^-+|-+$/g,'')['slice'](0x0,0x50)||'note';}function extractSnippet(a){let b='';try{const g=fs['statSync'](a);g['size']>MAX_FILE_BYTES?b=fs['readFileSync'](a,'utf8')['slice'](0x0,MAX_FILE_BYTES):b=fs['readFileSync'](a,'utf8');}catch{return{'lines':[],'imports':[],'exports':[],'figma':[]};}const c=b['split']('\x0a')['slice'](0x0,MAX_EXTRACT_LINES),d=[],e=[],f=[];for(const h of b['split']('\x0a')['slice'](0x0,0xc8)){const i=h['match'](/from\s+['"]([^'"]+)['"]/);if(i)d['push'](i[0x1]);(/^export\s+(default\s+)?(function|const|class|async)/['test'](h['trim']())||/^export\s*\{/['test'](h['trim']()))&&e['push'](h['trim']()['slice'](0x0,0x78));const j=h['match'](/https?:\/\/(?:www\.)?figma\.com\/[^\s)'"]+/gi);if(j)f['push'](...j);}return{'lines':c,'imports':[...new Set(d)]['slice'](0x0,0x14),'exports':e['slice'](0x0,0xf),'figma':[...new Set(f)]['slice'](0x0,0x5)};}function classifyFile(a,b){for(const d of CLASSIFIERS){if(d['match'](a,b))return d;}return null;}function frontmatter(a){const b=['---'];for(const [c,d]of Object['entries'](a)){if(Array['isArray'](d))b['push'](c+':\x20['+d['map'](e=>JSON['stringify'](e))['join'](',\x20')+']');else{if(typeof d==='boolean')b['push'](c+':\x20'+d);else b['push'](c+':\x20'+JSON['stringify'](String(d)));}}return b['push']('---'),b['join']('\x0a');}function buildDraftNote({title:a,rel:b,classifier:c,extract:d,related:e}){const f=new Date()['toISOString']()['split']('T')[0x0],g=sha1(d['lines']['join']('\x0a')),h=frontmatter({'title':a,'tags':c['tags'],'source':b,'category':c['category'],'collected_at':f,'content_hash':g,'enriched':![]}),i=e['length']?e['map'](k=>'-\x20[['+k+']]')['join']('\x0a'):'-\x20_(none\x20yet)_',j=['##\x20Purpose','Draft\x20extract\x20from\x20`'+b+'`.\x20Awaiting\x20Kavi\x20enrichment.','','##\x20Raw\x20extract','```',d['lines']['join']('\x0a')['slice'](0x0,0xbb8),'```','','##\x20Exports',d['exports']['length']?d['exports']['map'](k=>'-\x20`'+k+'`')['join']('\x0a'):'-\x20_(none\x20detected)_','','##\x20Imports',d['imports']['length']?d['imports']['map'](k=>'-\x20`'+k+'`')['join']('\x0a'):'-\x20_(none\x20detected)_','',...d['figma']['length']?['##\x20Figma\x20links',...d['figma']['map'](k=>'-\x20'+k),'']:[],'##\x20Related',i,'']['join']('\x0a');return{'fm':h,'body':j,'contentHash':g};}function scanAndDraft({cwd:a,vaultPath:b,dryRun:d}){const e=loadGitignore(a),g=walkFiles(a,e),h=[],i=[];for(const l of g){const o=classifyFile(l['rel'],l['base']);if(!o)continue;const p=extractSnippet(l['full']),q=slugify(l['base']);i['push']({...l,'classifier':o,'extract':p,'title':q});}const j=i['filter'](r=>r['classifier']['folder']==='Pages')['map'](r=>'Pages/'+r['title']),k=i['filter'](r=>r['classifier']['folder']==='Components')['map'](r=>'Components/'+r['title']);for(const r of i){let s=[];if(r['classifier']['folder']==='Components')s=j['filter'](y=>{const z=y['split']('/')[0x1]['toLowerCase']();return r['extract']['imports']['some'](A=>A['toLowerCase']()['includes'](r['title']['toLowerCase']()))||z['includes'](r['title']['toLowerCase']()['slice'](0x0,0x6));})['slice'](0x0,0x5);else r['classifier']['folder']==='Pages'&&(s=k['filter'](y=>{const z=y['split']('/')[0x1]['toLowerCase']();return r['extract']['imports']['some'](A=>A['toLowerCase']()['includes'](z));})['slice'](0x0,0x8));const {fm:t,body:u,contentHash:v}=buildDraftNote({'title':r['title'],'rel':r['rel'],'classifier':r['classifier'],'extract':r['extract'],'related':s});let w=path['join'](r['classifier']['folder'],r['title']+'.md'),x=0x2;while(h['some'](y=>y['vaultRel']===w)){w=path['join'](r['classifier']['folder'],r['title']+'-'+x+'.md'),x+=0x1;}h['push']({'vaultRel':w,'fullOut':path['join'](b,w),'title':r['title'],'classifier':r['classifier'],'contentHash':v,'draft':t+'\x0a\x0a'+u,'sourceRel':r['rel']});}if(!d){fs['mkdirSync'](b,{'recursive':!![]});for(const y of h){if(fs['existsSync'](y['fullOut']))try{const z=fs['readFileSync'](y['fullOut'],'utf8'),A=z['match'](/content_hash:\s*"?([a-f0-9]+)"?/),B=/enriched:\s*true/['test'](z);if(A&&A[0x1]===y['contentHash']&&B){y['skipped']=!![],y['draft']=z;continue;}}catch{}fs['mkdirSync'](path['dirname'](y['fullOut']),{'recursive':!![]}),fs['writeFileSync'](y['fullOut'],y['draft']);}writeMeta(b,h,a);}return{'filesScanned':g['length'],'notesWritten':h['filter'](C=>!C['skipped'])['length'],'notesSkipped':h['filter'](C=>C['skipped'])['length'],'notes':h};}function writeMeta(a,b,c){const d={};for(const f of b){const g=f['classifier']['folder'];if(!d[g])d[g]=[];d[g]['push'](f);}const e=['---','title:\x20\x22Knowledge\x20Vault\x20Index\x22','tags:\x20[meta,\x20moc]','collected_at:\x20\x22'+new Date()['toISOString']()['split']('T')[0x0]+'\x22','---','','#\x20Knowledge\x20Vault\x20Index','','Source\x20repo:\x20`'+c+'`','','##\x20Map\x20of\x20content',''];for(const [h,i]of Object['entries'](d)['sort']()){e['push']('###\x20'+h,'');for(const j of i){const k=j['vaultRel']['replace'](/\.md$/,'')['replace'](/\\/g,'/');e['push']('-\x20[['+k+'|'+j['title']+']]\x20—\x20`'+j['sourceRel']+'`');}e['push']('');}fs['mkdirSync'](path['join'](a,'_meta'),{'recursive':!![]}),fs['writeFileSync'](path['join'](a,'_meta','index.md'),e['join']('\x0a')),fs['writeFileSync'](path['join'](a,'_meta','last-sync.md'),['---','title:\x20\x22Last\x20sync\x22','---','','#\x20Last\x20sync','','-\x20**When:**\x20'+new Date()['toISOString'](),'-\x20**Repo:**\x20'+c,'-\x20**Notes:**\x20'+b['length'],'']['join']('\x0a'));}function loadKaviCard(){const a=path['join'](PACKAGE_ROOT,'agents','cards','kavi.md');if(fs['existsSync'](a))return fs['readFileSync'](a,'utf8');return'You\x20are\x20Kavi,\x20the\x20Knowledge\x20Archivist.