[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_153":154,"_155":156},"routes/_app.x.$handle.$name",{"_94":93,"_32":45,"_102":84,"_27":103,"_13":14,"_104":-5,"_11":105,"_106":107,"_108":109,"_110":111,"_112":113,"_20":-5,"_114":-5,"_115":-5,"_116":25,"_117":118,"_119":120,"_121":-5,"_122":-5,"_123":124},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":-5,"_18":-5},"project",{"_43":44,"_32":45,"_38":46,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":-5,"_53":-5,"_18":54,"_55":-5,"_56":25,"_57":-5,"_58":-5,"_59":60,"_61":62,"_63":-5,"_64":-5,"_65":-5,"_66":-5,"_67":68,"_69":68,"_70":71,"_72":71,"_73":71,"_74":68,"_75":68,"_76":77,"_78":79,"_80":81,"_82":25},"version",4,"changes",[19],"aiChatInfo","userId",{"_20":21,"_22":23,"_24":25,"_26":25,"_27":28,"_29":28},"file",{"_30":31,"_32":28,"_33":37,"_35":36,"_38":39,"_40":14,"_41":42,"_24":25},"previous",{"_30":31,"_32":28,"_33":34,"_35":36},"deleted",false,"deletedHere","path","Quinn-v3-backend.ts","latestPath","fileId","a55e920c-3689-11f1-ba83-42b51c65c3df","name","fileType","script","content","// ============================================================\n// QUINN v3 — Autonomous Quant Trading Agent\n// Val Town Backend (index.ts)\n//\n// Stack: OpenRouter (Llama 3.3 70b free) + Alpaca Markets\n// + Tavily Search + Val Town SQLite\n//\n// Hardening:\n// ✓ Model locked: meta-llama/llama-3.3-70b-instruct:free\n// ✓ Retry w/ exponential backoff (skips 400s)\n// ✓ Context trimming (last 6 messages, avoids 8k limit)\n// ✓ 1000ms scan pacing (free tier rate limit)\n// ✓ non_marginable_buying_power gating (GFV prevention)\n// ✓ Ghost trade reverse lookup on startup\n// ✓ Liquidity Sniper (10-day ADV, dual API shape)\n// ✓ Drawdown circuit breaker (%, not fixed $)\n// ✓ Fill reconciliation + PnL calculation\n// ✓ Structured post-mortems w/ signal JSON\n// ✓ Meta-Learner observer → auto-tunes after 20 trades\n// ✓ Confidence Score 0-100 (replaces boolean gates)\n// ✓ PDT day-trade counter (3/5d limit)\n// ✓ Swing mode: flags same-day open+close\n// ✓ Black-Scholes + Monte Carlo (2000 paths)\n// ✓ Friction cost modeled from real spread\n// ============================================================\n\nimport { sqlite } from \"https://esm.town/v/std/sqlite\";\nimport { email } from \"https://esm.town/v/std/email\";\n\nexport default async function (req: Request) {\n const headers = {\n \"Access-Control-Allow-Origin\": \"*\",\n \"Access-Control-Allow-Methods\": \"POST, OPTIONS\",\n \"Access-Control-Allow-Headers\": \"Content-Type\",\n };\n\n if (req.method === \"OPTIONS\") return new Response(null, { headers });\n if (req.method !== \"POST\") {\n return new Response(JSON.stringify({ status: \"QUINN v3 online\" }), {\n headers: { ...headers, \"Content-Type\": \"application/json\" },\n });\n }\n\n // ── ENV ──────────────────────────────────────────────────\n const OPENROUTER_KEY = Deno.env.get(\"OPENROUTER_KEY\");\n const TAVILY_KEY = Deno.env.get(\"TAVILY_KEY\");\n const ALPACA_KEY = Deno.env.get(\"ALPACA_KEY\");\n const ALPACA_SECRET = Deno.env.get(\"ALPACA_SECRET\");\n const ALPACA_BASE = Deno.env.get(\"ALPACA_LIVE\") === \"true\"\n ? \"https://api.alpaca.markets\"\n : \"https://paper-api.alpaca.markets\";\n const QUINN_MODEL = \"meta-llama/llama-3.3-70b-instruct:free\";\n\n if (!OPENROUTER_KEY) {\n return Response.json({ error: \"OPENROUTER_KEY missing.\" }, { headers });\n }\n\n // ── DB SETUP ─────────────────────────────────────────────\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS quinn_trades (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n symbol TEXT,\n side TEXT,\n qty REAL,\n entry_price REAL,\n exit_price REAL,\n reason TEXT,\n sentiment REAL,\n mc_prob REAL,\n friction_pct REAL,\n adv INTEGER,\n bs_price REAL,\n confidence_score INTEGER,\n status TEXT DEFAULT 'pending',\n alpaca_id TEXT,\n pnl REAL,\n pnl_vs_theoretical REAL,\n postmortem TEXT,\n postmortem_signals TEXT,\n is_day_trade INTEGER DEFAULT 0,\n created_at DATETIME DEFAULT CURRENT_TIMESTAMP,\n closed_at DATETIME\n )\n `);\n\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS quinn_watchlist (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n symbol TEXT UNIQUE,\n added_at DATETIME DEFAULT CURRENT_TIMESTAMP\n )\n `);\n\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS quinn_settings (\n key TEXT PRIMARY KEY,\n value TEXT,\n updated_at DATETIME DEFAULT CURRENT_TIMESTAMP\n )\n `);\n\n await sqlite.execute(`\n CREATE TABLE IF NOT EXISTS quinn_memory (\n id INTEGER PRIMARY KEY AUTOINCREMENT,\n key TEXT UNIQUE,\n content TEXT,\n updated_at DATETIME DEFAULT CURRENT_TIMESTAMP\n )\n `);\n\n // Seed defaults — only inserts if key doesn't exist yet\n const defaults: [string, string][] = [\n [\"max_position_usd\", \"10\"],\n [\"min_confidence_score\", \"65\"],\n [\"min_sentiment\", \"0.3\"],\n [\"max_friction_pct\", \"0.30\"],\n [\"min_adv\", \"50000\"],\n [\"autonomous_mode\", \"true\"],\n [\"circuit_breaker_pct\", \"0.15\"],\n [\"circuit_breaker_window_days\", \"3\"],\n [\"circuit_breaker_triggered\", \"false\"],\n [\"meta_learner_min_trades\", \"20\"],\n [\"max_day_trades_5d\", \"3\"],\n [\"swing_mode\", \"true\"],\n ];\n for (const [k, v] of defaults) {\n await sqlite.execute({\n sql: `INSERT OR IGNORE INTO quinn_settings (key, value) VALUES (?, ?)`,\n args: [k, v],\n });\n }\n\n await sqlite.execute(`\n INSERT OR IGNORE INTO quinn_watchlist (symbol) VALUES\n ('AAPL'),('NVDA'),('TSLA'),('MSFT'),('SPY'),('META'),('AMD'),('AMZN')\n `);\n\n // ── SETTINGS LOADER ──────────────────────────────────────\n async function loadSettings(): Promise> {\n const res = await sqlite.execute(`SELECT key, value FROM quinn_settings`);\n const cfg: Record = {};\n res.rows.forEach((r: any) => {\n cfg[r[0]] = r[1];\n });\n return cfg;\n }\n\n // ── MATH ENGINE ──────────────────────────────────────────\n function normCDF(x: number): number {\n const a = [\n 0.254829592,\n -0.284496736,\n 1.421413741,\n -1.453152027,\n 1.061405429,\n ];\n const p = 0.3275911;\n const sign = x < 0 ? -1 : 1;\n x = Math.abs(x) / Math.SQRT2;\n const t = 1 / (1 + p * x);\n const y = 1 -\n (((((a[4] * t + a[3]) * t) + a[2]) * t + a[1]) * t + a[0]) * t *\n Math.exp(-x * x);\n return 0.5 * (1 + sign * y);\n }\n\n function normPDF(x: number): number {\n return