/* store.jsx — cellar data model, persistence, status logic, AI identify */

const THIS_YEAR = 2026;
const LS_KEY = "cellar_wines_v2";

// ---------- backup ----------
const AUTO_BACKUP_LS = "cellar_auto_bk";
let _autoBackup = (()=>{ try{ return localStorage.getItem(AUTO_BACKUP_LS)!=="false"; }catch(e){ return true; } })();
let _bkTimer = null;
function _dl(blob,name){ const url=URL.createObjectURL(blob); const a=Object.assign(document.createElement("a"),{href:url,download:name}); document.body.appendChild(a); a.click(); document.body.removeChild(a); setTimeout(()=>URL.revokeObjectURL(url),2000); }

// ---------- persistence ----------
let _wines = null;
const listeners = new Set();
function uid(){ return "w_" + Math.random().toString(36).slice(2,9) + Date.now().toString(36).slice(-3); }

function load(){
  if(_wines) return _wines;
  try{
    const raw = localStorage.getItem(LS_KEY);
    if(raw){ _wines = JSON.parse(raw); }
  }catch(e){}
  if(!_wines){
    _wines = [];
    persist();
  }
  return _wines;
}
function persist(){ try{ localStorage.setItem(LS_KEY, JSON.stringify(_wines)); }catch(e){} listeners.forEach(fn=>fn()); }

const Cellar = {
  all(){ return load(); },
  get(id){ return load().find(w=>w.id===id); },
  add(w){ const item = { id:uid(), addedAt:Date.now(), qty:1, photo:null, ...w }; _wines=[item, ...load()]; persist(); Cellar.scheduleAutoBackup(); return item; },
  // Find an existing wine that matches producer+cuvée+vintage+color (case-insensitive)
  findDuplicate(w){
    const norm = s => (s||"").trim().toLowerCase();
    return load().find(e =>
      norm(e.producer)===norm(w.producer) &&
      norm(e.cuvee)===norm(w.cuvee) &&
      e.vintage==w.vintage &&
      norm(e.color)===norm(w.color)
    );
  },
  // Add or merge: if exact bottle exists, increment qty and return existing; otherwise add new
  addOrMerge(w, addQty=1){
    const existing = Cellar.findDuplicate(w);
    if(existing){
      Cellar.setQty(existing.id, (existing.qty||1) + addQty);
      return { ...Cellar.get(existing.id), merged:true };
    }
    return { ...Cellar.add({ ...w, qty:addQty }), merged:false };
  },
  update(id, patch){ _wines = load().map(w=> w.id===id ? {...w, ...patch} : w); persist(); },
  setQty(id, qty){ qty=Math.max(0,qty); if(qty===0){ Cellar.remove(id); return; } Cellar.update(id,{qty}); },
  remove(id){ _wines = load().filter(w=>w.id!==id); persist(); },
  openCoravin(id){ Cellar.update(id, { coravin:{ openedAt: Date.now() } }); },
  reseal(id){ Cellar.update(id, { coravin:null }); },
  finishBottle(id){ const w=Cellar.get(id); if(!w) return; const qty=(w.qty||1)-1; if(qty<=0){ Cellar.remove(id); } else { Cellar.update(id, { qty, coravin:null }); } },
  subscribe(fn){ listeners.add(fn); return ()=>listeners.delete(fn); },
  exportBackup(){
    const date=new Date().toISOString().slice(0,10);
    const payload={version:2,exportedAt:new Date().toISOString(),count:load().length,wines:load()};
    _dl(new Blob([JSON.stringify(payload,null,2)],{type:"application/json"}),"cochez-cellar-"+date+".json");
  },
  importBackup(file){
    return new Promise((resolve,reject)=>{
      const r=new FileReader();
      r.onload=e=>{
        try{
          const data=JSON.parse(e.target.result);
          const wines=Array.isArray(data)?data:(Array.isArray(data.wines)?data.wines:null);
          if(!wines) throw new Error("No wine data found in this file.");
          _wines=wines; persist(); resolve(wines.length);
        }catch(err){ reject(err); }
      };
      r.onerror=()=>reject(new Error("Could not read file."));
      r.readAsText(file);
    });
  },
  getAutoBackup(){ return _autoBackup; },
  setAutoBackup(on){ _autoBackup=!!on; try{ localStorage.setItem(AUTO_BACKUP_LS,String(_autoBackup)); }catch(e){} },
  scheduleAutoBackup(){
    if(!_autoBackup) return;
    clearTimeout(_bkTimer);
    _bkTimer=setTimeout(()=>Cellar.exportBackup(),3000);
  },
};