\x20Rewrite\x20draft\x20extracts\x20into\x20concise\x20Obsidian\x20notes.\x20Do\x20not\x20invent\x20files\x20or\x20APIs.';}function hasAnyApiKey(){return!!(process.env.ANTHROPIC_API_KEY||process.env.OPENAI_API_KEY||process.env.GEMINI_API_KEY||process.env.GOOGLE_API_KEY||process.env.ZAI_API_KEY||process.env.ZHIPU_API_KEY);}function pickEnrichProvider(a){if(process.env.OPENAI_API_KEY)return{'provider':'openai','model':a['orchestrator']?.['tiers']?.['structured']?.['model']||'gpt-4o-mini'};if(process.env.GEMINI_API_KEY||process.env.GOOGLE_API_KEY)return{'provider':'google','model':'gemini-2.5-flash'};if(process.env.ZAI_API_KEY||process.env.ZHIPU_API_KEY)return{'provider':'zai','model':'glm-4.5-flash'};if(process.env.ANTHROPIC_API_KEY)return{'provider':'anthropic','model':'claude-haiku-4-5-20251001'};return null;}function parseEnrichedBatch(a,b){const c=a['split'](/---NOTE\s+\d+---/i)['map'](d=>d['trim']())['filter'](Boolean);if(c['length']>=b)return c['slice'](0x0,b);if(c['length']===0x1&&b===0x1)return c;while(c['length']<b)c['push'](null);return c['slice'](0x0,b);}async function enrichNotes({notes:a,limit:b,provider:c,model:d}){const {callLlm:e,loadConfig:f}=require('./orchestrator/run'),g=f(),h=c&&d?{'provider':c,'model':d}:pickEnrichProvider(g);if(!h)return{'enriched':0x0,'skipped':a['length'],'reason':'No\x20API\x20key\x20set\x20—\x20wrote\x20draft-only\x20vault.','llm_calls':0x0,'input_tokens_est':0x0,'output_tokens_est':0x0};const k=a['filter'](q=>!q['skipped'])['slice'](0x0,b==null?a['length']:b),l=[loadKaviCard(),'','For\x20each\x20draft\x20note\x20below,\x20rewrite\x20ONLY\x20the\x20body\x20(keep\x20the\x20YAML\x20frontmatter\x20I\x20give\x20you).','Output\x20format:\x20for\x20each\x20note,\x20emit\x20exactly:','---NOTE\x201---','<full\x20markdown\x20including\x20the\x20original\x20frontmatter\x20with\x20enriched:\x20true>','---NOTE\x202---','...','Rules:\x20purpose\x20+\x20key\x20facts\x20+\x20Related\x20wikilinks.\x20Do\x20NOT\x20invent\x20files,\x20APIs,\x20or\x20components\x20not\x20in\x20the\x20Raw\x20extract.']['join']('\x0a');let m=0x0,n=0x0,o=0x0,p=0x0;for(let q=0x0;q<k['length'];q+=BATCH_SIZE){const r=k['slice'](q,q+BATCH_SIZE),s=r['map']((v,w)=>'---NOTE\x20'+(w+0x1)+'---\x0a'+v['draft'])['join']('\x0a\x0a'),t=await e({'provider':h['provider'],'model':h['model'],'system':l,'user':s,'maxTokens':0xfa0});n+=0x1,o+=Math['ceil']((l['length']+s['length'])/0x4),p+=Math['ceil'](t['length']/0x4);const u=parseEnrichedBatch(t,r['length']);for(let v=0x0;v<r['length'];v++){let w=u[v];if(!w)continue;!/enriched:\s*true/['test'](w)&&(w=w['replace'](/enriched:\s*false/,'enriched:\x20true'),!/enriched:/['test'](w)&&w['startsWith']('---')&&(w=w['replace'](/^---\n/,'---\x0aenriched:\x20true\x0a'))),!/content_hash:/['test'](w)&&r[v]['contentHash']&&(w=w['replace'](/^---\n/,'---\x0acontent_hash:\x20\x22'+r[v]['contentHash']+'\x22\x0a')),fs['writeFileSync'](r[v]['fullOut'],w),r[v]['draft']=w,m+=0x1;}}return{'enriched':m,'skipped':a['length']-m,'reason':null,'llm_calls':n,'input_tokens_est':o,'output_tokens_est':p,'provider':h['provider'],'model':h['model']};}async function collect(a={}){const b=path['resolve'](a['cwd']||process['cwd']()),c=a['project']||session['getProjectId'](b),e=path['resolve'](a['vault']||path['join'](os['homedir'](),'.analyzthis_design','vaults',c)),f=a['enrich']!==![],g=!!a['dryRun'],h=a['limit']!=null?a['limit']:null,i=a['web']!==![],j=a['discoverSources']!==![],k=loadCollectConfig(),{resolveTargets:l}=require('./platforms'),m=l(a['target']||'cursor')||['cursor'];!session['show']({'project':c})&&session['init']({'project':c});const o=[];if(g)o['push']('dry-run');else{if(j)o['push']('discover');if(i)o['push']('web');if(f)o['push']('enrich');o['push']('sync');}console['log']('\x0a📚\x20Kavi\x20—\x20Collect\x20Knowledge'),console['log']('\x20\x20\x20Repo:\x20\x20'+b),console['log']('\x20\x20\x20Vault:\x20'+e),console['log']('\x20\x20\x20Mode:\x20\x20'+(o['join']('\x20+\x20')||'scan')+'\x0a');const q=scanAndDraft({'cwd':b,'vaultPath':e,'dryRun':g}),r=j?discoverKnowledgeSources({'cwd':b,'config':k}):[];console['log']('\x20\x20\x20Scanned\x20'+q['filesScanned']+'\x20file(s)\x20→\x20'+q['notes']['length']+'\x20note(s)\x20('+q['notesWritten']+'\x20written,\x20'+q['notesSkipped']+'\x20unchanged)');if(j&&r['length']>0x0){console['log']('\x0a🔍\x20Knowledge\x20sources:\x20'+r['length']+'\x20discovered\x20(Obsidian\x20vaults,\x20wikis,\x20referenced\x20paths)');for(const B of r['slice'](0x0,0x6)){console['log']('\x20\x20\x20•\x20['+B['confidence']+']\x20'+B['type']+':\x20'+B['path']);}if(r['length']>0x6)console['log']('\x20\x20\x20…\x20and\x20'+(r['length']-0x6)+'\x20more');}else j&&!g&&console['log']('\x0a🔍\x20Knowledge\x20sources:\x20none\x20discovered\x20(add\x20collect.source_paths\x20in\x20config\x20or\x20link\x20vaults\x20in\x20README)');const s=i?discoverWebUrls({'cwd':b,'notes':q['notes'],'config':k,'webLimit':a['webLimit']}):{'urls':[],'queries':[],'limit':0x0,'totalDiscovered':0x0};if(i&&(s['urls']['length']>0x0||s['queries']['length']>0x0))console['log']('\x0a🌐\x20Web\x20research:\x20'+s['urls']['length']+'\x20URL(s)'+(s['queries']['length']?',\x20'+s['queries']['length']+'\x20query\x20stub(s)':'')+(s['totalDiscovered']>s['urls']['length']?'