Math.exp(-0.5 * x * x) / Math.sqrt(2 * Math.PI);\n }\n\n function bsCalc(\n type: \"call\" | \"put\",\n S: number,\n K: number,\n T: number,\n r: number,\n sig: number,\n q = 0,\n ) {\n if (T < 1e-8) {\n return {\n price: type === \"call\" ? Math.max(S - K, 0) : Math.max(K - S, 0),\n delta: 0,\n gamma: 0,\n vega: 0,\n theta: 0,\n };\n }\n const sqT = Math.sqrt(T), eqT = Math.exp(-q * T), erT = Math.exp(-r * T);\n const d1 = (Math.log(S / K) + (r - q + 0.5 * sig * sig) * T) / (sig * sqT);\n const d2 = d1 - sig * sqT;\n let price: number, delta: number;\n if (type === \"call\") {\n price = S * eqT * normCDF(d1) - K * erT * normCDF(d2);\n delta = eqT * normCDF(d1);\n } else {\n price = K * erT * normCDF(-d2) - S * eqT * normCDF(-d1);\n delta = -eqT * normCDF(-d1);\n }\n const gamma = eqT * normPDF(d1) / (S * sig * sqT);\n const vega = S * eqT * normPDF(d1) * sqT / 100;\n const theta = type === \"call\"\n ? (-S * normPDF(d1) * sig * eqT / (2 * sqT) - r * K * erT * normCDF(d2) +\n q * S * eqT * normCDF(d1)) / 365\n : (-S * normPDF(d1) * sig * eqT / (2 * sqT) + r * K * erT * normCDF(-d2) -\n q * S * eqT * normCDF(-d1)) / 365;\n return { price, delta, gamma, vega, theta };\n }\n\n function monteCarloEdge(\n S: number,\n K: number,\n T: number,\n r: number,\n sig: number,\n paths = 2000,\n ) {\n let callSum = 0, putSum = 0, profitCount = 0;\n for (let i = 0; i < paths; i++) {\n const u1 = Math.max(Math.random(), 1e-10), u2 = Math.random();\n const z = Math.sqrt(-2 * Math.log(u1)) * Math.cos(2 * Math.PI * u2);\n const ST = S *\n Math.exp((r - 0.5 * sig * sig) * T + sig * Math.sqrt(T) * z);\n callSum += Math.max(ST - K, 0);\n putSum += Math.max(K - ST, 0);\n if (ST > S * 1.01) profitCount++;\n }\n const disc = Math.exp(-r * T);\n return {\n callFV: (callSum / paths) * disc,\n putFV: (putSum / paths) * disc,\n probProfit: profitCount / paths,\n };\n }\n\n // Confidence Score 0–100\n // Replaces four separate boolean gates with one auditable number.\n // Quinn trades only when score ≥ min_confidence_score (default 65).\n function calcConfidence(\n sentiment: number,\n mcProb: number,\n frictionPct: number,\n isLiquid: boolean,\n ): number {\n const sentScore = Math.min(25, Math.max(0, sentiment * 50));\n const mcScore = Math.min(25, Math.max(0, (mcProb - 0.5) * 250));\n const frictScore = Math.min(25, Math.max(0, (0.30 - frictionPct) * 125));\n const liqScore = isLiquid ? 25 : 0;\n return Math.round(sentScore + mcScore + frictScore + liqScore);\n }\n\n // ── ALPACA LAYER ─────────────────────────────────────────\n const aH: Record = ALPACA_KEY\n ? {\n \"APCA-API-KEY-ID\": ALPACA_KEY,\n \"APCA-API-SECRET-KEY\": ALPACA_SECRET || \"\",\n \"Content-Type\": \"application/json\",\n }\n : {};\n\n async function alpacaGet(path: string): Promise {\n if (!ALPACA_KEY) return null;\n try {\n const r = await fetch(`${ALPACA_BASE}${path}`, { headers: aH });\n return await r.json();\n } catch {\n return null;\n }\n }\n\n async function getAlpacaAccount(): Promise {\n return alpacaGet(\"/v2/account\");\n }\n async function getAlpacaPositions(): Promise {\n return (await alpacaGet(\"/v2/positions\")) || [];\n }\n async function getAlpacaOrders(status = \"all\", limit = 50): Promise {\n return (await alpacaGet(`/v2/orders?status=${status}&limit=${limit}`)) ||\n [];\n }\n\n // Settled cash only — GFV prevention\n async function getAvailableCash(): Promise {\n const acct = await getAlpacaAccount();\n return parseFloat(acct?.non_marginable_buying_power || \"0\");\n }\n\n async function getMarketQuote(symbol: string): Promise<\n {\n price: number;\n bid: number;\n ask: number;\n spread: number;\n } | null\n > {\n if (!ALPACA_KEY) return null;\n try {\n const r = await fetch(\n `https://data.alpaca.markets/v2/stocks/${symbol}/quotes/latest`,\n { headers: aH },\n );\n const d = await r.json();\n const q = d.quote;\n if (!q) return null;\n const bid = q.bp || 0, ask = q.ap || 0;\n return { price: (ask + bid) / 2, bid, ask, spread: ask - bid };\n } catch {\n return null;\n }\n }\n\n // ── PDT GUARD ─────────────────────────────────────────────\n // Counts round-trips (open+close same calendar day) in last 5 business days.\n // Blocks trade if count ≥ max_day_trades_5d.\n async function checkPDT(): Promise<\n { ok: boolean; dayTradeCount: number; limit: number }\n > {\n try {\n const cfg = await loadSettings();\n const limit = parseInt(cfg.max_day_trades_5d || \"3\");\n const swingMode = cfg.swing_mode === \"true\";\n\n const cutoff = new Date(Date.now() - 5 * 24 * 60 * 60 * 1000)\n .toISOString();\n const res = await sqlite.execute({\n sql: `SELECT created_at, closed_at FROM quinn_trades\n WHERE is_day_trade = 1 AND created_at > ? AND status = 'closed'`,\n args: [cutoff],\n });\n\n const dayTradeCount = res.rows.length;\n const ok = swingMode || dayTradeCount < limit;\n\n return { ok, dayTradeCount, limit };\n } catch {\n return { ok: true, dayTradeCount: 0, limit: 3 };\n }\n }\n\n // ── CIRCUIT BREAKER ──────────────────────────────────────\n async function checkDrawdown(): Promise<{\n tripped: boolean;\n reason: string;\n loss3d: number;\n }> {\n try {\n const cfg = await loadSettings();\n if (cfg.circuit_breaker_triggered === \"true\") {\n return {\n tripped: true,\n reason:\n \"Previously tripped. Reset circuit_breaker_triggered to false to resume.\",\n loss3d: 0,\n };\n }\n\n const windowDays = parseInt(cfg.circuit_breaker_window_days || \"3\");\n const limitPct = parseFloat(cfg.circuit_breaker_pct || \"0.15\");\n const cutoff = new Date(Date.now() - windowDays * 24 * 60 * 60 * 1000)\n .toISOString();\n\n const lossRes = await sqlite.execute({\n sql: `SELECT COALESCE(SUM(pnl),0) FROM quinn_trades\n WHERE closed_at > ? AND status='closed' AND pnl < 0`,\n args: [cutoff],\n });\n\n const loss3d = Math.abs(parseFloat(String(lossRes.rows[0]?.[0] || \"0\")));\n if (loss3d === 0) {\n return { tripped: false, reason: \"No recent losses.\", loss3d: 0 };\n }\n\n const acct = await getAlpacaAccount();\n const equity = parseFloat(acct?.equity || \"0\");\n if (equity <= 0) {\n return { tripped: false, reason: \"Cannot read equity.