ATN — Agentic Trading Network

\n \n \n \n \n \n \n

Cognitive Resolution Cycle

\n 1.Observe\\u2192\n 2.Hypothesis\\u2192\n 3.Proof\\u2192\n 4.Contradiction\\u2192\n 5.Invalidation\\u2192\n 6.Resolution\\u2192\n 7.Memory\\u2192\n 8.Reuse\n

Recent MarketCases

Recent Pipeline Runs

All MarketCases

\n \n \n \n \n

Pipeline Control

Manually trigger the CNOS 8-stage cycle for all configured symbols.

\n \n

Pipeline Run History

Telegram Notifications

Enable Telegram notifications

Bot Token

Chat ID

\n \n

Twelve Data API (Multiple Keys with Rotation)

Add multiple free keys for automatic rotation. Free tier: 8 req/min, 800/day per key. N keys = N\\u00d7 limits. Get keys at twelvedata.com

Enable Twelve Data as primary data source

Data Source Mode

API Keys (one per line)

\n \n \n \n

Pipeline Configuration

Enable automated pipeline

Symbols (comma-separated)

Timeframe

Interval (minutes)

Desk Configuration

Desk ID

Min Confidence for Case

Auto-generate MarketCases from observations

Auto-resolve cases (CONFIRMED/REJECTED)

Contradictions required for refinement

Notification Preferences

Notify on new MarketCase

Notify on case resolution

Notify on contradiction detected

Notify on errors

Market Data Tools (MCP-style)

search_symbols

Search and resolve symbols across exchanges

get_quote

Real-time quote snapshot

get_technicals

Technical indicators (RSI, MACD, EMAs, Bollinger, ATR, etc.)

\n \n

get_technicals_rating

Aggregated buy/sell/neutral rating with breakdown

\n \n

analyze_multi_timeframe

Trend alignment across 15m, 1H, 4H, 1D

compute_levels

Support/resistance, swing levels, FVGs, premium/discount

\n \n

get_news

get_ohlcv

Historical OHLCV bars

\n \n

Resolved Memory Search

Search verified, attributable, reusable knowledge in the network memory.