\x20('+s['totalDiscovered']+'\x20found,\x20capped\x20at\x20'+s['limit']+')':''));else i&&!g&&console['log']('\x0a🌐\x20Web\x20research:\x20no\x20URLs\x20found\x20(add\x20collect.web_urls\x20in\x20~/.analyzthis_design/config.json\x20or\x20links\x20in\x20PRDs/README)');if(g){const C={};for(const D of q['notes']){C[D['classifier']['folder']]=(C[D['classifier']['folder']]||0x0)+0x1;}return console['log']('\x0a──\x20Dry\x20run\x20preview\x20──'),console['log'](JSON['stringify']({'vaultPath':e,'projectId':c,'byFolder':C,'knowledgeSources':r['map'](E=>({'path':E['path'],'type':E['type'],'confidence':E['confidence'],'discovered_from':E['discovered_from'],'markdown_files':E['stats']['markdown_files']})),'webUrls':s['urls']['map'](E=>({'url':E['url'],'source':E['source']})),'webQueries':s['queries'],'sample':q['notes']['slice'](0x0,0x8)['map'](E=>E['vaultRel'])},null,0x2)),console['log'](''),{'vaultPath':e,'projectId':c,'dryRun':!![],...q,'sources':r,'web':s,'enrich':null,'sync':null};}let u={'written':[]};j&&r['length']>0x0&&(console['log']('\x0a⏳\x20Writing\x20source\x20manifest\x20notes\x20(Sources/*.md)...'),u=writeSourceManifestNotes({'discoveries':r,'vaultPath':e,'dryRun':![]}),console['log']('\x20\x20\x20Wrote\x20'+u['written']['length']+'\x20manifest\x20note(s)\x20+\x20_meta/knowledge-sources.md'));let v={'fetched':[],'failed':[],'queries':[]};i&&(s['urls']['length']>0x0||s['queries']['length']>0x0)&&(console['log']('\x0a⏳\x20Fetching\x20web\x20research\x20into\x20web-context.md...'),v=await collectWebResearch({'urls':s['urls'],'queries':s['queries'],'projectId':c,'dryRun':![]}),console['log']('\x20\x20\x20Fetched\x20'+v['fetched']['length']+'\x20URL(s)'+(v['failed']['length']?',\x20'+v['failed']['length']+'\x20failed':'')));let w={'enriched':0x0,'skipped':q['notes']['length'],'reason':'enrichment\x20disabled','llm_calls':0x0,'input_tokens_est':0x0,'output_tokens_est':0x0};f&&(!hasAnyApiKey()?(console['log']('\x0a\x20\x20⚠\x20\x20No\x20LLM\x20API\x20key\x20found\x20—\x20writing\x20draft-only\x20vault.\x20Set\x20OPENAI_API_KEY\x20/\x20GEMINI_API_KEY\x20/\x20ANTHROPIC_API_KEY\x20/\x20ZAI_API_KEY\x20to\x20enrich.'),w['reason']='No\x20API\x20key\x20set\x20—\x20wrote\x20draft-only\x20vault.'):(console['log']('\x0a⏳\x20Enriching\x20notes\x20with\x20Kavi\x20(LLM)...'),w=await enrichNotes({'notes':q['notes'],'limit':h,'provider':a['provider'],'model':a['model']}),console['log']('\x20\x20\x20Enriched\x20'+w['enriched']+'\x20note(s)'+(w['reason']?'\x20('+w['reason']+')':''))));console['log']('\x0a⏳\x20Wiring\x20knowledge\x20bank...');const x=j?connectDiscoveredSources({'discoveries':r,'kaviVaultPath':e,'config':k,'autoConnect':k['collect']?.['auto_connect_discovered']!==![]}):{'connected':[],'skipped':[]};if(x['connected']['length']>0x0){console['log']('\x20\x20\x20Auto-connected\x20'+x['connected']['length']+'\x20source(s):');for(const E of x['connected'])console['log']('\x20\x20\x20•\x20'+E);}knowledge['connect']({'vaultPath':e,'tags':[],'include':[]});const y=knowledge['sync']({'targets':m});cache['invalidatePrefix']('kb:');if(y['message'])console['log']('\x20\x20⚠\x20\x20'+y['message']);else{console['log']('✅\x20Synced\x20'+y['synced']+'\x20note(s)\x20into\x20the\x20knowledge\x20bank.');for(const F of y['copiedTo']||[])console['log']('\x20\x20\x20•\x20'+F);}const z=session['show']({'project':c})||{},A=Array['isArray'](z['content_sources'])?z['content_sources']:[];if(!A['includes']('codebase_collect'))A['push']('codebase_collect');return session['update']({'project':c,'patch':{'content_sources':A,'vault_path':e,'last_collect_at':new Date()['toISOString'](),'digest':{...z['digest']||{},'task_map_summary':'Collected\x20knowledge\x20vault\x20('+q['notes']['length']+'\x20notes)\x20from\x20'+b,'experts':['kavi']},'metrics':{...z['metrics']||{},'llm_calls':w['llm_calls']||0x0,'experts_run':['kavi'],'input_tokens_est':w['input_tokens_est']||0x0,'output_tokens_est':w['output_tokens_est']||0x0,'mode':'collect','effort_log':w['model']?[{'persona':'kavi','effort':'standard','model':w['model'],'scoped_mode':null}]:[],'cost_usd':z['metrics']?.['cost_usd']||0x0}}}),console['log']('\x0a✅\x20Collect\x20complete.'),console['log']('\x20\x20\x20Vault:\x20'+e),console['log']('\x20\x20\x20Open\x20in\x20Obsidian,\x20or\x20run\x20/persona-orchestrator\x20—\x20personas\x20now\x20read\x20this\x20knowledge\x20bank.\x0a'),{'vaultPath':e,'projectId':c,'dryRun':![],'filesScanned':q['filesScanned'],'notes':q['notes']['length'],'notesWritten':q['notesWritten'],'notesSkipped':q['notesSkipped'],'web':v,'sources':{'discovered':r,'manifest':u,'connected':x['connected'],'skipped':x['skipped']},'enrich':w,'sync':y};}module['exports']={'collect':collect,'scanAndDraft':scanAndDraft,'classifyFile':classifyFile,'walkFiles':walkFiles,'discoverWebUrls':discoverWebUrls,'collectWebResearch':collectWebResearch,'extractUrlsFromText':extractUrlsFromText,'isResearchableUrl':isResearchableUrl,'HARD_EXCLUDES':HARD_EXCLUDES};
+'use strict';
+
+/**
+ * Kavi — Collect Knowledge (v1.11)
+ *
+ * Scans the current codebase, writes an Obsidian-compatible vault, optionally
+ * LLM-enriches notes, then auto-connects + syncs into the knowledge bank so
+ * all critique personas read company context first.