\", loss3d };\n }\n\n const lossPct = loss3d / equity;\n if (lossPct >= limitPct) {\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_settings (key,value,updated_at) VALUES ('circuit_breaker_triggered','true',CURRENT_TIMESTAMP)`,\n args: [],\n });\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_settings (key,value,updated_at) VALUES ('autonomous_mode','false',CURRENT_TIMESTAMP)`,\n args: [],\n });\n await email({\n subject: \"[QUINN v3] ⚠️ CIRCUIT BREAKER TRIPPED\",\n text: `Quinn halted.\\n\\n3-day loss: $${loss3d.toFixed(2)} (${\n (lossPct * 100).toFixed(1)\n }% of equity $${equity.toFixed(2)})\\nLimit: ${\n (limitPct * 100).toFixed(0)\n }%\\n\\nReset: set circuit_breaker_triggered=false and autonomous_mode=true.`,\n });\n return {\n tripped: true,\n reason: `${\n (lossPct * 100).toFixed(1)\n }% loss in ${windowDays}d breached ${\n (limitPct * 100).toFixed(0)\n }% limit.`,\n loss3d,\n };\n }\n\n return {\n tripped: false,\n reason: `${(lossPct * 100).toFixed(1)}% loss within ${\n (limitPct * 100).toFixed(0)\n }% limit.`,\n loss3d,\n };\n } catch (e: any) {\n return {\n tripped: false,\n reason: `CB check error: ${e.message}`,\n loss3d: 0,\n };\n }\n }\n\n // ── GHOST TRADE REVERSE LOOKUP ───────────────────────────\n // On every startup, cross-reference Alpaca filled orders against DB.\n // Any filled order not in DB is a ghost trade — log it immediately.\n async function resolveGhostTrades(): Promise {\n if (!ALPACA_KEY) return \"No Alpaca key.\";\n try {\n const alpacaOrders = await getAlpacaOrders(\"filled\", 50);\n let ghosts = 0;\n\n for (const o of alpacaOrders) {\n const existing = await sqlite.execute({\n sql: `SELECT id FROM quinn_trades WHERE alpaca_id = ?`,\n args: [o.id],\n });\n if (!existing.rows.length && o.status === \"filled\") {\n await sqlite.execute({\n sql: `INSERT INTO quinn_trades\n (symbol,side,qty,entry_price,status,alpaca_id,reason,confidence_score)\n VALUES (?,?,?,?,'open',?,'GHOST — recovered from Alpaca',0)`,\n args: [o.symbol, o.side, o.filled_qty, o.filled_avg_price, o.id],\n });\n ghosts++;\n }\n }\n\n return ghosts > 0\n ? `⚠️ Recovered ${ghosts} ghost trade(s). Review immediately.`\n : \"No ghost trades found.\";\n } catch (e: any) {\n return `Ghost check error: ${e.message}`;\n }\n }\n\n // ── FILL RECONCILIATION ──────────────────────────────────\n async function reconcileFills(): Promise {\n if (!ALPACA_KEY) return \"No Alpaca key.\";\n try {\n const pending = await sqlite.execute(\n `SELECT id,alpaca_id,symbol,side,qty,entry_price,bs_price,\n sentiment,mc_prob,friction_pct,confidence_score\n FROM quinn_trades WHERE status IN ('submitted','pending') AND alpaca_id IS NOT NULL`,\n );\n\n if (!pending.rows.length) return \"Nothing to reconcile.\";\n let reconciled = 0;\n\n for (const row of pending.rows) {\n const [\n id,\n alpacaId,\n symbol,\n side,\n qty,\n entryPrice,\n bsPrice,\n sentiment,\n mcProb,\n frictionPct,\n confScore,\n ] = row as any[];\n if (!alpacaId) continue;\n\n try {\n const o = await alpacaGet(`/v2/orders/${alpacaId}`);\n if (!o) continue;\n\n if (o.status === \"filled\") {\n const fillPrice = parseFloat(o.filled_avg_price || \"0\");\n const fillQty = parseFloat(o.filled_qty || String(qty));\n\n if (side === \"buy\") {\n await sqlite.execute({\n sql:\n `UPDATE quinn_trades SET status='open', entry_price=?, qty=? WHERE id=?`,\n args: [fillPrice, fillQty, id],\n });\n }\n\n if (side === \"sell\") {\n const openTrade = await sqlite.execute({\n sql: `SELECT id,entry_price,qty FROM quinn_trades\n WHERE symbol=? AND side='buy' AND status='open'\n ORDER BY created_at ASC LIMIT 1`,\n args: [symbol],\n });\n\n let pnl = 0, theorPnl = 0;\n if (openTrade.rows.length) {\n const [openId, openEntry, openQty] = openTrade.rows[0] as any[];\n pnl = (fillPrice - parseFloat(String(openEntry))) *\n parseFloat(String(openQty));\n theorPnl = parseFloat(String(bsPrice || 0)) *\n parseFloat(String(openQty));\n const pnlVsTheor = pnl - theorPnl;\n\n // Is this a day trade?\n const buyDateRes = await sqlite.execute({\n sql: `SELECT created_at FROM quinn_trades WHERE id=?`,\n args: [openId],\n });\n const buyDate = new Date(String(buyDateRes.rows[0]?.[0] || \"\"))\n .toDateString();\n const sellDate = new Date().toDateString();\n const isDayTrade = buyDate === sellDate ? 1 : 0;\n\n // Post-mortem\n const note = pnl > 0\n ? `WIN: ${symbol} +$${pnl.toFixed(2)}. Entry $${\n parseFloat(String(openEntry)).toFixed(2)\n }, exit $${fillPrice.toFixed(2)}. Conf=${confScore}. ${\n pnlVsTheor >= 0\n ? `Beat BS by $${pnlVsTheor.toFixed(2)}`\n : `Below BS by $${\n Math.abs(pnlVsTheor).toFixed(2)\n } — friction higher than modeled`\n }.`\n : `LOSS: ${symbol} $${pnl.toFixed(2)}. Entry $${\n parseFloat(String(openEntry)).toFixed(2)\n }, exit $${\n fillPrice.toFixed(2)\n }. Conf=${confScore}, Sentiment=${\n parseFloat(String(sentiment)).toFixed(2)\n }, MC=${parseFloat(String(mcProb)).toFixed(2)}. ${\n parseFloat(String(sentiment)) < 0.2\n ? \"Low sentiment was a warning.\"\n : \"Math looked good — likely macro/news event overwhelmed signal.\"\n }`;\n\n const signals = JSON.stringify({\n entryPrice: openEntry,\n exitPrice: fillPrice,\n sentiment,\n mcProb,\n frictionPct,\n bsPrice,\n confScore,\n actualPnl: pnl,\n theoreticalPnl: theorPnl,\n pnlVsTheoretical: pnlVsTheor,\n isDayTrade,\n });\n\n // Update both legs\n for (const closeId of [id, openId]) {\n await sqlite.execute({\n sql: `UPDATE quinn_trades SET\n status='closed', exit_price=?, pnl=?, pnl_vs_theoretical=?,\n postmortem=?, postmortem_signals=?, is_day_trade=?, closed_at=CURRENT_TIMESTAMP\n WHERE id=?`,\n args: [\n fillPrice,\n pnl,\n pnlVsTheor,\n note,\n signals,\n isDayTrade,\n closeId,\n ],\n });\n }\n\n // Save lesson to memory\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_memory (key,content,updated_at)\n VALUES (?,?,CURRENT_TIMESTAMP)`,\n args: [`pm_${symbol}_${Date.now()}`, note],\n });\n } else {\n await sqlite.execute({\n sql:\n `UPDATE quinn_trades SET status='closed',exit_price=?,closed_at=CURRENT_TIMESTAMP WHERE id=?`,\n args: [fillPrice, id],\n });\n }\n }\n reconciled++;\n } else if ([\"canceled\", \"expired\", \"rejected\"].includes(o.status)) {\n await sqlite.execute({\n sql: `UPDATE quinn_trades SET status=? WHERE id=?`,\n args: [o.status, id],\n });\n }\n } catch { /* skip this order */ }\n }\n\n return `Reconciled ${reconciled}/${pending.rows.length} orders.