\n\n\n\n`;\n}\n",[],[239,273],"eea8db28-62d6-4b17-b332-5bb98ead0d95","cnos.ts","e58f2dba-8b1a-11f1-aa6c-1607ee4eb77e",7,["D",1785308306186],"// ─── CNOS CORE: 8-Stage Cognitive Resolution Cycle ───────────────────\n// The missing cognitive layer (Layer 6) that completes the stack.\n// Observation → Hypothesis → Proof → Contradiction → Invalidation\n// → Resolution → Memory → Reuse\n\nimport {\n createObservation,\n createMarketCase,\n updateCaseResolution,\n addContradiction,\n incrementReuse,\n getStats,\n logPipelineRun,\n getRecentCases,\n getCasesByResolution,\n searchMemory,\n initDB,\n uuid,\n} from \"./db.ts\";\n\nimport { getOHLCV, getTechnicals, computeLevels } from \"./market-data.ts\";\n\nimport { runDesk, type ObservationResult } from \"./agents.ts\";\n\nimport {\n getSettings,\n type ATNSettings,\n} from \"./settings.ts\";\n\nimport {\n notifyNewCase,\n notifyResolution,\n notifyContradiction,\n notifyPipelineRun,\n notifyError,\n} from \"./telegram.ts\";\n\n// ─── Stage 1: Observation ────────────────────────────────────────────\n// Ingest raw agent outputs from a private desk. Strip private data.\nexport async function stage1_observation(\n deskId: string,\n symbol: string,\n timeframe: string,\n): Promise {\n const { observations } = await runDesk(deskId, symbol, timeframe);\n // Privacy firewall: execution agent output is NEVER included\n // observations only contains shared agents (macro, structure, liquidity, risk, memory)\n await logPipelineRun(\"OBSERVATION\", \"COMPLETE\", `Desk ${deskId} observed ${symbol} on ${timeframe}`, observations.length);\n return observations;\n}\n\n// ─── Stage 2: Structured Observation / Hypothesis ───────────────────\n// Convert the strongest observation into a testable MarketCase\nexport async function stage2_hypothesis(\n observations: ObservationResult[],\n deskId: string,\n symbol: string,\n timeframe: string,\n settings: ATNSettings,\n): Promise {\n const valid = observations\n .filter((o) => o.hypothesis && o.confidence >= settings.minConfidenceForCase)\n .sort((a, b) => b.confidence - a.confidence);\n\n if (valid.length === 0) {\n await logPipelineRun(\"HYPOTHESIS\", \"SKIP\", `No observations met minimum confidence ${settings.minConfidenceForCase}`, 0);\n return null;\n }\n\n const best = valid[0];\n const caseId = await createMarketCase({\n desk_id: deskId,\n symbol,\n timeframe,\n hypothesis: best.hypothesis!,\n expected_outcome: best.expected_outcome,\n reasoning: best.summary,\n invalidation_level: best.invalidation_level,\n conviction: best.confidence,\n time_horizon: timeframe,\n meta_tags: best.meta_tags,\n evidence: observations.filter((o) => o.observation_id).map((o) => o.observation_id),\n });\n\n await logPipelineRun(\"HYPOTHESIS\", \"COMPLETE\", `Created MarketCase for ${symbol} conviction ${best.confidence}`, 1);\n\n // Telegram notification\n await notifyNewCase({\n symbol,\n timeframe,\n hypothesis: best.hypothesis,\n conviction: best.confidence,\n resolution: \"PENDING\",\n meta_tags: best.meta_tags,\n });\n\n return caseId;\n}\n\n// ─── Stage 3: Proof ──────────────────────────────────────────────────\n// Automated backtesting — verify the hypothesis against historical data\nexport async function stage3_proof(\n caseId: string,\n symbol: string,\n timeframe: string,\n): Promise<{ passed: boolean; details: any }> {\n const ohlcv = await getOHLCV(symbol, timeframe, 50);\n if (ohlcv.error) {\n await logPipelineRun(\"PROOF\", \"FAIL\", `Could not fetch data for ${symbol}`, 0);\n return { passed: false, details: { error: ohlcv.error } };\n }\n\n const bars = ohlcv.bars;\n const last20 = bars.slice(-20);\n const closes = last20.map((b: any) => b.close);\n\n // Simulated proof: check if recent price action supports the hypothesis\n const recentChange = ((closes[closes.length - 1] - closes[0]) / closes[0]) * 100;\n const avgVolume = last20.reduce((a: number, b: any) => a + b.volume, 0) / last20.length;\n const lastVolume = bars[bars.length - 1].volume;\n const volumeSurge = lastVolume > avgVolume * 1.5;\n\n // Win rate proxy: how many of the last 10 bars closed green\n const last10 = bars.slice(-10);\n const greenBars = last10.filter((b: any) => b.close > b.open).length;\n const winRate = (greenBars / last10.length) * 100;\n\n // Profit factor proxy\n const gains = last10.filter((b: any) => b.close > b.open).reduce((a: number, b: any) => a + (b.close - b.open), 0);\n const losses = last10.filter((b: any) => b.close <= b.open).reduce((a: number, b: any) => a + Math.abs(b.close - b.open), 0);\n const profitFactor = losses > 0 ? gains / losses : gains > 0 ? 3 : 1;\n\n // Max drawdown in last 20 bars\n let peak = closes[0];\n let maxDD = 0;\n for (const c of closes) {\n if (c > peak) peak = c;\n const dd = ((peak - c) / peak) * 100;\n if (dd > maxDD) maxDD = dd;\n }\n\n const proofArtifact = {\n recentChange: recentChange.toFixed(2) + \"%\",\n winRate: winRate.toFixed(1) + \"%\",\n profitFactor: profitFactor.toFixed(2),\n maxDrawdown: maxDD.toFixed(2) + \"%\",\n volumeSurge,\n barsAnalyzed: last20.length,\n };\n\n const passed = winRate >= 50 && profitFactor >= 1.0 && maxDD < 15;\n\n await logPipelineRun(\"PROOF\", passed ? \"PASS\" : \"FAIL\",\n `${symbol}: WR ${winRate.toFixed(0)}% PF ${profitFactor.toFixed(2)} DD ${maxDD.toFixed(1)}%`, 1);\n\n // Store proof artifact on the case\n const { sqlite } = await import(\"https://esm.town/v/std/sqlite/main.ts\");\n await sqlite.execute({\n sql: \"UPDATE market_cases SET proof_artifact = ? WHERE id = ?