+ *
+ * Usage (via CLI):
+ *   npx analyzthis_design collect [--vault path] [--dry-run] [--no-enrich] [--limit N] [--target cursor|all]
+ */
+
+const fs     = require('fs');
+const path   = require('path');
+const os     = require('os');
+const crypto = require('crypto');
+const session  = require('./session');
+const knowledge = require('./knowledge');
+const cache    = require('./cache');
+const research = require('./research');
+const {
+  discoverKnowledgeSources,
+  writeSourceManifestNotes,
+  connectDiscoveredSources,
+} = require('./source-discovery');
+
+const { resolvePackageRoot } = require('./platforms');
+const PACKAGE_ROOT = resolvePackageRoot(__dirname);
+
+const HARD_EXCLUDES = new Set([
+  'node_modules', 'dist', 'build', '.next', '.nuxt', 'coverage', '.git',
+  '.turbo', '.vercel', '.cache', 'out', '.analyzthis_design', 'vendor',
+  '__pycache__', '.venv', 'venv', 'target',
+]);
+
+const MAX_FILE_BYTES = 200_000;
+const MAX_EXTRACT_LINES = 40;
+const BATCH_SIZE = 6;
+const DEFAULT_WEB_LIMIT = 10;
+const WEB_SOURCE_FOLDERS = new Set(['PRDs', 'Product', 'Research', 'Design']);
+const URL_RE = /https?:\/\/[^\s)'">\]`]+/gi;
+
+function loadCollectConfig() {
+  const configPath = path.join(os.homedir(), '.analyzthis_design', 'config.json');
+  if (!fs.existsSync(configPath)) return {};
+  try { return JSON.parse(fs.readFileSync(configPath, 'utf8')); }
+  catch { return {}; }
+}
+
+function normalizeUrl(url) {
+  return String(url)
+    .replace(/^[`'"]+|[`'"]+$/g, '')
+    .replace(/[.,;:!?)]+$/g, '');
+}
+
+function isResearchableUrl(url) {
+  if (!url || url.length < 12) return false;
+  if (/[{}]/.test(url)) return false;
+  try {
+    const u = new URL(url);
+    if (!['http:', 'https:'].includes(u.protocol)) return false;
+    const host = u.hostname.toLowerCase();
+    if (!host.includes('.')) return false;
+    if (host === 'localhost' || host.startsWith('127.') || host === '0.0.0.0') return false;
+    const pathname = u.pathname.toLowerCase();
+    if (/\.(png|jpe?g|gif|svg|ico|webp|woff2?|ttf|eot|css|js|map|zip|tar|gz|pdf)$/i.test(pathname)) return false;
+    if (host === 'www.npmjs.com' || host === 'npmjs.com') return false;
+    if (host === 'example.com' || host === 'www.example.com') return false;
+    if (host === 'github.com' && /\/(blob|raw|tree)\//.test(u.pathname)) return false;
+    return true;
+  } catch {
+    return false;
+  }
+}
+
+function extractUrlsFromText(text) {
+  const found = String(text).match(URL_RE) || [];
+  return found.map(normalizeUrl).filter(isResearchableUrl);
+}
+
+/**
+ * Discover external URLs from config + markdown in PRDs/Product/Research/Design/README.
+ */
+function discoverWebUrls({ cwd, notes, config, webLimit }) {
+  const collectCfg = config.collect || {};
+  const limit = webLimit ?? collectCfg.web_limit ?? DEFAULT_WEB_LIMIT;
+  const seen = new Set();
+  const urls = [];
+
+  const add = (url, source) => {
+    const u = normalizeUrl(url);
+    if (!isResearchableUrl(u) || seen.has(u)) return;
+    seen.add(u);
+    urls.push({ url: u, source });
+  };
+
+  for (const u of [...(collectCfg.web_urls || []), ...(config.research?.urls || [])]) {
+    add(u, 'config');
+  }
+
+  if (collectCfg.web_from_repo !== false) {
+    const readme = path.join(cwd, 'README.md');
+    if (fs.existsSync(readme)) {
+      try {
+        for (const u of extractUrlsFromText(fs.readFileSync(readme, 'utf8'))) add(u, 'README.md');
+      } catch { /* ignore */ }
+    }
+    for (const note of notes) {
+      if (!WEB_SOURCE_FOLDERS.has(note.classifier.folder)) continue;
+      const full = path.join(cwd, note.sourceRel);
+      try {
+        for (const u of extractUrlsFromText(fs.readFileSync(full, 'utf8'))) add(u, note.sourceRel);
+      } catch { /* ignore */ }
+    }
+  }
+
+  const queries = collectCfg.web_queries || config.research?.queries || [];
+  return { urls: urls.slice(0, limit), queries: queries.slice(0, limit), limit, totalDiscovered: urls.length };
+}
+
+/**
+ * Fetch discovered URLs (and optional query stubs) into session web-context.md.
+ */
+async function collectWebResearch({ urls, queries, projectId, dryRun }) {
+  const results = { fetched: [], failed: [], queries: [], dryRun: !!dryRun };
+  if (dryRun) {
+    results.urls = urls;
+    return results;
+  }
+
+  for (const { url, source } of urls) {
+    try {
+      console.log(`   • ${url}  (from ${source})`);
+      const r = await research.researchUrl({ url, project: projectId });
+      results.fetched.push({ url, source, chars: r.chars });
+    } catch (err) {
+      results.failed.push({ url, source, error: err.message });
+      console.log(`     ⚠  ${err.message}`);
+    }
+  }
+
+  for (const query of queries) {
+    try {
+      console.log(`   • query: "${query}"`);
+      const r = await research.researchQuery({ query, project: projectId });
+      results.queries.push({ query, mode: r.mode, chars: r.chars });
+    } catch (err) {
+      results.failed.push({ query, error: err.message });
+      console.log(`     ⚠  ${err.message}`);
+    }
+  }
+
+  return results;
+}
+
+// ─── Classification rules (source → vault folder + category) ─────────────────
+
+const CLASSIFIERS = [
+  {
+    folder: 'PRDs',
+    category: 'prd',
+    tags: ['prd'],
+    match: (rel, base) =>
+      /^(readme|contributing|changelog)\.md$/i.test(base) ||
+      /\b(prd|requirements|user[-_]?stor|acceptance|epic|jtbd)\b/i.test(rel) ||
+      /^docs\//i.test(rel) && /\.md$/i.test(base),
+  },
+  {
+    folder: 'Brand',
+    category: 'brand',
+    tags: ['brand', 'design'],
+    match: (rel, base) =>
+      /tailwind\.config\./i.test(base) ||
+      /^(globals|global|tokens|theme|variables)\.(css|scss|ts|js)$/i.test(base) ||
+      /design[-_]?system|brand|tokens?\./i.test(rel) ||
+      /\.css$/i.test(base) && /(theme|token|color|brand)/i.test(rel),
+  },
+  {
+    folder: 'Product',
+    category: 'product',
+    tags: ['product'],
+    match: (rel) => /\b(roadmap|vision|product|okr|kpi|north[-_]?star)\b/i.test(rel) && /\.md$/i.test(rel),
+  },