// React hook
function useCellar(){
  const [, force] = React.useReducer(x=>x+1, 0);
  React.useEffect(()=> Cellar.subscribe(force), []);
  return Cellar.all();
}

// ---------- status logic ----------
function statusOf(w, year=THIS_YEAR){
  if(w.drinkTo && year > w.drinkTo) return "past";
  if(w.peakFrom && w.peakTo && year >= w.peakFrom && year <= w.peakTo) return "now";
  if(w.peakFrom && year < w.peakFrom){
    return (w.drinkFrom && year >= w.drinkFrom) ? "soon" : "hold";
  }
  if(w.peakTo && year > w.peakTo) return "soon"; // past peak but drinkable → drink up
  return "hold";
}
const STATUS_LABEL = { now:"Drink now", soon:"Drink soon", hold:"Hold", past:"Past peak" };
function statusLabel(w){
  const s = statusOf(w);
  if(s==="hold" && w.peakFrom) return `Hold → ${w.peakFrom}`;
  return STATUS_LABEL[s];
}
function windowText(w){
  if(!w.drinkFrom) return "—";
  return `${w.drinkFrom}–${w.drinkTo}` + (w.peakFrom?`  ·  peak ${w.peakFrom}–${w.peakTo}`:"");
}
// position 0..1 of "now" across the drink window, + peak band
function meterGeom(w){
  const a=w.drinkFrom, b=w.drinkTo; if(!a||!b||b<=a) return null;
  const clamp=v=>Math.max(0,Math.min(1,v));
  return {
    now: clamp((THIS_YEAR-a)/(b-a)),
    peakL: clamp((w.peakFrom-a)/(b-a)),
    peakR: clamp((w.peakTo-a)/(b-a)),
  };
}
const fmt$ = n => n==null ? "—" : "$" + Math.round(n).toLocaleString();
const COLOR_HEX = { Red:"#7a2233", White:"#c8a93f", "Rosé":"#d98a8f", Rose:"#d98a8f", Sparkling:"#b88a2a" };

// ---------- professional critic ratings ----------
const RATING_SOURCES = { RP:"Robert Parker", JS:"James Suckling", WS:"Wine Spectator", AG:"Vinous", JR:"Jancis Robinson", D:"Decanter", WE:"Wine Enthusiast" };
function ratingsList(w){
  if(!w||!w.ratings) return [];
  return Object.entries(w.ratings).filter(([k,v])=>v!=null&&v>0).sort((a,b)=>b[1]-a[1]);
}

// ---------- Coravin preservation (3-year window) ----------
const CORAVIN_DAYS = 365*3;
function coravinInfo(w){
  if(!w || !w.coravin || !w.coravin.openedAt) return null;
  const total = CORAVIN_DAYS*86400000;
  const left = (w.coravin.openedAt + total) - Date.now();
  return { total, leftMs:left, leftDays:Math.ceil(left/86400000), frac:Math.max(0,Math.min(1,left/total)), expired:left<=0, drinkBy:w.coravin.openedAt+total, openedAt:w.coravin.openedAt };
}
function coravinText(w){
  const c = coravinInfo(w); if(!c) return null;
  if(c.expired) return "Window passed — drink now";
  const d = c.leftDays;
  if(d>=365){ const y=Math.floor(d/365); const mo=Math.round((d%365)/30.44); return mo ? (y+"y "+mo+"mo left") : (y+"y left"); }
  if(d>=60) return Math.round(d/30.44)+" months left";
  if(d>=1) return d+(d===1?" day left":" days left");
  return "Drink today";
}