\",\n args: [JSON.stringify(proofArtifact), caseId],\n });\n\n return { passed, details: proofArtifact };\n}\n\n// ─── Stage 4: Contradiction ──────────────────────────────────────────\n// Compare hypothesis against other cases in the network\nexport async function stage4_contradiction(\n caseId: string,\n symbol: string,\n hypothesis: string,\n settings: ATNSettings,\n): Promise<{ contradictionsFound: number; contradictions: any[] }> {\n // Find other pending or resolved cases for the same symbol\n const { sqlite } = await import(\"https://esm.town/v/std/sqlite/main.ts\");\n const result = await sqlite.execute({\n sql: `SELECT * FROM market_cases WHERE symbol = ? AND id != ? ORDER BY created_at DESC LIMIT 10`,\n args: [symbol, caseId],\n });\n\n const otherCases = result.rows as any[];\n const contradictions: any[] = [];\n\n for (const other of otherCases) {\n const otherHypo = (other.hypothesis as string) ?? \"\";\n const otherResolution = (other.resolution as string) ?? \"PENDING\";\n\n // Check for directional contradiction\n const thisBullish = hypothesis.toLowerCase().includes(\"upside\") ||\n hypothesis.toLowerCase().includes(\"bullish\") ||\n hypothesis.toLowerCase().includes(\"continuation toward resistance\") ||\n hypothesis.toLowerCase().includes(\"risk-on\");\n const otherBearish = otherHypo.toLowerCase().includes(\"downside\") ||\n otherHypo.toLowerCase().includes(\"bearish\") ||\n otherHypo.toLowerCase().includes(\"continuation toward support\") ||\n otherHypo.toLowerCase().includes(\"risk-off\");\n\n const thisBearish = hypothesis.toLowerCase().includes(\"downside\") ||\n hypothesis.toLowerCase().includes(\"bearish\") ||\n hypothesis.toLowerCase().includes(\"continuation toward support\") ||\n hypothesis.toLowerCase().includes(\"risk-off\");\n const otherBullish = otherHypo.toLowerCase().includes(\"upside\") ||\n otherHypo.toLowerCase().includes(\"bullish\") ||\n otherHypo.toLowerCase().includes(\"continuation toward resistance\") ||\n otherHypo.toLowerCase().includes(\"risk-on\");\n\n if ((thisBullish && otherBearish) || (thisBearish && otherBullish)) {\n const contradictionText = `New ${thisBullish ? \"bullish\" : \"bearish\"} hypothesis on ${symbol} contradicts ${otherBullish ? \"bullish\" : \"bearish\"} case ${other.id.substring(0, 8)} (${otherResolution})`;\n const cId = await addContradiction(caseId, \"desk_beta_02\", contradictionText, other.id);\n contradictions.push({ id: cId, text: contradictionText, againstCase: other.id });\n }\n\n // Check if contradicts a CONFIRMED case\n if (otherResolution === \"CONFIRMED\" && (thisBullish !== otherBullish)) {\n const contradictionText = `New hypothesis on ${symbol} contradicts a CONFIRMED case ${other.id.substring(0, 8)}`;\n const cId = await addContradiction(caseId, \"desk_gamma_03\", contradictionText, other.id);\n contradictions.push({ id: cId, text: contradictionText, againstCase: other.id });\n }\n }\n\n await logPipelineRun(\"CONTRADICTION\", contradictions.length > 0 ? \"FLAGGED\" : \"CLEAR\",\n `${symbol}: ${contradictions.length} contradictions found`, contradictions.length);\n\n // Notify contradictions\n for (const c of contradictions) {\n await notifyContradiction({\n caseSymbol: symbol,\n contradictionText: c.text,\n challengingDesk: \"network\",\n });\n }\n\n return { contradictionsFound: contradictions.length, contradictions };\n}\n\n// ─── Stage 5: Invalidation ───────────────────────────────────────────\n// Check if the hypothesis has been invalidated by current market data\nexport async function stage5_invalidation(\n caseId: string,\n symbol: string,\n timeframe: string,\n invalidationLevel: string | null,\n): Promise<{ invalidated: boolean; reason: string }> {\n if (!invalidationLevel) {\n return { invalidated: false, reason: \"No invalidation level set\" };\n }\n\n const ohlcv = await getOHLCV(symbol, timeframe, 5);\n if (ohlcv.error) {\n return { invalidated: false, reason: \"Could not fetch data for invalidation check\" };\n }\n\n const lastBar = ohlcv.bars[ohlcv.bars.length - 1];\n const lastPrice = lastBar.close;\n\n // Parse invalidation level from text (extract number)\n const numMatch = invalidationLevel.match(/[\\d,.]+/);\n if (!numMatch) {\n // Non-numeric invalidation (e.g., \"VIX spikes above 30\") — can't auto-check\n return { invalidated: false, reason: `Manual check needed: ${invalidationLevel}` };\n }\n\n const level = parseFloat(numMatch[0].replace(/,/g, \"\"));\n const lower = invalidationLevel.toLowerCase();\n\n let invalidated = false;\n let reason = \"\";\n\n if (lower.includes(\"below\")) {\n if (lastPrice < level) {\n invalidated = true;\n reason = `Price ${lastPrice.toFixed(2)} dropped below invalidation level ${level}`;\n }\n } else if (lower.includes(\"above\")) {\n if (lastPrice > level) {\n invalidated = true;\n reason = `Price ${lastPrice.toFixed(2)} exceeded invalidation level ${level}`;\n }\n } else {\n // Generic: check if price is near the level\n const distance = Math.abs(lastPrice - level) / level * 100;\n if (distance < 1) {\n invalidated = true;\n reason = `Price ${lastPrice.toFixed(2)} within 1% of invalidation level ${level}`;\n }\n }\n\n await logPipelineRun(\"INVALIDATION\", invalidated ? \"INVALIDATED\" : \"VALID\",\n `${symbol}: ${invalidated ? \"INVALIDATED — \" + reason : \"Still valid\"}`, 1);\n\n return { invalidated, reason };\n}\n\n// ─── Stage 6: Resolution ─────────────────────────────────────────────\n// Combine proof + contradiction + invalidation into a final resolution\nexport async function stage6_resolution(\n caseId: string,\n symbol: string,\n proofPassed: boolean,\n contradictionsFound: number,\n invalidated: boolean,\n invalidationReason: string,\n settings: ATNSettings,\n): Promise<{ resolution: string; reason: string }> {\n let resolution = \"PENDING\";\n let reason = \"\";\n\n if (invalidated) {\n resolution = \"REJECTED\";\n reason = `Invalidated: ${invalidationReason}`;\n } else if (!proofPassed) {\n resolution = \"REJECTED\";\n reason = \"Proof stage failed — insufficient statistical evidence (win rate, profit factor, or drawdown)\";\n } else if (contradictionsFound >= settings.contradictionsRequired) {\n // Has enough contradictions to require refinement\n resolution = \"REFINED\";\n reason = `${contradictionsFound} contradictions found — hypothesis needs refinement before confirmation`;\n } else if (proofPassed && contradictionsFound < settings.contradictionsRequired) {\n resolution = \"CONFIRMED\";\n reason = `Passed proof stage with ${contradictionsFound} contradictions (threshold: ${settings.contradictionsRequired})`;\n }\n\n if (!settings.autoResolve && resolution !== \"REJECTED\") {\n resolution = \"PENDING\";\n reason = \"Auto-resolve disabled — awaiting manual review\";\n }\n\n await updateCaseResolution(caseId, resolution, reason);\n await logPipelineRun(\"RESOLUTION\", resolution, `${symbol}: ${resolution} — ${reason}`, 1);\n\n // Notify\n await notifyResolution({\n symbol,\n resolution,\n resolution_reason: reason,\n contradictions_resolved: contradictionsFound,\n });\n\n return { resolution, reason };\n}\n\n// ─── Stage 7: Resolved Memory ────────────────────────────────────────\n// Verified cases enter the shared memory store\nexport async function stage7_memory(caseId: string, resolution: string): Promise {\n if (resolution === \"CONFIRMED\" || resolution === \"REJECTED\") {\n // Memory entry is already created in updateCaseResolution\n await logPipelineRun(\"MEMORY\", \"STORED\", `Case ${caseId.substring(0, 8)} entered resolved memory as ${resolution}`, 1);\n } else {\n await logPipelineRun(\"MEMORY\", \"SKIP\", `Case ${caseId.substring(0, 8)} not resolved — stays pending`, 0);\n }\n}\n\n// ─── Stage 8: Reuse ──────────────────────────────────────────────────\n// Check if any new observations match existing resolved memory\nexport async function stage8_reuse(symbol: string): Promise<{ matches: number }> {\n const matches = await searchMemory(symbol, 5);\n for (const m of matches) {\n await incrementReuse(m.id);\n }\n await logPipelineRun(\"REUSE\", \"COMPLETE\", `${symbol}: ${matches.length} memory matches found and reuse count incremented`, matches.length);\n return { matches: matches.length };\n}\n\n// ─── Full CNOS Cycle ─────────────────────────────────────────────────\n// Runs all 8 stages for a single symbol\nexport async function runCNOSCycle(\n deskId: string,\n symbol: string,\n timeframe: string,\n): Promise<{\n symbol: string;\n caseId: string | null;\n resolution: string;\n proofPassed: boolean;\n contradictions: number;\n stages: string[];\n}> {\n const settings = await getSettings();\n const stages: string[] = [];\n\n // Stage 1: Observation\n stages.push(\"OBSERVATION\");\n const observations = await stage1_observation(deskId, symbol, timeframe);\n\n // Stage 2: Hypothesis\n stages.push(\"HYPOTHESIS\");\n const caseId = await stage2_hypothesis(observations, deskId, symbol, timeframe, settings);\n if (!caseId) {\n return { symbol, caseId: null, resolution: \"SKIPPED\", proofPassed: false, contradictions: 0, stages };\n }\n\n // Stage 3: Proof\n stages.push(\"PROOF\");\n const proof = await stage3_proof(caseId, symbol, timeframe);\n\n // Stage 4: Contradiction\n stages.push(\"CONTRADICTION\");\n const { sqlite } = await import(\"https://esm.town/v/std/sqlite/main.ts\");\n const caseRow = await sqlite.execute({ sql: \"SELECT * FROM market_cases WHERE id = ?\", args: [caseId] });\n const hypothesis = (caseRow.rows[0] as any)?.hypothesis ?? \"\";\n const invalidationLevel = (caseRow.rows[0] as any)?.invalidation_level ?? null;\n const contradictionResult = await stage4_contradiction(caseId, symbol, hypothesis, settings);\n\n // Stage 5: Invalidation\n stages.push(\"INVALIDATION\");\n const invalidationResult = await stage5_invalidation(caseId, symbol, timeframe, invalidationLevel);\n\n // Stage 6: Resolution\n stages.push(\"RESOLUTION\");\n const resolutionResult = await stage6_resolution(\n caseId, symbol,\n proof.passed,\n contradictionResult.contradictionsFound,\n invalidationResult.invalidated,\n invalidationResult.reason,\n settings,\n );\n\n // Stage 7: Memory\n stages.push(\"MEMORY\");\n await stage7_memory(caseId, resolutionResult.resolution);\n\n // Stage 8: Reuse\n stages.push(\"REUSE\");\n await stage8_reuse(symbol);\n\n await notifyPipelineRun(\"FULL_CYCLE\", 1, `${symbol} completed all 8 stages → ${resolutionResult.resolution}`);\n\n return {\n symbol,\n caseId,\n resolution: resolutionResult.resolution,\n proofPassed: proof.passed,\n contradictions: contradictionResult.contradictionsFound,\n stages,\n };\n}\n\n// ─── Run Pipeline for All Configured Symbols ─────────────────────────\nexport async function runPipeline(): Promise<{\n processed: number;\n results: any[];\n errors: string[];\n}> {\n await initDB();\n const settings = await getSettings();\n const results: any[] = [];\n const errors: string[] = [];\n\n if (!settings.pipelineEnabled) {\n return { processed: 0, results: [], errors: [\"Pipeline is disabled in settings\"] };\n }\n\n for (const symbol of settings.pipelineSymbols) {\n try {\n const result = await runCNOSCycle(settings.deskId, symbol, settings.pipelineTimeframe);\n results.push(result);\n } catch (e: any) {\n errors.push(`${symbol}: ${e.message}`);\n await notifyError(`${symbol} pipeline error: ${e.message}`);\n }\n }\n\n await notifyPipelineRun(\"PIPELINE_BATCH\", results.length, `Processed ${results.length} symbols, ${errors.length} errors`);\n\n return { processed: results.length, results, errors };\n}\n",[],[239,280],"d39cd854-4396-41b7-a9ed-7a9385e451c6","agents.ts","104b7258-8b1a-11f1-bd5c-1607ee4eb77e",["D",1785307948387],"// ─── Private Agentic Desk Agents ─────────────────────────────────────\n// Each desk runs 6 specialized agents. Agents produce structured\n// observations that are shared with the network. Execution stays private.\n\nimport {\n getQuote,\n getTechnicals,\n getTechnicalsRating,\n analyzeMultiTimeframe,\n computeLevels,\n getOHLCV,\n getNews,\n} from \"./market-data.ts\";\n\nimport { createObservation, createMarketCase, searchMemory } from \"./db.ts\";\n\nexport type AgentType =\n | \"macro\"\n | \"structure\"\n | \"liquidity\"\n | \"risk\"\n | \"memory\"\n | \"execution\";\n\nexport interface ObservationResult {\n observation_id: string;\n agent_type: AgentType;\n symbol: string;\n timeframe: string;\n observation_type: string;\n summary: string;\n data: any;\n confidence: number;\n hypothesis?: string;\n expected_outcome?: string;\n invalidation_level?: string;\n meta_tags?: string[];\n}\n\n// ─── Macro Agent ─────────────────────────────────────────────────────\n// Economic regime, cross-asset flows, macro context\nexport async function runMacroAgent(\n deskId: string,\n symbol: string,\n): Promise {\n const mtf = await analyzeMultiTimeframe(symbol);\n const news = await getNews(symbol);\n\n // Macro context from cross-asset\n const spyQuote = await getQuote(\"SPY\");\n const vixQuote = await getQuote(\"VIX\");\n const dxyQuote = await getQuote(\"DXY\");\n\n const riskRegime = vixQuote?.price > 25 ? \"RISK-OFF\" : vixQuote?.price > 18 ? \"NEUTRAL\" : \"RISK-ON\";\n const dxyTrend = dxyQuote?.changePercent > 0 ? \"DXY_RISING\" : \"DXY_FALLING\";\n\n const observation = {\n riskRegime,\n dxyTrend,\n vix: vixQuote?.price,\n spyChange: spyQuote?.changePercent,\n multiTimeframeBias: mtf.bias,\n alignmentScore: mtf.alignmentScore,\n newsCount: Array.isArray(news) ? news.length : 0,\n topNews: Array.isArray(news) ? news.slice(0, 3).map((n: any) => n.title) : [],\n };\n\n const obsId = await createObservation({\n desk_id: deskId,\n agent_type: \"macro\",\n symbol,\n timeframe: \"1D\",\n observation_type: \"macro_context\",\n data_payload: observation,\n confidence: 65,\n provenance: \"macro_agent\",\n });\n\n const hypothesis = `IF ${riskRegime} regime persists with VIX at ${vixQuote?.price?.toFixed(1)} THEN ${symbol} likely ${mtf.bias === \"BULLISH\" ? \"continues upside\" : \"faces headwind\"} BECAUSE cross-asset flows favor ${riskRegime === \"RISK-ON\" ? \"risk assets\" : \"safe haven\"}`;\n const invalidation = riskRegime === \"RISK-OFF\" ? \"VIX drops below 18\" : \"VIX spikes above 30\";\n\n return {\n observation_id: obsId,\n agent_type: \"macro\",\n symbol,\n timeframe: \"1D\",\n observation_type: \"macro_context\",\n summary: `${riskRegime} regime. ${mtf.bias} bias on ${symbol}. VIX: ${vixQuote?.price?.toFixed(1)}`,\n data: observation,\n confidence: 65,\n hypothesis,\n expected_outcome: mtf.bias === \"BULLISH\" ? \"Price appreciation\" : \"Price decline or consolidation\",\n invalidation_level: invalidation,\n meta_tags: [symbol, \"macro\", riskRegime, dxyTrend, mtf.bias],\n };\n}\n\n// ─── Structure Agent ─────────────────────────────────────────────────\n// Price action, SMC, swing analysis, multi-timeframe alignment\nexport async function runStructureAgent(\n deskId: string,\n symbol: string,\n timeframe: string = \"4H\",\n): Promise {\n const tech = await getTechnicals(symbol, timeframe);\n const levels = await computeLevels(symbol, timeframe);\n const rating = await getTechnicalsRating(symbol, timeframe);\n\n if (tech.error) {\n return {\n observation_id: \"\",\n agent_type: \"structure\",\n symbol,\n timeframe,\n observation_type: \"structure_analysis\",\n summary: `Error: ${tech.error}`,\n data: {},\n confidence: 0,\n };\n }\n\n const i = tech.indicators;\n const trend = i.ema.ema20 > i.ema.ema50 ? \"UPTREND\" : i.ema.ema20 < i.ema.ema50 ? \"DOWNTREND\" : \"RANGE\";\n const rsiSignal = i.rsi < 30 ? \"OVERSOLD\" : i.rsi > 70 ? \"OVERBOUGHT\" : \"NEUTRAL\";\n const macdSignal = i.macd?.histogram > 0 ? \"BULLISH\" : \"BEARISH\";\n const bbPosition = tech.lastPrice > i.bollinger?.upper ? \"UPPER_BAND\" :\n tech.lastPrice < i.bollinger?.lower ? \"LOWER_BAND\" : \"MID_RANGE\";\n\n const structure = {\n trend,\n rsi: i.rsi,\n rsiSignal,\n macdSignal,\n