+  {
+    folder: 'Research',
+    category: 'research',
+    tags: ['research'],
+    match: (rel) => /\b(research|interview|survey|insight|persona|usability)\b/i.test(rel) && /\.md$/i.test(rel),
+  },
+  {
+    folder: 'Pages',
+    category: 'design',
+    tags: ['design', 'page'],
+    match: (rel, base) =>
+      /(^|\/)(app|pages|src\/pages)\//i.test(rel) &&
+      (/page\.(tsx|jsx|ts|js|vue|svelte)$/i.test(base) || /^index\.(tsx|jsx|ts|js)$/i.test(base)),
+  },
+  {
+    folder: 'Components',
+    category: 'design',
+    tags: ['design', 'component'],
+    match: (rel, base) =>
+      /(^|\/)(components|src\/ui|src\/components|ui)\//i.test(rel) &&
+      /\.(tsx|jsx|vue|svelte)$/i.test(base) &&
+      !/\.(test|spec|stories)\./i.test(base),
+  },
+  {
+    folder: 'Design',
+    category: 'design',
+    tags: ['design'],
+    match: (rel, base) =>
+      /\.md$/i.test(base) && /\b(design|figma|wireframe|layout|ux|ui)\b/i.test(rel),
+  },
+  {
+    folder: 'Tech',
+    category: 'tech',
+    tags: ['tech'],
+    match: (rel, base) =>
+      base === 'package.json' ||
+      /^(tsconfig|next\.config|vite\.config|astro\.config|nuxt\.config)/i.test(base) ||
+      /(^|\/)(api|app\/api|pages\/api|server)\//i.test(rel) && /\.(ts|js)$/i.test(base),
+  },
+];
+
+// ─── Ignore helpers ──────────────────────────────────────────────────────────
+
+function loadGitignore(cwd) {
+  const patterns = [];
+  const gi = path.join(cwd, '.gitignore');
+  if (!fs.existsSync(gi)) return patterns;
+  for (const line of fs.readFileSync(gi, 'utf8').split('\n')) {
+    const t = line.trim();
+    if (!t || t.startsWith('#') || t.startsWith('!')) continue;
+    patterns.push(t.replace(/\/$/, ''));
+  }
+  return patterns;
+}
+
+function isIgnored(rel, gitPatterns) {
+  const parts = rel.split(path.sep);
+  if (parts.some((p) => HARD_EXCLUDES.has(p) || p.startsWith('.'))) {
+    // Allow .md under docs that aren't hidden dirs; still block .git etc.
+    if (parts.some((p) => HARD_EXCLUDES.has(p))) return true;
+    // Dotfiles at root of scan (e.g. .env) — skip
+    if (parts.some((p) => p.startsWith('.') && p !== '.')) return true;
+  }
+  for (const pat of gitPatterns) {
+    const clean = pat.replace(/^\*\*\//, '').replace(/\*\*/g, '');
+    if (!clean) continue;
+    if (rel === pat || rel.startsWith(pat + path.sep) || rel.includes(path.sep + pat + path.sep)) return true;
+    if (pat.endsWith('*') && rel.startsWith(pat.slice(0, -1))) return true;
+    // Simple glob: *.log
+    if (pat.startsWith('*.') && rel.endsWith(pat.slice(1))) return true;
+  }
+  return false;
+}
+
+function walkFiles(cwd, gitPatterns, out = []) {
+  const walk = (dir) => {
+    let entries;
+    try { entries = fs.readdirSync(dir, { withFileTypes: true }); }
+    catch { return; }
+    for (const entry of entries) {
+      const full = path.join(dir, entry.name);
+      const rel = path.relative(cwd, full);
+      if (isIgnored(rel, gitPatterns)) continue;
+      if (entry.isDirectory()) walk(full);
+      else if (entry.isFile()) out.push({ full, rel, base: entry.name });
+    }
+  };
+  walk(cwd);
+  return out;
+}
+
+// ─── Extract helpers ─────────────────────────────────────────────────────────
+
+function sha1(text) {
+  return crypto.createHash('sha1').update(text).digest('hex').slice(0, 12);
+}
+
+function slugify(name) {
+  return String(name)
+    .replace(/\.[^.]+$/, '')
+    .replace(/[^a-zA-Z0-9-_]+/g, '-')
+    .replace(/^-+|-+$/g, '')
+    .slice(0, 80) || 'note';
+}
+
+function extractSnippet(fullPath) {
+  let raw = '';
+  try {
+    const stat = fs.statSync(fullPath);
+    if (stat.size > MAX_FILE_BYTES) {
+      raw = fs.readFileSync(fullPath, 'utf8').slice(0, MAX_FILE_BYTES);
+    } else {
+      raw = fs.readFileSync(fullPath, 'utf8');
+    }
+  } catch {
+    return { lines: [], imports: [], exports: [], figma: [] };
+  }
+  const lines = raw.split('\n').slice(0, MAX_EXTRACT_LINES);
+  const imports = [];
+  const exports = [];
+  const figma = [];
+  for (const line of raw.split('\n').slice(0, 200)) {
+    const imp = line.match(/from\s+['"]([^'"]+)['"]/);
+    if (imp) imports.push(imp[1]);
+    if (/^export\s+(default\s+)?(function|const|class|async)/.test(line.trim()) ||
+        /^export\s*\{/.test(line.trim())) {
+      exports.push(line.trim().slice(0, 120));
+    }
+    const fig = line.match(/https?:\/\/(?:www\.)?figma\.com\/[^\s)'"]+/gi);
+    if (fig) figma.push(...fig);
+  }
+  return { lines, imports: [...new Set(imports)].slice(0, 20), exports: exports.slice(0, 15), figma: [...new Set(figma)].slice(0, 5) };
+}
+
+function classifyFile(rel, base) {
+  for (const c of CLASSIFIERS) {
+    if (c.match(rel, base)) return c;
+  }
+  return null;
+}
+
+function frontmatter(meta) {
+  const lines = ['---'];
+  for (const [k, v] of Object.entries(meta)) {
+    if (Array.isArray(v)) lines.push(`${k}: [${v.map((x) => JSON.stringify(x)).join(', ')}]`);
+    else if (typeof v === 'boolean') lines.push(`${k}: ${v}`);
+    else lines.push(`${k}: ${JSON.stringify(String(v))}`);
+  }
+  lines.push('---');
+  return lines.join('\n');
+}
+
+function buildDraftNote({ title, rel, classifier, extract, related }) {
+  const date = new Date().toISOString().split('T')[0];
+  const contentHash = sha1(extract.lines.join('\n'));
+  const fm = frontmatter({
+    title,
+    tags: classifier.tags,
+    source: rel,
+    category: classifier.category,
+    collected_at: date,
+    content_hash: contentHash,
+    enriched: false,
+  });
+  const relatedBlock = related.length
+    ? related.map((r) => `- [[${r}]]`).join('\n')
+    : '- _(none yet)_';
+  const body = [
+    '## Purpose',
+    `Draft extract from \`${rel}\`. Awaiting Kavi enrichment.`,
+    '',
+    '## Raw extract',
+    '```',
+    extract.lines.join('\n').slice(0, 3000),
+    '```',
+    '',
+    '## Exports',
+    extract.exports.length ? extract.exports.map((e) => `- \`${e}\``).join('\n') : '- _(none detected)_',
+    '',
+    '## Imports',
+    extract.imports.length ? extract.imports.map((i) => `- \`${i}\``).join('\n') : '- _(none detected)_',
