Agencio Predict - AI Fund Manager System Architecture

1. System Overview

The AI Fund Manager is an autonomous multi-strategy trading system that manages capital across multiple algorithms, asset classes, and brokers with real-time NAV tracking and automatic regime-based strategy adjustment.

Automatic Strategy Adjustment

Detects market regimes (BULL/BEAR/SIDEWAYS/CRISIS/RECOVERY) using PINN and Markov HMM, automatically switching to optimal strategies.

Multi-Asset Trading

Trades across Equities, Crypto, Forex, Futures, Options, and CFDs with unified position management.

Multiple Trading Modes

Supports Mock (simulated), Paper (realistic simulation), and Live (real broker execution) trading modes.

Multi-Broker Integration

Integrates with Alpaca, Binance, IBKR, Pepperstone, and Schwab for cross-market execution.

Multi-Symbol Execution

Executes trades across multiple symbols simultaneously with shared cash pool and correlation monitoring.

Real-Time NAV Tracking

Calculates and tracks NAV per share, total equity, daily/cumulative returns, and drawdown metrics.

2. Key Capabilities Summary

Production Ready: All 5 key capabilities are fully implemented and operational as of May 2026.

Capability	Description	Status
Auto-Adjust Strategies	Automatically switches between 6 regime-specific strategies based on PINN + Markov HMM regime detection (BULL/BEAR/SIDEWAYS/CRISIS/RECOVERY)	Active
Multiple Asset Classes	Trades Equities, Crypto, Forex, Futures, Options, CFDs across broker-specific universes	Active
Simulation & Live Modes	Mock ($100k simulated), Paper (realistic with slippage), Live (MFA-gated real execution)	Active
Multi-Trade Execution	Per-symbol entry/exit decisions, shared cash pool, portfolio-level risk, correlation monitoring	Active
Multiple Brokers	Alpaca, Binance, IBKR, Pepperstone, Schwab with unified adapter interface and credential encryption	Active

3. Full System Architecture

Complete architecture showing the Fund Manager orchestrating strategies, brokers, and regime detection.

flowchart TB subgraph FundManager["AI FUND MANAGER"] subgraph Orchestrator["Fund Orchestrator"] NAV["NAV Calculation"] Alloc["Capital Allocation"] Risk["Risk Limits"] Corr["Correlation Monitor"] Rebal["Auto Rebalancing"] end subgraph RegimeSelector["Regime Strategy Selector"] PINN["PINN Predictor"] Markov["Markov HMM"] Detector["Regime Detector"] Selector["Strategy Selector"] Fitness["Fitness Scoring"] end subgraph Strategies["Strategy Pool (6 Strategies)"] S1["Bull Market Momentum"] S2["Bull Factor Tilt"] S3["Bear Defensive"] S4["Bear Pairs Trading"] S5["Sideways Vol Premium"] S6["Crisis Risk Parity"] end subgraph Executor["Multi-Symbol Executor"] PriceVal["Price Validation"] PerSymbol["Per-Symbol Decisions"] SharedCash["Shared Cash Pool"] Slippage["Slippage Model"] end end subgraph DataFeeds["Market Data Feeds"] Yahoo["Yahoo Finance"] Finnhub["Finnhub"] CoinGecko["CoinGecko"] Polygon["Polygon.io"] Binance["Binance"] end subgraph Brokers["Broker Adapters"] Alpaca["Alpaca\n(Stocks, ETFs, Crypto)"] BinanceB["Binance\n(Crypto Spot + Futures)"] IBKR["IBKR\n(All Asset Classes)"] Pepper["Pepperstone\n(Forex, CFDs)"] Schwab["Schwab\n(US Stocks, Options)"] end subgraph Storage["Database"] Funds["predict.funds"] Positions["predict.fund_positions"] Trades["predict.fund_trades"] NAVHist["predict.fund_nav_history"] Audit["predict.fund_audit_log"] end DataFeeds --> PINN DataFeeds --> Markov PINN --> Detector Markov --> Detector Detector --> Selector Selector --> Fitness Fitness --> Strategies Orchestrator --> Strategies Strategies --> Executor Executor --> Brokers Brokers --> Storage Orchestrator --> Storage

4. Tick Loop Flow (60-Second Cycle)

The fund orchestrator runs a tick loop every 60 seconds for all active funds.