macdHistogram: i.macd?.histogram,\n bollingerPosition: bbPosition,\n ema20: i.ema.ema20,\n ema50: i.ema.ema50,\n ema200: i.ema.ema200,\n sma200: i.sma.sma200,\n adx: i.adx,\n rating: rating.rating,\n ratingScore: rating.score,\n support: levels.supportLevels,\n resistance: levels.resistanceLevels,\n swingHighs: levels.swingHighs,\n swingLows: levels.swingLows,\n equilibrium: levels.range?.equilibrium,\n fairValueGaps: levels.fairValueGaps,\n };\n\n const obsId = await createObservation({\n desk_id: deskId,\n agent_type: \"structure\",\n symbol,\n timeframe,\n observation_type: \"structure_analysis\",\n data_payload: structure,\n confidence: Math.min(90, rating.score + 10),\n provenance: \"structure_agent\",\n });\n\n const hypothesis = `IF ${symbol} ${trend} with RSI ${rsiSignal} (${i.rsi?.toFixed(1)}) and MACD ${macdSignal} THEN expect ${trend === \"UPTREND\" ? \"continuation toward resistance\" : trend === \"DOWNTREND\" ? \"continuation toward support\" : \"range-bound movement\"} BECAUSE structure confirms ${trend} with ADX ${i.adx?.toFixed(1)}`;\n const invalidation = trend === \"UPTREND\"\n ? `Close below ${levels.supportLevels?.[0]?.toFixed(2)} on ${timeframe}`\n : `Close above ${levels.resistanceLevels?.[0]?.toFixed(2)} on ${timeframe}`;\n\n return {\n observation_id: obsId,\n agent_type: \"structure\",\n symbol,\n timeframe,\n observation_type: \"structure_analysis\",\n summary: `${trend} | RSI ${i.rsi?.toFixed(1)} ${rsiSignal} | MACD ${macdSignal} | Rating ${rating.rating} (${rating.score})`,\n data: structure,\n confidence: Math.min(90, rating.score + 10),\n hypothesis,\n expected_outcome: trend === \"UPTREND\" ? \"Move to resistance\" : trend === \"DOWNTREND\" ? \"Move to support\" : \"Range continuation\",\n invalidation_level: invalidation,\n meta_tags: [symbol, timeframe, trend, rsiSignal, macdSignal, `rating_${rating.rating}`],\n };\n}\n\n// ─── Liquidity Agent ─────────────────────────────────────────────────\n// Volume profiling, order flow, FVG identification, premium/discount\nexport async function runLiquidityAgent(\n deskId: string,\n symbol: string,\n timeframe: string = \"4H\",\n): Promise {\n const ohlcv = await getOHLCV(symbol, timeframe, 50);\n const levels = await computeLevels(symbol, timeframe);\n\n if (ohlcv.error) {\n return {\n observation_id: \"\",\n agent_type: \"liquidity\",\n symbol,\n timeframe,\n observation_type: \"liquidity_analysis\",\n summary: `Error: ${ohlcv.error}`,\n data: {},\n confidence: 0,\n };\n }\n\n const bars = ohlcv.bars;\n const volumes = bars.map((b: any) => b.volume);\n const avgVolume = volumes.reduce((a: number, b: number) => a + b, 0) / volumes.length;\n const lastVolume = volumes[volumes.length - 1];\n const volumeRatio = lastVolume / avgVolume;\n\n // Volume-weighted price level (VWAP approximation)\n let pvSum = 0, volSum = 0;\n for (const b of bars) {\n pvSum += b.close * b.volume;\n volSum += b.volume;\n }\n const vwap = pvSum / volSum;\n const lastPrice = bars[bars.length - 1].close;\n const vwapDeviation = ((lastPrice - vwap) / vwap) * 100;\n\n // Liquidity zones from FVGs\n const liquidityZones = levels.fairValueGaps?.map((fvg: any) => ({\n top: fvg.top,\n bottom: fvg.bottom,\n type: \"FVG\",\n })) ?? [];\n\n // Volume profile buckets (simplified)\n const profileBuckets: Record = {};\n for (const b of bars) {\n const key = b.close.toFixed(2);\n profileBuckets[key] = (profileBuckets[key] || 0) + b.volume;\n }\n const pocLevel = Object.entries(profileBuckets).sort((a, b) => b[1] - a[1])[0]?.[0] ?? \"N/A\";\n\n const liquidity = {\n vwap,\n vwapDeviation: vwapDeviation.toFixed(2) + \"%\",\n volumeRatio: volumeRatio.toFixed(2),\n avgVolume,\n lastVolume,\n pointOfControl: pocLevel,\n equilibrium: levels.range?.equilibrium,\n premiumDiscount: levels.range?.equilibrium,\n liquidityZones,\n supportLevels: levels.supportLevels,\n resistanceLevels: levels.resistanceLevels,\n };\n\n const obsId = await createObservation({\n desk_id: deskId,\n agent_type: \"liquidity\",\n symbol,\n timeframe,\n observation_type: \"liquidity_analysis\",\n data_payload: liquidity,\n confidence: 60,\n provenance: \"liquidity_agent\",\n });\n\n const deviation = parseFloat(vwapDeviation.toFixed(2));\n const hypothesis = `IF ${symbol} is ${levels.range?.equilibrium} (${deviation > 0 ? \"+\" : \"\"}${deviation}% from VWAP) with volume ratio ${volumeRatio.toFixed(2)}x THEN expect ${levels.range?.equilibrium === \"PREMIUM\" ? \"mean reversion toward VWAP\" : \"bounce from discount zone\"} BECAUSE price is ${levels.range?.equilibrium === \"PREMIUM\" ? \"above\" : \"below\"} value area`;\n const invalidation = levels.range?.equilibrium === \"PREMIUM\"\n ? `Sustained close above ${levels.resistanceLevels?.[0]?.toFixed(2)}`\n : `Sustained close below ${levels.supportLevels?.