+    '',
+    ...(extract.figma.length ? ['## Figma links', ...extract.figma.map((u) => `- ${u}`), ''] : []),
+    '## Related',
+    relatedBlock,
+    '',
+  ].join('\n');
+  return { fm, body, contentHash };
+}
+
+// ─── Scan → draft vault ──────────────────────────────────────────────────────
+
+function scanAndDraft({ cwd, vaultPath, dryRun }) {
+  const gitPatterns = loadGitignore(cwd);
+  const files = walkFiles(cwd, gitPatterns);
+  const notes = []; // { vaultRel, fullOut, title, classifier, contentHash, draft }
+
+  // First pass: classify + extract
+  const candidates = [];
+  for (const f of files) {
+    const classifier = classifyFile(f.rel, f.base);
+    if (!classifier) continue;
+    const extract = extractSnippet(f.full);
+    const title = slugify(f.base);
+    candidates.push({ ...f, classifier, extract, title });
+  }
+
+  // Build simple related links: Components ↔ Pages by shared basename / import
+  const pageTitles = candidates.filter((c) => c.classifier.folder === 'Pages').map((c) => `Pages/${c.title}`);
+  const componentTitles = candidates.filter((c) => c.classifier.folder === 'Components').map((c) => `Components/${c.title}`);
+
+  for (const c of candidates) {
+    let related = [];
+    if (c.classifier.folder === 'Components') {
+      related = pageTitles.filter((p) => {
+        const pageName = p.split('/')[1].toLowerCase();
+        return c.extract.imports.some((i) => i.toLowerCase().includes(c.title.toLowerCase())) ||
+          pageName.includes(c.title.toLowerCase().slice(0, 6));
+      }).slice(0, 5);
+    } else if (c.classifier.folder === 'Pages') {
+      related = componentTitles.filter((comp) => {
+        const name = comp.split('/')[1].toLowerCase();
+        return c.extract.imports.some((i) => i.toLowerCase().includes(name));
+      }).slice(0, 8);
+    }
+
+    const { fm, body, contentHash } = buildDraftNote({
+      title: c.title,
+      rel: c.rel,
+      classifier: c.classifier,
+      extract: c.extract,
+      related,
+    });
+
+    // Disambiguate colliding titles within a folder
+    let vaultRel = path.join(c.classifier.folder, `${c.title}.md`);
+    let n = 2;
+    while (notes.some((x) => x.vaultRel === vaultRel)) {
+      vaultRel = path.join(c.classifier.folder, `${c.title}-${n}.md`);
+      n += 1;
+    }
+
+    notes.push({
+      vaultRel,
+      fullOut: path.join(vaultPath, vaultRel),
+      title: c.title,
+      classifier: c.classifier,
+      contentHash,
+      draft: `${fm}\n\n${body}`,
+      sourceRel: c.rel,
+    });
+  }
+
+  if (!dryRun) {
+    fs.mkdirSync(vaultPath, { recursive: true });
+    for (const note of notes) {
+      // Hash-based skip: if existing note has same content_hash, keep enriched body
+      if (fs.existsSync(note.fullOut)) {
+        try {
+          const existing = fs.readFileSync(note.fullOut, 'utf8');
+          const m = existing.match(/content_hash:\s*"?([a-f0-9]+)"?/);
+          const enriched = /enriched:\s*true/.test(existing);
+          if (m && m[1] === note.contentHash && enriched) {
+            note.skipped = true;
+            note.draft = existing; // keep previous enriched version
+            continue;
+          }
+        } catch { /* rewrite */ }
+      }
+      fs.mkdirSync(path.dirname(note.fullOut), { recursive: true });
+      fs.writeFileSync(note.fullOut, note.draft);
+    }
+    writeMeta(vaultPath, notes, cwd);
+  }
+
+  return {
+    filesScanned: files.length,
+    notesWritten: notes.filter((n) => !n.skipped).length,
+    notesSkipped: notes.filter((n) => n.skipped).length,
+    notes,
+  };
+}
+
+function writeMeta(vaultPath, notes, cwd) {
+  const byFolder = {};
+  for (const n of notes) {
+    const folder = n.classifier.folder;
+    if (!byFolder[folder]) byFolder[folder] = [];
+    byFolder[folder].push(n);
+  }
+
+  const indexLines = [
+    '---',
+    'title: "Knowledge Vault Index"',
+    'tags: [meta, moc]',
+    `collected_at: "${new Date().toISOString().split('T')[0]}"`,
+    '---',
+    '',
+    '# Knowledge Vault Index',
+    '',
+    `Source repo: \`${cwd}\``,
+    '',
+    '## Map of content',
+    '',
+  ];
+  for (const [folder, list] of Object.entries(byFolder).sort()) {
+    indexLines.push(`### ${folder}`, '');
+    for (const n of list) {
+      const link = n.vaultRel.replace(/\.md$/, '').replace(/\\/g, '/');
+      indexLines.push(`- [[${link}|${n.title}]] — \`${n.sourceRel}\``);
+    }
+    indexLines.push('');
+  }
+  fs.mkdirSync(path.join(vaultPath, '_meta'), { recursive: true });
+  fs.writeFileSync(path.join(vaultPath, '_meta', 'index.md'), indexLines.join('\n'));
+  fs.writeFileSync(path.join(vaultPath, '_meta', 'last-sync.md'), [
+    '---',
+    'title: "Last sync"',
+    '---',
+    '',
+    `# Last sync`,
+    '',
+    `- **When:** ${new Date().toISOString()}`,
+    `- **Repo:** ${cwd}`,
+    `- **Notes:** ${notes.length}`,
+    '',
+  ].join('\n'));
+}
+
+// ─── LLM enrichment ──────────────────────────────────────────────────────────
+
+function loadKaviCard() {
+  const p = path.join(PACKAGE_ROOT, 'agents', 'cards', 'kavi.md');
+  if (fs.existsSync(p)) return fs.readFileSync(p, 'utf8');
+  return 'You are Kavi, the Knowledge Archivist. Rewrite draft extracts into concise Obsidian notes. Do not invent files or APIs.';
+}
+
+function hasAnyApiKey() {
+  return !!(
+    process.env.ANTHROPIC_API_KEY ||
+    process.env.OPENAI_API_KEY ||
+    process.env.GEMINI_API_KEY ||
+    process.env.GOOGLE_API_KEY ||
+    process.env.ZAI_API_KEY ||
+    process.env.ZHIPU_API_KEY
+  );
+}
+
+function pickEnrichProvider(config) {
+  // Prefer cheap structured tier: openai mini → google flash → anthropic → zai
+  if (process.env.OPENAI_API_KEY) return { provider: 'openai', model: config.orchestrator?.tiers?.structured?.model || 'gpt-4o-mini' };
+  if (process.env.GEMINI_API_KEY || process.env.GOOGLE_API_KEY) return { provider: 'google', model: 'gemini-2.5-flash' };
+  if (process.env.ZAI_API_KEY || process.env.ZHIPU_API_KEY) return { provider: 'zai', model: 'glm-4.5-flash' };