`;\n } catch (e: any) {\n return `Reconcile error: ${e.message}`;\n }\n }\n\n // ── LIQUIDITY SNIPER ─────────────────────────────────────\n async function checkLiquidity(symbol: string, targetQty: number): Promise<{\n is_liquid: boolean;\n avg_10d_volume: number;\n reason: string;\n }> {\n if (!ALPACA_KEY) {\n return {\n is_liquid: true,\n avg_10d_volume: 999999,\n reason: \"No key — skipped\",\n };\n }\n try {\n const r = await fetch(\n `https://data.alpaca.markets/v2/stocks/${symbol}/bars?timeframe=1Day&limit=10`,\n { headers: aH },\n );\n const d = await r.json();\n // Handle both Alpaca response shapes\n const bars: any[] = Array.isArray(d.bars)\n ? d.bars\n : (d.bars?.[symbol] || []);\n\n if (!bars || bars.length < 5) {\n return {\n is_liquid: false,\n avg_10d_volume: 0,\n reason: `Insufficient history (${bars?.length || 0} days)`,\n };\n }\n\n const cfg = await loadSettings();\n const minAdv = parseInt(cfg.min_adv || \"50000\");\n const adv = bars.reduce((s: number, b: any) => s + (b.v || 0), 0) /\n bars.length;\n const isLiq = adv >= minAdv;\n\n return {\n is_liquid: isLiq,\n avg_10d_volume: Math.round(adv),\n reason: isLiq\n ? `ADV ${\n Math.round(adv).toLocaleString()\n } ≥ ${minAdv.toLocaleString()} floor`\n : `ADV ${\n Math.round(adv).toLocaleString()\n } below ${minAdv.toLocaleString()} — zombie stock risk`,\n };\n } catch (e: any) {\n return {\n is_liquid: false,\n avg_10d_volume: 0,\n reason: `Liq check error: ${e.message}`,\n };\n }\n }\n\n // ── WEB SEARCH ───────────────────────────────────────────\n async function searchWeb(query: string): Promise {\n if (!TAVILY_KEY) return \"[]\";\n try {\n const r = await fetch(\"https://api.tavily.com/search\", {\n method: \"POST\",\n headers: { \"Content-Type\": \"application/json\" },\n body: JSON.stringify({\n api_key: TAVILY_KEY,\n query,\n search_depth: \"basic\",\n max_results: 5,\n }),\n });\n const d = await r.json();\n if (!d.results) return \"[]\";\n return JSON.stringify(d.results.map((r: any) => ({\n title: r.title,\n content: r.content.substring(0, 300),\n })));\n } catch {\n return \"[]\";\n }\n }\n\n // ── SENTIMENT ENGINE ─────────────────────────────────────\n function scoreSentiment(raw: string): number {\n try {\n const parsed = JSON.parse(raw);\n const text = parsed.map((r: any) => r.title + \" \" + r.content).join(\" \")\n .toLowerCase();\n const bull =\n (text.match(\n /beat|surge|rally|upgrade|strong|record|growth|bullish|launch|partnership|guidance|beat|profit|dividend/g,\n ) || []).length;\n const bear =\n (text.match(\n /miss|drop|fall|downgrade|weak|loss|crash|bearish|lawsuit|investigation|recall|fraud|warning|cut|layoff|bankrupt/g,\n ) || []).length;\n const total = bull + bear;\n return total > 0 ? (bull - bear) / total : 0;\n } catch {\n return 0;\n }\n }\n\n // ── FULL STOCK ANALYSIS ──────────────────────────────────\n async function analyzeStock(\n symbol: string,\n volInput?: number,\n ): Promise {\n symbol = symbol.toUpperCase();\n const cfg = await loadSettings();\n const quote = await getMarketQuote(symbol);\n const S = quote?.price || 100;\n const spread = quote?.spread || 0.10;\n\n // News sentiment\n const rawNews = await searchWeb(\n `${symbol} stock news today earnings catalyst`,\n );\n const sentiment = scoreSentiment(rawNews);\n\n // Math\n const sig = volInput || 0.25;\n const T = 30 / 365; // 30-day horizon\n const r = 0.053; // risk-free rate\n const K = S; // ATM\n\n const bs = bsCalc(\"call\", S, K, T, r, sig);\n const mc = monteCarloEdge(S, K, T, r, sig, 2000);\n const commEst = 0.65;\n const frictPct = bs.price > 0\n ? ((commEst + spread / 2) * 5) / bs.price\n : 999;\n\n // Liquidity\n const maxPos = parseFloat(cfg.max_position_usd || \"10\");\n const suggestedQ = Math.max(1, Math.floor(maxPos / S));\n const liq = await checkLiquidity(symbol, suggestedQ);\n\n // Confidence score\n const conf = calcConfidence(\n sentiment,\n mc.probProfit,\n frictPct,\n liq.is_liquid,\n );\n const minConf = parseInt(cfg.min_confidence_score || \"65\");\n\n // Blockers\n const blockers: string[] = [];\n if (sentiment < parseFloat(cfg.min_sentiment || \"0.3\")) {\n blockers.push(`sentiment ${sentiment.toFixed(2)} < ${cfg.min_sentiment}`);\n }\n if (frictPct > parseFloat(cfg.max_friction_pct || \"0.30\")) {\n blockers.push(\n `friction ${(frictPct * 100).toFixed(1)}% > ${\n parseFloat(cfg.max_friction_pct || \"0.30\") * 100\n }%`,\n );\n }\n if (mc.probProfit < 0.52) {\n blockers.push(`MC prob ${(mc.probProfit * 100).toFixed(1)}% < 52%`);\n }\n if (!liq.is_liquid) blockers.push(liq.reason);\n if (conf < minConf) blockers.push(`confidence ${conf} < ${minConf}`);\n\n const verdict = blockers.length === 0 ? \"BUY\" : \"HOLD\";\n\n return JSON.stringify({\n symbol,\n verdict,\n confidence: conf,\n price: +S.toFixed(2),\n spread: +spread.toFixed(4),\n sentiment: +sentiment.toFixed(3),\n bsCallPrice: +bs.price.toFixed(4),\n bsDelta: +bs.delta.toFixed(4),\n bsGamma: +bs.gamma.toFixed(6),\n bsVega: +bs.vega.toFixed(4),\n bsTheta: +bs.theta.toFixed(4),\n mcCallFV: +mc.callFV.toFixed(4),\n probProfit: +mc.probProfit.toFixed(3),\n frictionPct: +frictPct.toFixed(3),\n adv: liq.avg_10d_volume,\n liquidityOk: liq.is_liquid,\n suggestedQty: suggestedQ,\n blockers,\n signals: {\n sentiment,\n mcProb: mc.probProfit,\n frictionPct: frictPct,\n bsPrice: bs.price,\n adv: liq.avg_10d_volume,\n confScore: conf,\n },\n });\n }\n\n // ── WATCHLIST SCAN ───────────────────────────────────────\n // 1000ms pacing between calls — free tier rate limit protection\n async function scanWatchlist(): Promise {\n const wl = await sqlite.execute(`SELECT symbol FROM quinn_watchlist`);\n const out = [];\n for (const row of wl.rows) {\n try {\n out.push(JSON.parse(await analyzeStock(String(row[0]))));\n } catch { /* skip */ }\n await new Promise((r) => setTimeout(r, 1000));\n }\n out.sort((a, b) => {\n if (a.verdict === \"BUY\" && b.verdict !== \"BUY\") return -1;\n if (b.verdict === \"BUY\" && a.verdict !== \"BUY\") return 1;\n return b.confidence - a.confidence;\n });\n return JSON.stringify(out);\n }\n\n // ── ORDER SUBMISSION ─────────────────────────────────────\n async function submitOrder(\n symbol: string,\n side: \"buy\" | \"sell\",\n qty: number,\n reason: string,\n signals: Record = {},\n ): Promise<{ success: boolean; orderId?: string; error?: string }> {\n if (!ALPACA_KEY) return { success: false, error: \"No Alpaca key.