flowchart TB Start["Scheduler: fund-tick-active (60s)"] --> LoadFunds["Load all active funds"] LoadFunds --> ForEach["For each fund"] subgraph TickLoop["Per-Fund Tick Loop"] ForEach --> Step1["Step 1: Fetch live prices"] Step1 --> Step2["Step 2: Update position values"] Step2 --> Step3["Step 3: Calculate fund equity"] Step3 --> Step4["Step 4: Update NAV per share"] Step4 --> Step41{"Step 4.1: Auto-switch enabled?"} Step41 -->|"Yes"| CheckRegime["Get current regime + confidence"] CheckRegime --> SwitchDecision{"Switch urgency?"} SwitchDecision -->|"Immediate"| ExecuteSwitch["Create new algorithm\nStart paper run\nUpdate allocations"] SwitchDecision -->|"None"| Step5 ExecuteSwitch --> Step5 Step41 -->|"No"| Step5 Step5["Step 5: Tick each strategy"] Step5 --> Step6["Step 6: Process Stock Hunter\n(if enabled)"] Step6 --> Step7["Step 7: Check risk limits"] Step7 --> Step8{"Risk breach?"} Step8 -->|"Yes"| Pause["Auto-pause fund"] Step8 -->|"No"| Step9["Step 8: Check rebalancing"] Step9 --> Step10["Step 9: Update telemetry"] end Step10 --> NextFund["Next fund"] Pause --> NextFund NextFund --> ForEach

Scheduler Jobs

Job Name	Interval	Description
`fund-tick-active`	60 seconds	Main tick loop for all active funds
`fund-nav-snapshot`	Daily (market close)	Record daily NAV snapshot for performance charts
`fund-rebalance-check`	1 hour	Check all funds for rebalancing triggers
`fund-correlation-check`	4 hours	Monitor cross-strategy correlations

5. Regime Detection System

The system uses PINN (Physics-Informed Neural Network) combined with Markov HMM to detect market regimes.

flowchart LR subgraph Input["Market Data (252 bars)"] Prices["Price History"] Volume["Volume Data"] VIX["VIX Index"] Yields["Treasury Yields"] end subgraph PINN["PINN Predictor"] Features["16 Input Features"] Physics["Physics Constraints\n(Kelly, Liquidation, Mean-Rev)"] Outputs["3 Output Heads"] CrisisProb["Crisis Probability"] end subgraph Markov["Markov HMM (4-State)"] States["BULL | BEAR | SIDEWAYS | CRISIS"] Trans["Transition Matrix"] Prob["State Probabilities"] end subgraph Combined["Combined Regime Detection"] Merge["Confidence Scoring"] Regime["Final Regime"] Warning["Crisis Warning Level"] end Input --> PINN Input --> Markov PINN --> CrisisProb Markov --> Prob CrisisProb --> Merge Prob --> Merge Merge --> Regime Merge --> Warning

Market Regimes

Regime	Characteristics	Detection Signals	Strategy Approach
BULL	Positive drift, low volatility	Hurst > 0.55, positive factor alpha, VIX < 20	Momentum, trend-following, let winners run
BEAR	Negative drift, high volatility	Hurst < 0.45, negative drift, VIX > 25	Defensive, quick profits, tight stops
SIDEWAYS	No drift, moderate volatility	Hurst ~ 0.50, ranging price action	Mean-reversion, volatility selling, range trading
CRISIS	Extreme volatility, correlations spike	VIX > 35, PINN crisis prob > 0.5, yield curve signals	Capital preservation, hedging, risk parity
RECOVERY	Transitioning from crisis/bear	VIX declining, positive momentum building	Cautious momentum, rebuilding positions

6. Regime-Specific Strategies

Six pre-built strategies optimized for specific market regimes.

mindmap root((Regime Strategies)) BULL Market Bull Market Momentum Trend-following Wide stops Trailing stops Kelly sizing Bull Factor Tilt High-beta tilt Small-cap SMB Factor exposure BEAR Market Bear Defensive Mean-reversion Oversold entries Tight stops Low beta pref Bear Pairs Trading Market-neutral Cointegrated pairs Z-score extremes SIDEWAYS Sideways Vol Premium Vol selling IV greater than RV Range trading CRISIS Crisis Risk Parity Max preservation Diversified defense Minimal exposure

Strategy Details

Strategy	Regimes	Risk Level	Key Parameters
Bull Market Momentum	BULL RECOVERY	Medium	Stop: 6%, Trailing: 4%, Max Position: $40k
Bull Factor Tilt	BULL	High	High-beta targets, SMB factor tilt, growth bias
Bear Defensive	BEAR CRISIS	Low	Stop: 2.5%, Trailing: 1.5%, Max Position: $15k
Bear Pairs Trading	BEAR SIDEWAYS	Medium	Market-neutral, cointegration required, z-score > 2
Sideways Vol Premium	SIDEWAYS	Medium	IV > RV spread, range-bound symbols, vol decay
Crisis Risk Parity	CRISIS	Low	Stop: 3%, Max DD: 5%, Daily Loss: 1.5%, Max Position: $10k