[0]?.toFixed(2)}`;\n\n return {\n observation_id: obsId,\n agent_type: \"liquidity\",\n symbol,\n timeframe,\n observation_type: \"liquidity_analysis\",\n summary: `${levels.range?.equilibrium} | VWAP dev: ${deviation > 0 ? \"+\" : \"\"}${deviation}% | Vol ratio: ${volumeRatio.toFixed(2)}x | POC: ${pocLevel}`,\n data: liquidity,\n confidence: 60,\n hypothesis,\n expected_outcome: levels.range?.equilibrium === \"PREMIUM\" ? \"Reversion to VWAP\" : \"Bounce from discount\",\n invalidation_level: invalidation,\n meta_tags: [symbol, timeframe, levels.range?.equilibrium, `vol_${volumeRatio > 1.5 ? \"high\" : \"normal\"}`],\n };\n}\n\n// ─── Risk Agent (PRIVATE — observations only, no execution shared) ───\nexport async function runRiskAgent(\n deskId: string,\n symbol: string,\n timeframe: string = \"4H\",\n): Promise {\n const tech = await getTechnicals(symbol, timeframe);\n if (tech.error) {\n return {\n observation_id: \"\",\n agent_type: \"risk\",\n symbol,\n timeframe,\n observation_type: \"risk_assessment\",\n summary: `Error: ${tech.error}`,\n data: {},\n confidence: 0,\n };\n }\n\n const atr = tech.indicators.atr ?? 0;\n const lastPrice = tech.lastPrice;\n const atrPercent = (atr / lastPrice) * 100;\n\n // Risk levels based on ATR\n const suggestedStop = atr * 1.5;\n const suggestedTarget = atr * 3;\n const riskReward = suggestedTarget / suggestedStop;\n\n // Correlation risk flag\n const adx = tech.indicators.adx ?? 0;\n const trendStrength = adx > 25 ? \"STRONG\" : adx > 20 ? \"MODERATE\" : \"WEAK\";\n\n // Risk score (higher = riskier)\n let riskScore = 0;\n if (atrPercent > 3) riskScore += 30;\n else if (atrPercent > 2) riskScore += 20;\n else riskScore += 10;\n if (adx < 20) riskScore += 20;\n if (tech.indicators.rsi > 70 || tech.indicators.rsi < 30) riskScore += 15;\n\n const riskLevel = riskScore > 50 ? \"HIGH\" : riskScore > 30 ? \"MODERATE\" : \"LOW\";\n\n const riskData = {\n atr,\n atrPercent: atrPercent.toFixed(2) + \"%\",\n suggestedStopDistance: suggestedStop.toFixed(2),\n suggestedTargetDistance: suggestedTarget.toFixed(2),\n riskRewardRatio: riskReward.toFixed(2),\n trendStrength,\n adx,\n rsi: tech.indicators.rsi,\n riskScore,\n riskLevel,\n };\n\n const obsId = await createObservation({\n desk_id: deskId,\n agent_type: \"risk\",\n symbol,\n timeframe,\n observation_type: \"risk_assessment\",\n data_payload: riskData,\n confidence: 55,\n privacy_class: \"shared\", // only risk metrics shared, not position sizing\n });\n\n return {\n observation_id: obsId,\n agent_type: \"risk\",\n symbol,\n timeframe,\n observation_type: \"risk_assessment\",\n summary: `Risk: ${riskLevel} | ATR: ${atrPercent.toFixed(2)}% | RR: ${riskReward.toFixed(2)} | Trend: ${trendStrength}`,\n data: riskData,\n confidence: 55,\n hypothesis: `IF ${symbol} has ${riskLevel} risk (ATR ${atrPercent.toFixed(2)}%) THEN position sizing should be ${riskLevel === \"HIGH\" ? \"reduced\" : \"standard\"} BECAUSE volatility and trend strength dictate exposure`,\n expected_outcome: riskLevel === \"HIGH\" ? \"Wider stops needed, smaller position\" : \"Standard risk parameters applicable\",\n invalidation_level: `ATR exceeds ${(atrPercent * 1.5).toFixed(2)}%`,\n meta_tags: [symbol, timeframe, `risk_${riskLevel}`, `atr_${atrPercent.toFixed(1)}%`],\n };\n}\n\n// ─── Memory Agent ────────────────────────────────────────────────────\n// RAG over desk's local memory + network resolved memory\nexport async function runMemoryAgent(\n deskId: string,\n symbol: string,\n): Promise {\n // Search resolved memory for similar cases\n const similar = await searchMemory(symbol, 5);\n\n const memoryData = {\n similarCasesFound: similar.length,\n cases: similar.map((c) => ({\n case_id: c.id,\n symbol: c.symbol,\n hypothesis: c.hypothesis,\n resolution: c.resolution,\n conviction: c.conviction,\n reuse_count: c.reuse_count,\n tags: c.meta_tags,\n })),\n topResolution: similar[0]?.resolution ?? \"NONE\",\n avgConviction: similar.length > 0\n ? Math.round(similar.reduce((a, c) => a + (c.conviction ?? 0), 0) / similar.length)\n : 0,\n };\n\n const obsId = await createObservation({\n desk_id: deskId,\n agent_type: \"memory\",\n symbol,\n timeframe: \"1D\",\n observation_type: \"memory_retrieval\",\n data_payload: memoryData,\n confidence: similar.length > 0 ? 70 : 40,\n provenance: \"memory_agent\",\n });\n\n const hasHistory = similar.length > 0;\n const hypothesis = hasHistory\n ? `IF ${symbol} has ${similar.length} resolved cases with avg conviction ${memoryData.avgConviction} THEN current analysis benefits from historical context BECAUSE resolved memory contains verified patterns`\n : `IF ${symbol} has no resolved memory THEN this observation establishes a new memory baseline BECAUSE first cases seed the network`;\n\n return {\n observation_id: obsId,\n agent_type: \"memory\",\n symbol,\n timeframe: \"1D\",\n observation_type: \"memory_retrieval\",\n summary: hasHistory\n ? `${similar.length} similar cases found | Avg conviction: ${memoryData.avgConviction} | Top resolution: ${memoryData.topResolution}`\n : \"No prior memory — establishing baseline\",\n data: memoryData,\n confidence: similar.length > 0 ? 70 : 40,\n hypothesis,\n expected_outcome: hasHistory ? \"Pattern-informed analysis\" : \"New memory entry created\",\n invalidation_level: \"N/A\",\n meta_tags: [symbol, \"memory\", hasHistory ? \"has_history\" : \"new_baseline\"],\n };\n}\n\n// ─── Execution Agent (PRIVATE — never shared) ────────────────────────\n// Produces only a private signal, no network egress\nexport async function runExecutionAgent(\n deskId: string,\n symbol: string,\n observations: ObservationResult[],\n): Promise<{ action: string; reasoning: string; private: boolean }> {\n const macro = observations.find((o) => o.agent_type === \"macro\");\n const structure = observations.find((o) => o.agent_type === \"structure\");\n const liquidity = observations.find((o) => o.agent_type === \"liquidity\");\n const risk = observations.find((o) => o.agent_type === \"risk\");\n\n const bullishCount = [macro, structure, liquidity].filter((o) =>\n o?.hypothesis?.toLowerCase().includes(\"upside\") ||\n o?.hypothesis?.toLowerCase().includes(\"bullish\") ||\n o?.hypothesis?.toLowerCase().includes(\"continuation toward resistance\") ||\n o?.hypothesis?.toLowerCase().includes(\"risk-on\")\n ).length;\n\n const riskLevel = risk?.data?.riskLevel ?? \"MODERATE\";\n\n let action = \"HOLD\";\n if (bullishCount >= 2 && riskLevel !