\" };\n\n // Gate 1: circuit breaker\n const cb = await checkDrawdown();\n if (cb.tripped) {\n return { success: false, error: `Circuit breaker: ${cb.reason}` };\n }\n\n // Gate 2: autonomous mode\n const cfg = await loadSettings();\n if (cfg.autonomous_mode !== \"true\") {\n return { success: false, error: \"Autonomous mode off. Analysis only.\" };\n }\n\n // Gate 3: PDT guard\n const pdt = await checkPDT();\n if (!pdt.ok) {\n return {\n success: false,\n error:\n `PDT limit reached (${pdt.dayTradeCount}/${pdt.limit} day trades in 5d). Enable swing_mode or wait.`,\n };\n }\n\n // Gate 4: settled cash check (GFV prevention)\n if (side === \"buy\") {\n const availCash = await getAvailableCash();\n const orderCost = qty * (signals.price || 0);\n if (orderCost > availCash) {\n return {\n success: false,\n error: `Insufficient settled cash. Need $${\n orderCost.toFixed(2)\n }, have $${availCash.toFixed(2)} non-marginable.`,\n };\n }\n }\n\n try {\n const res = await fetch(`${ALPACA_BASE}/v2/orders`, {\n method: \"POST\",\n headers: aH,\n body: JSON.stringify({\n symbol,\n qty: String(qty),\n side,\n type: \"market\",\n time_in_force: \"day\",\n }),\n });\n const data = await res.json();\n if (!res.ok) {\n return { success: false, error: data.message || JSON.stringify(data) };\n }\n\n await sqlite.execute({\n sql: `INSERT INTO quinn_trades\n (symbol,side,qty,entry_price,reason,sentiment,mc_prob,\n friction_pct,adv,bs_price,confidence_score,status,alpaca_id)\n VALUES (?,?,?,?,?,?,?,?,?,?,?,'submitted',?)`,\n args: [\n symbol,\n side,\n qty,\n signals.price || 0,\n reason,\n signals.sentiment || 0,\n signals.mcProb || 0,\n signals.frictionPct || 0,\n signals.adv || 0,\n signals.bsPrice || 0,\n signals.confScore || 0,\n data.id,\n ],\n });\n\n return { success: true, orderId: data.id };\n } catch (e: any) {\n return { success: false, error: e.message };\n }\n }\n\n async function closePosition(symbol: string, reason: string) {\n if (!ALPACA_KEY) return { success: false, error: \"No Alpaca key.\" };\n try {\n const pos = await alpacaGet(`/v2/positions/${symbol}`);\n if (!pos || pos.code) {\n return { success: false, error: \"No open position.\" };\n }\n return await submitOrder(\n symbol,\n \"sell\",\n Math.abs(parseFloat(pos.qty)),\n reason,\n );\n } catch (e: any) {\n return { success: false, error: e.message };\n }\n }\n\n // ── META-LEARNER ─────────────────────────────────────────\n async function analyzePerformance(): Promise {\n try {\n const cfg = await loadSettings();\n const minT = parseInt(cfg.meta_learner_min_trades || \"20\");\n const closed = await sqlite.execute(\n `SELECT symbol,pnl,pnl_vs_theoretical,postmortem_signals FROM quinn_trades\n WHERE status='closed' AND pnl IS NOT NULL ORDER BY closed_at DESC`,\n );\n\n const trades = closed.rows.map((r: any) => {\n let sig: any = {};\n try {\n sig = JSON.parse(r[3] || \"{}\");\n } catch {}\n return {\n symbol: r[0],\n pnl: parseFloat(r[1] || \"0\"),\n pnlVsT: parseFloat(r[2] || \"0\"),\n sig,\n };\n });\n\n if (!trades.length) {\n return JSON.stringify({ status: \"No closed trades yet.\" });\n }\n\n const n = trades.length;\n const winners = trades.filter((t) => t.pnl > 0);\n const losers = trades.filter((t) => t.pnl < 0);\n const winRate = winners.length / n;\n const totPnl = trades.reduce((s, t) => s + t.pnl, 0);\n const avgW = winners.length\n ? winners.reduce((s, t) => s + t.pnl, 0) / winners.length\n : 0;\n const avgL = losers.length\n ? losers.reduce((s, t) => s + t.pnl, 0) / losers.length\n : 0;\n const pf = avgL !== 0 ? Math.abs(avgW / avgL) : 999;\n\n // Signal correlations\n const wS = trades.filter((t) => t.sig.sentiment !== undefined);\n const hiS = wS.filter((t) => t.sig.sentiment >= 0.5);\n const loS = wS.filter((t) =>\n t.sig.sentiment >= 0 && t.sig.sentiment < 0.5\n );\n const hiSwr = hiS.length\n ? hiS.filter((t) => t.pnl > 0).length / hiS.length\n : 0;\n const loSwr = loS.length\n ? loS.filter((t) => t.pnl > 0).length / loS.length\n : 0;\n\n const hiC = wS.filter((t) => t.sig.confScore >= 75);\n const hiCwr = hiC.length\n ? hiC.filter((t) => t.pnl > 0).length / hiC.length\n : 0;\n\n const autoTuned: string[] = [];\n const suggestions: string[] = [];\n\n if (n >= minT) {\n // Auto-tune: raise min_sentiment if high-sentiment trades clearly outperform\n if (hiSwr > loSwr + 0.2 && loS.length >= 5) {\n const cur = parseFloat(cfg.min_sentiment || \"0.3\");\n const nxt = Math.min(0.65, cur + 0.05);\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_settings (key,value,updated_at) VALUES ('min_sentiment',?,CURRENT_TIMESTAMP)`,\n args: [String(nxt)],\n });\n autoTuned.push(\n `min_sentiment: ${cur} → ${nxt} (hi-sent WR ${\n (hiSwr * 100).toFixed(0)\n }% vs lo ${(loSwr * 100).toFixed(0)}%)`,\n );\n }\n // Auto-tune: tighten friction if profit factor poor\n if (pf < 1.2) {\n const cur = parseFloat(cfg.max_friction_pct || \"0.30\");\n const nxt = Math.max(0.12, cur - 0.03);\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_settings (key,value,updated_at) VALUES ('max_friction_pct',?,CURRENT_TIMESTAMP)`,\n args: [String(nxt)],\n });\n autoTuned.push(\n `max_friction_pct: ${cur} → ${nxt} (profit factor ${\n pf.toFixed(2)\n })`,\n );\n }\n // Auto-tune: raise confidence floor if high-conf trades outperform\n if (hiCwr > 0.65 && hiC.length >= 5) {\n const cur = parseInt(cfg.min_confidence_score || \"65\");\n const nxt = Math.min(85, cur + 5);\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_settings (key,value,updated_at) VALUES ('min_confidence_score',?,CURRENT_TIMESTAMP)`,\n args: [String(nxt)],\n });\n autoTuned.push(\n `min_confidence_score: ${cur} → ${nxt} (high-conf WR ${\n (hiCwr * 100).toFixed(0)\n }%)`,\n );\n }\n } else {\n suggestions.push(\n `Observer mode: ${minT - n} more closed trades before auto-tuning`,\n );\n }\n\n const report = {\n totalTrades: n,\n winRate: +winRate.toFixed(3),\n totalPnl: +totPnl.toFixed(2),\n avgWin: +avgW.toFixed(2),\n avgLoss: +avgL.toFixed(2),\n profitFactor: +pf.toFixed(3),\n sentimentEdge: {\n hiSentWR: +hiSwr.toFixed(3),\n loSentWR: +loSwr.toFixed(3),\n },\n confidenceEdge: { hiConfWR: +hiCwr.toFixed(3) },\n autoTuned: autoTuned.length\n ? autoTuned\n : [\"None — threshold not reached yet\"],\n suggestions,\n