// ---------- AI identify from OCR text ----------
async function identifyFromText(ocrText){
  const prompt = `You are a master sommelier and wine database. Text was OCR-scanned from a wine bottle label (it may be messy, partial, or contain noise). Identify the wine and return ONLY a JSON object (no markdown, no prose) with EXACTLY these keys:
{"producer":string,"cuvee":string,"vintage":number,"color":"Red"|"White"|"Rosé"|"Sparkling","country":string,"region":string,"subregion":string,"appellation":string,"classification":string,"varietal":string,"abv":number,"critScore":number,"drinkFrom":number,"drinkTo":number,"peakFrom":number,"peakTo":number,"valueLow":number,"valueHigh":number,"valueEst":number,"pairings":string[],"tasting":string,"body":number,"tannin":number,"acidity":number,"sweetness":number,"ratings":{"RP":number|null,"JS":number|null,"WS":number|null,"AG":number|null,"JR":number|null,"D":number|null,"WE":number|null},"confidence":number}
Rules: classification = the meaningful tier for that region (e.g. Rioja: Crianza/Reserva/Gran Reserva; Bordeaux: Cru Bourgeois/Grand Cru; else "Estate" or appellation tier). body/tannin/acidity/sweetness are 0-100. Estimate realistic current market value in USD. drink window and peak are 4-digit years. pairings = 3-5 specific foods. tasting = one vivid sentence. ratings = known professional critic scores for this specific wine+vintage (RP=Robert Parker, JS=James Suckling, WS=Wine Spectator, AG=Vinous, JR=Jancis Robinson converted to 100-pt, D=Decanter, WE=Wine Enthusiast); use null for unknown. critScore = the single highest score. confidence 0-100 = how sure you are of the identification. If a field is unknown, make your best expert estimate. Today is ${THIS_YEAR}.

OCR TEXT:
"""${ocrText.slice(0,1200)}"""`;
  const raw = await window.claude.complete(prompt);
  let obj;
  try{
    const m = raw.match(/\{[\s\S]*\}/);
    obj = JSON.parse(m ? m[0] : raw);
  }catch(e){
    throw new Error("Could not read that label. Try again or enter details manually.");
  }
  // coerce numbers
  ["vintage","abv","critScore","drinkFrom","drinkTo","peakFrom","peakTo","valueLow","valueHigh","valueEst","body","tannin","acidity","sweetness","confidence"]
    .forEach(k=>{ if(obj[k]!=null) obj[k]=Number(obj[k]); });
  if(!Array.isArray(obj.pairings)) obj.pairings = obj.pairings? [String(obj.pairings)] : [];
  return obj;
}

// ---------- AI identify directly from the label photo (vision) ----------
// Far more reliable than OCR for real bottles — gold-on-black, embossed,
// curved, or decorative labels that defeat text recognition.
const WINE_SCHEMA = `{"producer":string,"cuvee":string,"vintage":number,"color":"Red"|"White"|"Rosé"|"Sparkling","country":string,"region":string,"subregion":string,"appellation":string,"classification":string,"varietal":string,"abv":number,"critScore":number,"drinkFrom":number,"drinkTo":number,"peakFrom":number,"peakTo":number,"valueLow":number,"valueHigh":number,"valueEst":number,"pairings":string[],"tasting":string,"body":number,"tannin":number,"acidity":number,"sweetness":number,"ratings":{"RP":number|null,"JS":number|null,"WS":number|null,"AG":number|null,"JR":number|null,"D":number|null,"WE":number|null},"labelText":string,"confidence":number}`;

function parseWine(raw){
  let obj;
  try{
    const m = raw.match(/\{[\s\S]*\}/);
    obj = JSON.parse(m ? m[0] : raw);
  }catch(e){
    throw new Error("Could not read that label. Try again or enter details manually.");
  }
  ["vintage","abv","critScore","drinkFrom","drinkTo","peakFrom","peakTo","valueLow","valueHigh","valueEst","body","tannin","acidity","sweetness","confidence"]
    .forEach(k=>{ if(obj[k]!=null) obj[k]=Number(obj[k]); });
  if(obj.confidence!=null && obj.confidence<=1) obj.confidence = Math.round(obj.confidence*100); // accept 0-1 or 0-100
  if(!Array.isArray(obj.pairings)) obj.pairings = obj.pairings? [String(obj.pairings)] : [];
  // normalize ratings
  if(obj.ratings && typeof obj.ratings==="object"){
    Object.keys(obj.ratings).forEach(k=>{ if(obj.ratings[k]!=null){ obj.ratings[k]=Number(obj.ratings[k]); if(isNaN(obj.ratings[k])) obj.ratings[k]=null; } });
  } else { obj.ratings = {}; }
  // set critScore from highest rating if not already set
  if(!obj.critScore){ const vals=Object.values(obj.ratings||{}).filter(v=>v!=null&&v>0); if(vals.length) obj.critScore=Math.max(...vals); }
  return obj;
}