+  if (process.env.ANTHROPIC_API_KEY) return { provider: 'anthropic', model: 'claude-haiku-4-5-20251001' };
+  return null;
+}
+
+function parseEnrichedBatch(text, count) {
+  // Expect notes separated by ---NOTE N--- markers; fall back to whole text for single note
+  const parts = text.split(/---NOTE\s+\d+---/i).map((s) => s.trim()).filter(Boolean);
+  if (parts.length >= count) return parts.slice(0, count);
+  if (parts.length === 1 && count === 1) return parts;
+  // Uneven response — pad with nulls so we keep drafts for missing ones
+  while (parts.length < count) parts.push(null);
+  return parts.slice(0, count);
+}
+
+async function enrichNotes({ notes, limit, provider, model }) {
+  const { callLlm, loadConfig } = require('./orchestrator/run');
+  const config = loadConfig();
+  const chosen = provider && model
+    ? { provider, model }
+    : pickEnrichProvider(config);
+  if (!chosen) {
+    return { enriched: 0, skipped: notes.length, reason: 'No API key set — wrote draft-only vault.', llm_calls: 0, input_tokens_est: 0, output_tokens_est: 0 };
+  }
+
+  const toEnrich = notes.filter((n) => !n.skipped).slice(0, limit == null ? notes.length : limit);
+  const system = [
+    loadKaviCard(),
+    '',
+    'For each draft note below, rewrite ONLY the body (keep the YAML frontmatter I give you).',
+    'Output format: for each note, emit exactly:',
+    '---NOTE 1---',
+    '<full markdown including the original frontmatter with enriched: true>',
+    '---NOTE 2---',
+    '...',
+    'Rules: purpose + key facts + Related wikilinks. Do NOT invent files, APIs, or components not in the Raw extract.',
+  ].join('\n');
+
+  let enriched = 0;
+  let llmCalls = 0;
+  let inTok = 0;
+  let outTok = 0;
+
+  for (let i = 0; i < toEnrich.length; i += BATCH_SIZE) {
+    const batch = toEnrich.slice(i, i + BATCH_SIZE);
+    const user = batch.map((n, idx) => `---NOTE ${idx + 1}---\n${n.draft}`).join('\n\n');
+    const text = await callLlm({
+      provider: chosen.provider,
+      model: chosen.model,
+      system,
+      user,
+      maxTokens: 4000,
+    });
+    llmCalls += 1;
+    inTok += Math.ceil((system.length + user.length) / 4);
+    outTok += Math.ceil(text.length / 4);
+
+    const parts = parseEnrichedBatch(text, batch.length);
+    for (let j = 0; j < batch.length; j++) {
+      let body = parts[j];
+      if (!body) continue;
+      // Ensure enriched: true in frontmatter
+      if (!/enriched:\s*true/.test(body)) {
+        body = body.replace(/enriched:\s*false/, 'enriched: true');
+        if (!/enriched:/.test(body) && body.startsWith('---')) {
+          body = body.replace(/^---\n/, '---\nenriched: true\n');
+        }
+      }
+      // Preserve content_hash from draft
+      if (!/content_hash:/.test(body) && batch[j].contentHash) {
+        body = body.replace(/^---\n/, `---\ncontent_hash: "${batch[j].contentHash}"\n`);
+      }
+      fs.writeFileSync(batch[j].fullOut, body);
+      batch[j].draft = body;
+      enriched += 1;
+    }
+  }
+
+  return {
+    enriched,
+    skipped: notes.length - enriched,
+    reason: null,
+    llm_calls: llmCalls,
+    input_tokens_est: inTok,
+    output_tokens_est: outTok,
+    provider: chosen.provider,
+    model: chosen.model,
+  };
+}
+
+// ─── Public API ──────────────────────────────────────────────────────────────
+
+/**
+ * @param {{
+ *   cwd?: string,
+ *   vault?: string,
+ *   project?: string,
+ *   dryRun?: boolean,
+ *   enrich?: boolean,
+ *   limit?: number,
+ *   target?: string,
+ *   provider?: string,
+ *   model?: string,
+ *   web?: boolean,
+ *   webLimit?: number,
+ * }} opts
+ */
+async function collect(opts = {}) {
+  const cwd = path.resolve(opts.cwd || process.cwd());
+  const projectId = opts.project || session.getProjectId(cwd);
+  const vaultPath = path.resolve(opts.vault || path.join(os.homedir(), '.analyzthis_design', 'vaults', projectId));
+  const enrich = opts.enrich !== false;
+  const dryRun = !!opts.dryRun;
+  const limit = opts.limit != null ? opts.limit : null;
+  const fetchWeb = opts.web !== false;
+  const discoverSources = opts.discoverSources !== false;
+  const config = loadCollectConfig();
+  const { resolveTargets } = require('./platforms');
+  const targets = resolveTargets(opts.target || 'cursor') || ['cursor'];
+
+  // Ensure session exists
+  if (!session.show({ project: projectId })) {
+    session.init({ project: projectId });
+  }
+
+  const modeParts = [];
+  if (dryRun) modeParts.push('dry-run');
+  else {
+    if (discoverSources) modeParts.push('discover');
+    if (fetchWeb) modeParts.push('web');
+    if (enrich) modeParts.push('enrich');
+    modeParts.push('sync');
+  }
+
+  console.log(`\n📚 Kavi — Collect Knowledge`);
+  console.log(`   Repo:  ${cwd}`);
+  console.log(`   Vault: ${vaultPath}`);
+  console.log(`   Mode:  ${modeParts.join(' + ') || 'scan'}\n`);
+
+  const scan = scanAndDraft({ cwd, vaultPath, dryRun });
+
+  const sourceDiscoveries = discoverSources
+    ? discoverKnowledgeSources({ cwd, config })
+    : [];
+
+  console.log(`   Scanned ${scan.filesScanned} file(s) → ${scan.notes.length} note(s) (${scan.notesWritten} written, ${scan.notesSkipped} unchanged)`);
+
+  if (discoverSources && sourceDiscoveries.length > 0) {
+    console.log(`\n🔍 Knowledge sources: ${sourceDiscoveries.length} discovered (Obsidian vaults, wikis, referenced paths)`);
+    for (const d of sourceDiscoveries.slice(0, 6)) {
+      console.log(`   • [${d.confidence}] ${d.type}: ${d.path}`);
+    }
+    if (sourceDiscoveries.length > 6) console.log(`   … and ${sourceDiscoveries.length - 6} more`);
+  } else if (discoverSources && !dryRun) {
+    console.log(`\n🔍 Knowledge sources: none discovered (add collect.source_paths in config or link vaults in README)`);
+  }
+
+  const webPlan = fetchWeb
+    ? discoverWebUrls({ cwd, notes: scan.notes, config, webLimit: opts.webLimit })
+    : { urls: [], queries: [], limit: 0, totalDiscovered: 0 };
+
+  if (fetchWeb && (webPlan.urls.length > 0 || webPlan.queries.length > 0)) {