observerMode: n < minT,\n };\n\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_memory (key,content,updated_at) VALUES ('last_perf_analysis',?,CURRENT_TIMESTAMP)`,\n args: [JSON.stringify(report)],\n });\n\n return JSON.stringify(report);\n } catch (e: any) {\n return `Perf error: ${e.message}`;\n }\n }\n\n // ── PORTFOLIO ────────────────────────────────────────────\n async function getPortfolio(): Promise {\n const [acct, pos, orders] = await Promise.all([\n getAlpacaAccount(),\n getAlpacaPositions(),\n getAlpacaOrders(\"all\", 20),\n ]);\n const cash = parseFloat(acct?.non_marginable_buying_power || \"0\");\n return JSON.stringify({\n account: acct,\n availableCash: cash,\n positions: pos,\n recentOrders: orders,\n });\n }\n\n async function getTradeLedger(limit = 20): Promise {\n const r = await sqlite.execute({\n sql:\n `SELECT id,symbol,side,qty,entry_price,exit_price,pnl,pnl_vs_theoretical,\n confidence_score,status,postmortem,is_day_trade,created_at,closed_at\n FROM quinn_trades ORDER BY created_at DESC LIMIT ?`,\n args: [limit],\n });\n return JSON.stringify(r.rows);\n }\n\n async function updateWatchlist(\n symbol: string,\n action: \"add\" | \"remove\",\n ): Promise {\n symbol = symbol.toUpperCase();\n if (action === \"add\") {\n await sqlite.execute({\n sql: `INSERT OR IGNORE INTO quinn_watchlist (symbol) VALUES (?)`,\n args: [symbol],\n });\n return `Added ${symbol}.`;\n } else {\n await sqlite.execute({\n sql: `DELETE FROM quinn_watchlist WHERE symbol=?`,\n args: [symbol],\n });\n return `Removed ${symbol}.`;\n }\n }\n\n async function updateSetting(key: string, value: string): Promise {\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_settings (key,value,updated_at) VALUES (?,?,CURRENT_TIMESTAMP)`,\n args: [key, value],\n });\n return `${key} = ${value}`;\n }\n\n async function sendAlert(subject: string, body: string): Promise {\n try {\n await email({ subject: `[QUINN v3] ${subject}`, text: body });\n return `Sent: ${subject}`;\n } catch (e: any) {\n return `Email error: ${e.message}`;\n }\n }\n\n async function rememberThis(key: string, content: string): Promise {\n await sqlite.execute({\n sql:\n `INSERT OR REPLACE INTO quinn_memory (key,content,updated_at) VALUES (?,?,CURRENT_TIMESTAMP)`,\n args: [key.toLowerCase().replace(/\\s+/g, \"_\"), content],\n });\n return `Saved: ${key}`;\n }\n\n // ── TOOLS REGISTRY ───────────────────────────────────────\n const tools = [\n {\n type: \"function\",\n function: {\n name: \"analyzeStock\",\n description:\n \"Full quant analysis: live quote + Black-Scholes + Monte Carlo + liquidity + sentiment + confidence score. Returns BUY/HOLD verdict.\",\n parameters: {\n type: \"object\",\n properties: {\n symbol: { type: \"string\" },\n volatility: {\n type: \"number\",\n description: \"Annualized vol decimal e.g. 0.25\",\n },\n },\n required: [\"symbol\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"scanWatchlist\",\n description:\n \"Scan all watchlist symbols, run full analysis on each (1s pacing), return sorted by confidence score.\",\n parameters: { type: \"object\", properties: {}, required: [] },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"checkLiquidity\",\n description: \"10-day ADV liquidity check. Call before any trade.\",\n parameters: {\n type: \"object\",\n properties: {\n symbol: { type: \"string\" },\n target_qty: { type: \"number\" },\n },\n required: [\"symbol\", \"target_qty\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"checkDrawdown\",\n description: \"Check circuit breaker status. Call before any trade.\",\n parameters: { type: \"object\", properties: {}, required: [] },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"checkPDT\",\n description: \"Check Pattern Day Trader count in last 5 business days.\",\n parameters: { type: \"object\", properties: {}, required: [] },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"reconcileFills\",\n description:\n \"Sync Alpaca fills to DB, compute PnL, write post-mortems.\",\n parameters: { type: \"object\", properties: {}, required: [] },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"resolveGhostTrades\",\n description:\n \"Cross-reference Alpaca filled orders against DB. Recover any missing trades.\",\n parameters: { type: \"object\", properties: {}, required: [] },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"analyzePerformance\",\n description:\n \"Meta-learner: analyze closed trade win rates by signal, auto-tune settings after 20 trades.\",\n parameters: { type: \"object\", properties: {}, required: [] },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"executeTrade\",\n description:\n \"Submit market order to Alpaca. ONLY after analyzeStock returns BUY with confidence ≥ min_confidence_score, circuit breaker clear, PDT ok, liquidity ok, settled cash sufficient.\",\n parameters: {\n type: \"object\",\n properties: {\n symbol: { type: \"string\" },\n side: { type: \"string\", enum: [\"buy\", \"sell\"] },\n qty: { type: \"number\" },\n reason: { type: \"string\" },\n signals: {\n type: \"object\",\n description: \"Pass signals from analyzeStock for ledger storage.\",\n },\n },\n required: [\"symbol\", \"side\", \"qty\", \"reason\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"closePosition\",\n description: \"Close an open position.\",\n parameters: {\n type: \"object\",\n properties: {\n symbol: { type: \"string\" },\n reason: { type: \"string\" },\n },\n required: [\"symbol\", \"reason\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"getPortfolio\",\n description:\n \"Get Alpaca account, available settled cash (non_marginable_buying_power), and open positions.\",\n parameters: { type: \"object\", properties: {}, required: [] },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"getTradeLedger\",\n description: \"Get trade history with PnL and post-mortems.\",\n parameters: {\n type: \"object\",\n properties: { limit: { type: \"number\" } },\n required: [],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"updateWatchlist\",\n description: \"Add or remove symbol from watchlist.\",\n parameters: {\n type: \"object\",\n properties: {\n symbol: { type: \"string\" },\n action: { type: \"string\", enum: [\"add\", \"remove\"] },\n },\n required: [\"symbol\", \"action\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"updateSetting\",\n description: \"Update a Quinn config setting.