async function identifyFromImage(imageDataUrl){
  if(!window.claude || !window.claude.complete){
    throw new Error("The wine-ID service isn’t connected here. Scanning needs the Claude AI bridge, which is available in the Anthropic preview — a plain static host (e.g. Netlify) needs a small API proxy first.");
  }
  const m = /^data:(image\/[a-z0-9.+-]+);base64,(.*)$/i.exec(imageDataUrl) || [];
  const media_type = m[1] || "image/jpeg";
  const data = m[2] || (imageDataUrl.split(",")[1] || imageDataUrl);
  const instruction = `You are a master sommelier with an encyclopedic wine database. Study this photo of a wine bottle label and identify the wine. Read every legible word — including small print, foil/gold lettering, and embossed or low-contrast text. Use producer logos and visual cues. Then return ONLY a JSON object (no markdown, no prose) with EXACTLY these keys:
${WINE_SCHEMA}
Rules: classification = the meaningful tier for that region (e.g. Rioja: Crianza/Reserva/Gran Reserva; Bordeaux: Cru Bourgeois/Grand Cru; else "Estate" or appellation tier). body/tannin/acidity/sweetness are 0-100. Estimate realistic current market value in USD. drink window and peak are 4-digit years. pairings = 3-5 specific foods. tasting = one vivid sentence. labelText = the exact words you can read on the label, comma-separated. ratings = known professional critic scores for this specific wine+vintage (RP=Robert Parker/Wine Advocate, JS=James Suckling, WS=Wine Spectator, AG=Antonio Galloni/Vinous, JR=Jancis Robinson converted to 100-pt, D=Decanter, WE=Wine Enthusiast); use null for any you don't know — only include scores you're confident about. critScore = the single highest/most authoritative score. confidence 0-100 = how sure you are of the identification. If the producer or wine is unclear, still give your best expert guess and lower the confidence. If a field is unknown, make your best expert estimate. Today is ${THIS_YEAR}.`;
  const raw = await window.claude.complete({ messages:[{ role:"user", content:[
    { type:"image", source:{ type:"base64", media_type, data } },
    { type:"text", text:instruction }
  ]}]});
  return parseWine(raw);
}

// Refresh just the ratings for a wine already in the cellar (text-only, no image)
async function refreshRatings(id){
  if(!window.claude || !window.claude.complete) throw new Error("AI service not connected.");
  const w = Cellar.get(id);
  if(!w) throw new Error("Wine not found.");
  const prompt = `You are a master sommelier. Return ONLY a JSON object (no markdown) with professional critic scores for this specific wine:
Producer: ${w.producer}
Cuvée: ${w.cuvee||""}
Vintage: ${w.vintage||"NV"}
Varietal: ${w.varietal||""}
Region: ${w.region||""}, ${w.country||""}
Classification: ${w.classification||""}

Keys: {"RP":number|null,"JS":number|null,"WS":number|null,"AG":number|null,"JR":number|null,"D":number|null,"WE":number|null,"critScore":number}
RP=Robert Parker, JS=James Suckling, WS=Wine Spectator, AG=Vinous, JR=Jancis Robinson (100-pt), D=Decanter, WE=Wine Enthusiast. Use null for scores you don't know. critScore = the highest. Only include scores you're confident are real published scores for this exact wine+vintage.`;
  const raw = await window.claude.complete(prompt);
  let obj;
  try{ const m=raw.match(/\{[\s\S]*\}/); obj=JSON.parse(m?m[0]:raw); }catch(e){ throw new Error("Couldn't parse ratings."); }
  const ratings = {};
  Object.keys(RATING_SOURCES).forEach(k=>{ if(obj[k]!=null) ratings[k]=Number(obj[k]); });
  const critScore = obj.critScore ? Number(obj.critScore) : Math.max(0,...Object.values(ratings).filter(v=>v!=null&&v>0));
  Cellar.update(id, { ratings, critScore: critScore||w.critScore });
  return ratings;
}

Object.assign(window, { Cellar, useCellar, statusOf, statusLabel, STATUS_LABEL, windowText, meterGeom, fmt$, COLOR_HEX, RATING_SOURCES, ratingsList, refreshRatings, identifyFromText, identifyFromImage, coravinInfo, coravinText, CORAVIN_DAYS, THIS_YEAR });