+    console.log(`\n🌐 Web research: ${webPlan.urls.length} URL(s)${webPlan.queries.length ? `, ${webPlan.queries.length} query stub(s)` : ''}${webPlan.totalDiscovered > webPlan.urls.length ? ` (${webPlan.totalDiscovered} found, capped at ${webPlan.limit})` : ''}`);
+  } else if (fetchWeb && !dryRun) {
+    console.log(`\n🌐 Web research: no URLs found (add collect.web_urls in ~/.analyzthis_design/config.json or links in PRDs/README)`);
+  }
+
+  if (dryRun) {
+    const byFolder = {};
+    for (const n of scan.notes) {
+      byFolder[n.classifier.folder] = (byFolder[n.classifier.folder] || 0) + 1;
+    }
+    console.log('\n── Dry run preview ──');
+    console.log(JSON.stringify({
+      vaultPath,
+      projectId,
+      byFolder,
+      knowledgeSources: sourceDiscoveries.map((d) => ({
+        path: d.path,
+        type: d.type,
+        confidence: d.confidence,
+        discovered_from: d.discovered_from,
+        markdown_files: d.stats.markdown_files,
+      })),
+      webUrls: webPlan.urls.map((u) => ({ url: u.url, source: u.source })),
+      webQueries: webPlan.queries,
+      sample: scan.notes.slice(0, 8).map((n) => n.vaultRel),
+    }, null, 2));
+    console.log('');
+    return { vaultPath, projectId, dryRun: true, ...scan, sources: sourceDiscoveries, web: webPlan, enrich: null, sync: null };
+  }
+
+  let sourceManifest = { written: [] };
+  if (discoverSources && sourceDiscoveries.length > 0) {
+    console.log(`\n⏳ Writing source manifest notes (Sources/*.md)...`);
+    sourceManifest = writeSourceManifestNotes({
+      discoveries: sourceDiscoveries,
+      vaultPath,
+      dryRun: false,
+    });
+    console.log(`   Wrote ${sourceManifest.written.length} manifest note(s) + _meta/knowledge-sources.md`);
+  }
+
+  let webResult = { fetched: [], failed: [], queries: [] };
+  if (fetchWeb && (webPlan.urls.length > 0 || webPlan.queries.length > 0)) {
+    console.log(`\n⏳ Fetching web research into web-context.md...`);
+    webResult = await collectWebResearch({
+      urls: webPlan.urls,
+      queries: webPlan.queries,
+      projectId,
+      dryRun: false,
+    });
+    console.log(`   Fetched ${webResult.fetched.length} URL(s)${webResult.failed.length ? `, ${webResult.failed.length} failed` : ''}`);
+  }
+
+  let enrichResult = { enriched: 0, skipped: scan.notes.length, reason: 'enrichment disabled', llm_calls: 0, input_tokens_est: 0, output_tokens_est: 0 };
+  if (enrich) {
+    if (!hasAnyApiKey()) {
+      console.log(`\n  ⚠  No LLM API key found — writing draft-only vault. Set OPENAI_API_KEY / GEMINI_API_KEY / ANTHROPIC_API_KEY / ZAI_API_KEY to enrich.`);
+      enrichResult.reason = 'No API key set — wrote draft-only vault.';
+    } else {
+      console.log(`\n⏳ Enriching notes with Kavi (LLM)...`);
+      enrichResult = await enrichNotes({
+        notes: scan.notes,
+        limit,
+        provider: opts.provider,
+        model: opts.model,
+      });
+      console.log(`   Enriched ${enrichResult.enriched} note(s)${enrichResult.reason ? ` (${enrichResult.reason})` : ''}`);
+    }
+  }
+
+  // Auto-connect discovered sources + Kavi vault, then sync knowledge bank
+  console.log(`\n⏳ Wiring knowledge bank...`);
+  const connectResult = discoverSources
+    ? connectDiscoveredSources({
+      discoveries: sourceDiscoveries,
+      kaviVaultPath: vaultPath,
+      config,
+      autoConnect: config.collect?.auto_connect_discovered !== false,
+    })
+    : { connected: [], skipped: [] };
+  if (connectResult.connected.length > 0) {
+    console.log(`   Auto-connected ${connectResult.connected.length} source(s):`);
+    for (const p of connectResult.connected) console.log(`   • ${p}`);
+  }
+  knowledge.connect({ vaultPath, tags: [], include: [] });
+  const syncResult = knowledge.sync({ targets });
+  cache.invalidatePrefix('kb:');
+
+  if (syncResult.message) {
+    console.log(`  ⚠  ${syncResult.message}`);
+  } else {
+    console.log(`✅ Synced ${syncResult.synced} note(s) into the knowledge bank.`);
+    for (const t of syncResult.copiedTo || []) console.log(`   • ${t}`);
+  }
+
+  // Persist session fields
+  const state = session.show({ project: projectId }) || {};
+  const sources = Array.isArray(state.content_sources) ? state.content_sources : [];
+  if (!sources.includes('codebase_collect')) sources.push('codebase_collect');
+  session.update({
+    project: projectId,
+    patch: {
+      content_sources: sources,
+      vault_path: vaultPath,
+      last_collect_at: new Date().toISOString(),
+      digest: {
+        ...(state.digest || {}),
+        task_map_summary: `Collected knowledge vault (${scan.notes.length} notes) from ${cwd}`,
+        experts: ['kavi'],
+      },
+      metrics: {
+        ...(state.metrics || {}),
+        llm_calls: enrichResult.llm_calls || 0,
+        experts_run: ['kavi'],
+        input_tokens_est: enrichResult.input_tokens_est || 0,
+        output_tokens_est: enrichResult.output_tokens_est || 0,
+        mode: 'collect',
+        effort_log: enrichResult.model
+          ? [{ persona: 'kavi', effort: 'standard', model: enrichResult.model, scoped_mode: null }]
+          : [],
+        cost_usd: state.metrics?.cost_usd || 0,
+      },
+    },
+  });
+
+  console.log(`\n✅ Collect complete.`);
+  console.log(`   Vault: ${vaultPath}`);
+  console.log(`   Open in Obsidian, or run /persona-orchestrator — personas now read this knowledge bank.\n`);
+
+  return {
+    vaultPath,
+    projectId,
+    dryRun: false,
+    filesScanned: scan.filesScanned,
+    notes: scan.notes.length,
+    notesWritten: scan.notesWritten,
+    notesSkipped: scan.notesSkipped,
+    web: webResult,
+    sources: {
+      discovered: sourceDiscoveries,
+      manifest: sourceManifest,
+      connected: connectResult.connected,
+      skipped: connectResult.skipped,
+    },
+    enrich: enrichResult,
+    sync: syncResult,
+  };
+}
+
+module.exports = {
+  collect,
+  scanAndDraft,
+  classifyFile,
+  walkFiles,
+  discoverWebUrls,
+  collectWebResearch,
+  extractUrlsFromText,
+  isResearchableUrl,
+  HARD_EXCLUDES,
+};