\",\n parameters: {\n type: \"object\",\n properties: { key: { type: \"string\" }, value: { type: \"string\" } },\n required: [\"key\", \"value\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"sendAlert\",\n description: \"Email Void a trade alert or report.\",\n parameters: {\n type: \"object\",\n properties: { subject: { type: \"string\" }, body: { type: \"string\" } },\n required: [\"subject\", \"body\"],\n },\n },\n },\n {\n type: \"function\",\n function: {\n name: \"rememberThis\",\n description: \"Save a note to Quinn's memory.\",\n parameters: {\n type: \"object\",\n properties: { key: { type: \"string\" }, content: { type: \"string\" } },\n required: [\"key\", \"content\"],\n },\n },\n },\n ];\n\n // ── CONTEXT LOADER ───────────────────────────────────────\n let ctx = \"\";\n try {\n const [cfg, wl, mem, trades, cb, pdt] = await Promise.all([\n sqlite.execute(`SELECT key,value FROM quinn_settings`),\n sqlite.execute(`SELECT symbol FROM quinn_watchlist`),\n sqlite.execute(\n `SELECT key,content FROM quinn_memory ORDER BY updated_at DESC LIMIT 6`,\n ),\n sqlite.execute(\n `SELECT symbol,side,qty,status,confidence_score,pnl,created_at FROM quinn_trades ORDER BY created_at DESC LIMIT 5`,\n ),\n checkDrawdown(),\n checkPDT(),\n ]);\n\n ctx += `\\n=== CONFIG ===\\n${\n cfg.rows.map((r: any) => `${r[0]}: ${r[1]}`).join(\"\\n\")\n }`;\n ctx += `\\n\\n=== WATCHLIST ===\\n${wl.rows.map((r: any) => r[0]).join(\", \")}`;\n if (mem.rows.length) {\n ctx += `\\n\\n=== LESSONS ===\\n${\n mem.rows.map((r: any) => `${r[0]}: ${r[1]}`).join(\"\\n\")\n }`;\n }\n if (trades.rows.length) {\n ctx += `\\n\\n=== RECENT TRADES ===\\n`;\n ctx += trades.rows.map((r: any) =>\n `${r[6]}: ${r[1]} ${r[0]} x${r[2]} conf=${r[4]} [${r[3]}]${\n r[5] ? ` PnL $${parseFloat(r[5]).toFixed(2)}` : \"\"\n }`\n ).join(\"\\n\");\n }\n ctx += `\\n\\n=== SAFETY ===`;\n ctx += `\\nCircuit breaker: ${\n cb.tripped ? `⚠️ TRIPPED — ${cb.reason}` : `Clear (${cb.reason})`\n }`;\n ctx += `\\nPDT: ${pdt.dayTradeCount}/${pdt.limit} day trades used in 5d`;\n ctx += `\\nAlpaca: ${\n ALPACA_BASE.includes(\"paper\") ? \"PAPER TRADING\" : \"⚠️ LIVE TRADING\"\n }`;\n ctx += `\\nModel: ${QUINN_MODEL}`;\n } catch {\n ctx = \"\";\n }\n\n // ── SYSTEM PROMPT ────────────────────────────────────────\n const systemPrompt =\n `You are QUINN v3, an elite autonomous quant trading agent built by Void. You trade US stocks via Alpaca Markets using a multi-signal quant framework: Black-Scholes pricing, Monte Carlo simulation, news sentiment, liquidity analysis, friction modeling, and a composite Confidence Score (0-100).\n\nMANDATORY TRADE SEQUENCE — follow every time:\n1. reconcileFills — sync fills first, always\n2. resolveGhostTrades — recover any missing orders\n3. checkDrawdown — if tripped, STOP. No trades.\n4. checkPDT — if limit hit and swing_mode=false, STOP.\n5. scanWatchlist or analyzeStock — find candidates\n6. For BUY candidates: confidence ≥ min_confidence_score AND is_liquid = true\n7. getPortfolio — verify non_marginable_buying_power covers order cost\n8. executeTrade — pass full signals object for ledger\n9. sendAlert — always notify Void after execution\n\nABSOLUTE RISK RULES:\n- Never use account.equity or account.buying_power for sizing — ONLY non_marginable_buying_power (settled cash, GFV prevention)\n- Never trade when circuit_breaker_triggered = true\n- Never trade when autonomous_mode = false\n- Never trade illiquid stocks (is_liquid = false)\n- Never trade on negative sentiment (bear keywords: lawsuit/fraud/investigation/crash)\n- Never exceed max_position_usd per single trade\n- Never place a sell order without a confirmed matching buy in the ledger\n- In swing_mode: buy Monday, target exit Tuesday+ (crosses midnight, avoids PDT)\n\nCONFIDENCE SCORE LOGIC:\n- 0-40: Do not trade under any circumstance\n- 41-64: Analyze only, log findings, wait\n- 65-79: Trade allowed at reduced position size (50% of max_position_usd)\n- 80-100: Full position size\n\nMETA-LEARNER:\n- Run analyzePerformance periodically or when asked\n- Read lesson notes from memory before trading\n- Trust auto-tuned settings — they are evidence-based\n\nYou are cold, precise, and disciplined. No FOMO. No hope trades. No revenge trades. Only statistical edge.\n${ctx}`;\n\n // ── MODEL CALL WITH RETRY ────────────────────────────────\n let body: any;\n try {\n body = await req.json();\n } catch {\n return Response.json({ error: \"Invalid JSON.\" }, { headers });\n }\n\n // Context trim: last 6 messages only (avoids 8k token limit on free model)\n const messages: any[] = (body.messages || []).slice(-6);\n\n const callModel = async (msgs: any[], useTools = true): Promise => {\n const payload: any = {\n model: QUINN_MODEL,\n messages: msgs,\n temperature: 0.2,\n max_tokens: 2048,\n };\n if (useTools) {\n payload.tools = tools;\n payload.tool_choice = \"auto\";\n }\n\n // Retry with exponential backoff — skip retry on 400 (bad request)\n for (let attempt = 0; attempt <= 2; attempt++) {\n try {\n const res = await fetch(\n \"https://openrouter.ai/api/v1/chat/completions\",\n {\n method: \"POST\",\n headers: {\n \"Authorization\": `Bearer ${OPENROUTER_KEY}`,\n \"Content-Type\": \"application/json\",\n \"HTTP-Referer\": \"https://quinn.codeberg.page\",\n \"X-Title\": \"QUINN\",\n },\n body: JSON.stringify(payload),\n },\n );\n\n // Don't retry on client errors (malformed request)\n if (res.status === 400) {\n const e = await res.json();\n throw new Error(`400: ${e.error?.message || JSON.stringify(e)}`);\n }\n\n // Retry on rate limit or server error\n if (res.status === 429 || res.status >= 500) {\n if (attempt === 2) {\n throw new Error(`API failed after 3 attempts (HTTP ${res.status})`);\n }\n await new Promise((r) => setTimeout(r, 2000 * (attempt + 1)));\n continue;\n }\n\n return await res.json();\n } catch (e: any) {\n if (attempt === 2) throw e;\n await new Promise((r) => setTimeout(r, 2000 * (attempt + 1)));\n }\n }\n };\n\n // Background maintenance on every request\n Promise.all([reconcileFills(), resolveGhostTrades()]).catch(() => {});\n\n let response: any;\n try {\n response = await callModel([\n { role: \"system\", content: systemPrompt },\n ...messages,\n ]);\n } catch (e: any) {\n return Response.json({ error: `Model call failed: ${e.message}` }, {\n headers,\n });\n }\n\n if (response?.error) {\n return Response.json({\n error: response.error.message || JSON.stringify(response.error),\n }, { headers });\n }\n\n const message = response?.choices?.