== \"HIGH\") action = \"LONG_SIGNAL\";\n else if (bullishCount === 0 && riskLevel !== \"HIGH\") action = \"SHORT_SIGNAL\";\n else if (riskLevel === \"HIGH\") action = \"REDUCE_EXPOSURE\";\n\n return {\n action,\n reasoning: `${bullishCount}/3 agents bullish. Risk: ${riskLevel}. Action: ${action}`,\n private: true, // NEVER shared with network\n };\n}\n\n// ─── Run Full Desk ───────────────────────────────────────────────────\nexport async function runDesk(\n deskId: string,\n symbol: string,\n timeframe: string = \"4H\",\n): Promise<{\n observations: ObservationResult[];\n execution: { action: string; reasoning: string; private: boolean };\n caseId?: string;\n}> {\n // Run all shared agents in parallel\n const [macro, structure, liquidity, risk, memory] = await Promise.all([\n runMacroAgent(deskId, symbol),\n runStructureAgent(deskId, symbol, timeframe),\n runLiquidityAgent(deskId, symbol, timeframe),\n runRiskAgent(deskId, symbol, timeframe),\n runMemoryAgent(deskId, symbol),\n ]);\n\n const observations = [macro, structure, liquidity, risk, memory];\n\n // Run execution agent PRIVATELY (never shared)\n const execution = await runExecutionAgent(deskId, symbol, observations);\n\n // Create a MarketCase from the strongest observation\n const sorted = observations\n .filter((o) => o.hypothesis && o.confidence > 0)\n .sort((a, b) => b.confidence - a.confidence);\n\n let caseId: string | undefined;\n if (sorted.length > 0 && sorted[0].hypothesis) {\n const best = sorted[0];\n caseId = await createMarketCase({\n desk_id: deskId,\n symbol,\n timeframe,\n hypothesis: best.hypothesis,\n expected_outcome: best.expected_outcome,\n reasoning: best.summary,\n invalidation_level: best.invalidation_level,\n conviction: best.confidence,\n time_horizon: timeframe,\n meta_tags: best.meta_tags,\n evidence: observations.filter((o) => o.observation_id).map((o) => o.observation_id),\n });\n }\n\n return { observations, execution, caseId };\n}\n",[],[239,288],"fa6ed36c-77e7-408a-b21f-b7e68e58f55f",["D",1785307426186],[211,212,213,214,215,216,217,218],[239],"e6ecdc51-6b7e-4b1f-b8fa-bf39bd9bc5e7","pipeline.ts","eb58ae1a-8b1a-11f1-9fc9-1607ee4eb77e","14a2dc12-cf92-4d0e-8c2d-8e10cb4dec84",8,["D",1785308315894],"interval","// ─── Automated Pipeline Cron Job ──────────────────────────────────────\n// Runs the full CNOS 8-stage cycle for all configured symbols on a schedule.\n// Schedule and symbols are controlled from the dashboard settings.\n\nimport { runPipeline } from \"../lib/cnos.ts\";\nimport { getSettings, saveSettings } from \"../lib/settings.ts\";\nimport { initDB } from \"../lib/db.ts\";\n\nexport default async function (interval: Interval) {\n await initDB();\n const settings = await getSettings();\n\n if (!settings.pipelineEnabled) {\n console.log(\"Pipeline is disabled — skipping run\");\n return;\n }\n\n console.log(`Pipeline run started at ${new Date().toISOString()}`);\n console.log(`Symbols: ${settings.pipelineSymbols.join(\", \")}`);\n console.log(`Timeframe: ${settings.pipelineTimeframe}`);\n console.log(`Desk: ${settings.deskId}`);\n\n const result = await runPipeline();\n\n console.log(`Pipeline complete: ${result.processed} symbols processed`);\n if (result.errors.length > 0) {\n console.log(`Errors: ${result.errors.join(\"; \")}`);\n }\n for (const r of result.results) {\n console.log(` ${r.symbol} → ${r.resolution} (proof: ${r.proofPassed}, contradictions: ${r.contradictions})`);\n }\n}\n",[],[308,299],"cron","28ed7aac-99c7-4431-9761-2ef1d84e3c27",["D",1785308315894],[209],[308],["D",1785309220363],[],[18,69],["D",1785308525726],[207],[18],["D",1785307366530],["D",1789476648181],{"_11":12,"_160":75},{"_11":12,"_155":156,"_74":75,"_31":73,"_157":158,"_159":30},{"_43":146,"_13":16,"_37":40,"_147":109,"_148":-5,"_149":-5,"_150":-5,"_151":-5,"_152":-5,"_153":14,"_154":12},"ctx",{"_57":-5,"_179":357,"_358":359,"_360":361,"_362":30,"_363":30,"_364":30,"_365":24,"_366":-5,"_367":-7,"_368":-7,"_369":26,"_370":30,"_371":30,"_372":-5,"_373":30,"_374":30,"_375":30,"_376":-5},"ENV",{"_328":329,"_330":331,"_332":16,"_333":334,"_335":336,"_337":338,"_339":340,"_341":342,"_343":344,"_345":346,"_347":348,"_349":350,"_351":352,"_353":354,"_355":356},"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"]

ATN — AGENTIC TRADING NETWORK

CNOS Cognitive Resolution Layer · Layer 6

Cognitive Resolution Cycle

Recent MarketCases

Recent Pipeline Runs

All MarketCases

Pipeline Control

Pipeline Run History

Telegram Notifications

Twelve Data API (Multiple Keys with Rotation)

Pipeline Configuration

Desk Configuration

Notification Preferences

Market Data Tools (MCP-style)

search_symbols

get_quote

get_technicals

get_technicals_rating

analyze_multi_timeframe

compute_levels

get_news

get_ohlcv

Resolved Memory Search

History · api