[0]?.message;\n if (!message) {\n return Response.json({ error: \"No response from model.\" }, { headers });\n }\n\n // ── TOOL DISPATCHER ──────────────────────────────────────\n if (message.tool_calls?.length > 0) {\n const tc = message.tool_calls[0];\n const name = tc.function.name;\n let args: any = {};\n try {\n args = JSON.parse(tc.function.arguments || \"{}\");\n } catch {\n args = {};\n }\n\n let result = \"\";\n let extra: any = { tool_used: name };\n\n switch (name) {\n case \"analyzeStock\": {\n result = await analyzeStock(args.symbol, args.volatility);\n try {\n const p = JSON.parse(result);\n extra.analysis = p;\n extra.verdict = p.verdict;\n extra.confidence = p.confidence;\n } catch {}\n break;\n }\n case \"scanWatchlist\": {\n result = await scanWatchlist();\n try {\n extra.scan_results = JSON.parse(result);\n } catch {}\n break;\n }\n case \"checkLiquidity\": {\n result = JSON.stringify(\n await checkLiquidity(args.symbol, args.target_qty || 1),\n );\n break;\n }\n case \"checkDrawdown\": {\n const r = await checkDrawdown();\n result = JSON.stringify(r);\n extra.circuit_breaker = r;\n break;\n }\n case \"checkPDT\": {\n const r = await checkPDT();\n result = JSON.stringify(r);\n extra.pdt = r;\n break;\n }\n case \"reconcileFills\": {\n result = await reconcileFills();\n break;\n }\n case \"resolveGhostTrades\": {\n result = await resolveGhostTrades();\n break;\n }\n case \"analyzePerformance\": {\n result = await analyzePerformance();\n try {\n extra.performance = JSON.parse(result);\n } catch {}\n break;\n }\n case \"executeTrade\": {\n const q = await getMarketQuote(args.symbol);\n const signals = { ...(args.signals || {}), price: q?.price || 0 };\n const r = await submitOrder(\n args.symbol,\n args.side,\n args.qty,\n args.reason,\n signals,\n );\n result = JSON.stringify(r);\n extra.trade_executed = r.success;\n extra.trade_symbol = args.symbol;\n extra.trade_side = args.side;\n extra.trade_qty = args.qty;\n break;\n }\n case \"closePosition\": {\n result = JSON.stringify(await closePosition(args.symbol, args.reason));\n break;\n }\n case \"getPortfolio\": {\n result = await getPortfolio();\n try {\n extra.portfolio = JSON.parse(result);\n } catch {}\n break;\n }\n case \"getTradeLedger\": {\n result = await getTradeLedger(args.limit || 20);\n try {\n extra.ledger = JSON.parse(result);\n } catch {}\n break;\n }\n case \"updateWatchlist\": {\n result = await updateWatchlist(args.symbol, args.action);\n break;\n }\n case \"updateSetting\": {\n result = await updateSetting(args.key, args.value);\n break;\n }\n case \"sendAlert\": {\n result = await sendAlert(args.subject, args.body);\n break;\n }\n case \"rememberThis\": {\n result = await rememberThis(args.key, args.content);\n break;\n }\n default:\n result = `Unknown tool: ${name}`;\n }\n\n let finalResp: any;\n try {\n finalResp = await callModel([\n { role: \"system\", content: systemPrompt },\n ...messages,\n message,\n { role: \"tool\", tool_call_id: tc.id, name, content: result },\n ], false);\n } catch (e: any) {\n return Response.json({ error: `Final call failed: ${e.message}` }, {\n headers,\n });\n }\n\n const reply = finalResp?.choices?.[0]?.message?.content || \"No response.\";\n return Response.json({ reply, ...extra }, { headers });\n }\n\n return Response.json({ reply: message.content || \"No response.\" }, {\n headers,\n });\n}","http","createdAt",["D",1776036357045],"branchVersion","branchId","a55c5366-3689-11f1-ba83-42b51c65c3df","id","a55ba98e-3689-11f1-ba83-42b51c65c3df","Quinn-AI",["D",1776010023785],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","imageUrl","61537aff-745c-45d8-8e13-c99e165fb31a","pinnedFileId","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","019d8271-e53c-751a-8a25-767506d4c4fb","tursoName","tursoHostname","tursoMeasuredAt","pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_94":93,"_101":97},"org",{"_94":93,"_95":-5,"_96":97,"_18":54,"_98":99,"_100":25},"branch",{"_13":83,"_32":84,"_43":42,"_85":44,"_86":-5,"_87":-5,"_88":-5,"_89":-5,"_90":-5,"_91":45,"_92":93},"isOwned",25,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","infinarscientiacloud","handle","displayName","avatar","https://avatars.githubusercontent.com/u/275509936?v=4","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_43":44,"_32":45,"_38":149,"_47":-5,"_48":25,"_49":50,"_51":50,"_52":-5,"_53":-5,"_18":54,"_55":-5,"_56":25,"_57":-5,"_58":-5,"_59":60,"_61":62,"_63":-5,"_64":-5,"_65":-5,"_66":-5,"_67":68,"_69":68,"_70":71,"_72":71,"_73":71,"_74":68,"_75":68,"_76":150,"_78":151,"_80":152,"_82":25},"tree",[144],"files",[144],"counts",{"_133":60,"_134":135,"_136":60,"_137":138,"_139":140},"contents",[125],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles",true,"grants","orgs","customDomains",{},{"_43":126,"_32":28,"_30":31,"_127":-5,"_41":42,"_40":14,"_128":25,"_38":129,"_13":130,"_33":37,"_35":36,"_24":25,"_131":-5,"_27":132},"f5369b70-36c6-11f1-b613-42b51c65c3df","parentId","latest",["D",1776036357045],3,"label",[28],"pulls","branches",1,"forks","envVars",{"_141":14,"_143":60},"connections",{"_141":60,"_142":60},"count","unfilledCount","emptyCount",{"_43":126,"_32":28,"_30":31,"_127":-5,"_41":42,"_40":14,"_128":25,"_38":145,"_13":130,"_33":37,"_35":36,"_24":25,"_131":-5,"_146":147,"_27":148},["D",1776036357045],"children",[],[28],["D",1776010023785],{"_94":93,"_101":97},{"_94":93,"_95":-5,"_96":97,"_18":54,"_98":99,"_100":25},{"_13":83,"_32":84,"_43":42,"_85":44,"_86":-5,"_87":-5,"_88":-5,"_89":-5,"_90":-5,"_91":45,"_92":93},"ctx",{"_76":-5,"_122":186,"_187":188,"_189":190,"_191":25,"_192":25,"_193":25,"_194":120,"_195":-5,"_196":-7,"_197":-7,"_198":60,"_199":25,"_200":25,"_201":-5,"_202":25,"_203":25,"_204":25,"_205":-5},"ENV",{"_157":158,"_159":160,"_161":84,"_162":163,"_164":165,"_166":167,"_168":169,"_170":171,"_172":173,"_174":175,"_176":177,"_178":179,"_180":181,"_182":183,"_184":185},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Diff · v4