7. Automatic Strategy Switching

The fund automatically switches strategies when regime conditions change.

sequenceDiagram participant Scheduler as Tick Loop (60s) participant Orchestrator as Fund Orchestrator participant Selector as Regime Strategy Selector participant PINN as PINN + Markov participant Executor as Algorithm Executor participant DB as Database Scheduler->>Orchestrator: tick(fundId) Orchestrator->>Orchestrator: Check: regimeAutoSwitchEnabled? alt Auto-Switch Enabled Orchestrator->>Selector: checkRegimeSwitch(fundId, currentStrategy) Selector->>PINN: detectRegime(priceHistory) PINN-->>Selector: { regime: BEAR, confidence: 0.78, crisisWarning: 0.15 } Selector->>Selector: evaluateFitness(allStrategies, regime) Selector->>Selector: computeSwitchUrgency() Selector-->>Orchestrator: { shouldSwitch: true, urgency: 'immediate', newStrategy: 'Bear Defensive' } alt Urgency = Immediate Orchestrator->>Executor: stopRun(oldRunId) Orchestrator->>DB: createAlgorithm(bearDefensiveAST) Orchestrator->>Executor: startRun(newAlgoId, 'paper') Orchestrator->>DB: updateFundStrategy(newAlgoId) Orchestrator->>DB: logAudit('regime_switch_executed') end end Orchestrator->>Orchestrator: Continue tick loop...

Switch Urgency Levels

Urgency	Trigger	Action
Immediate	Crisis warning > threshold OR fitness diff > 50%	Auto-execute switch if enabled
Soon	Fitness diff > 30% AND confidence >= threshold	Recommend in UI, user decides
Optional	Fitness diff > 15%	Show as suggestion
None	Current strategy is optimal	No action needed

8. Configurable Thresholds

Each fund has UI-configurable thresholds for regime detection and strategy switching.

Threshold	Default	Range	Description
`markovRegimeProbMin`	0.60	0.30 - 0.90	Minimum Markov regime probability to confirm regime
`hurstTrendingThreshold`	0.55	0.45 - 0.70	Hurst exponent threshold for trending confirmation
`garchVolatilityCap`	25	15 - 50	GARCH volatility cap (%) for normal trading
`crisisWarningThreshold`	0.30	0.10 - 0.60	Crisis warning level to trigger defensive switch
`regimeSwitchConfidenceMin`	0.70	0.50 - 0.95	Confidence required to trigger automatic switch
`minTicksBetweenSwitches`	10	1 - 60	Minimum ticks (minutes) before allowing another switch

Threshold Presets

Conservative

Higher thresholds for more certainty. Lower crisis warning for earlier defensive moves. Longer wait between switches.

{ markovRegimeProbMin: 0.70, crisisWarningThreshold: 0.20, minTicksBetweenSwitches: 20 }

Moderate (Default)

Balanced thresholds suitable for most market conditions.

{ markovRegimeProbMin: 0.60, crisisWarningThreshold: 0.30, minTicksBetweenSwitches: 10 }

Aggressive

Lower thresholds for faster regime response. Higher crisis tolerance. Shorter intervals between switches.

{ markovRegimeProbMin: 0.50, crisisWarningThreshold: 0.40, minTicksBetweenSwitches: 5 }

9. Multi-Asset Class Support

The fund supports trading across 6 asset classes with broker-specific routing.

flowchart TB subgraph AssetClasses["Asset Classes"] Equities["Equities\nAAPL, MSFT, SPY, QQQ"] Crypto["Crypto\nBTC, ETH, SOL"] Forex["Forex\nEUR/USD, GBP/USD"] Futures["Futures\nES, NQ, CL, GC"] Options["Options\nSPY calls/puts"] CFDs["CFDs\nIndices, Commodities"] end subgraph Brokers["Broker Routing"] Alpaca["Alpaca"] Binance["Binance"] IBKR["IBKR"] Pepper["Pepperstone"] Schwab["Schwab"] end Equities --> Alpaca Equities --> IBKR Equities --> Schwab Crypto --> Alpaca Crypto --> Binance Forex --> Pepper Forex --> IBKR Futures --> IBKR Options --> IBKR Options --> Schwab CFDs --> Pepper

Asset Class	Supported Brokers	Example Symbols	Special Features
Equities	Alpaca, IBKR, Schwab	AAPL, MSFT, SPY, QQQ	Commission-free (Alpaca/Schwab)
Crypto	Alpaca, Binance	BTC, ETH, SOL	24/7 trading, futures support (Binance)
Forex	Pepperstone, IBKR	EUR/USD, GBP/USD	70+ pairs, leverage available
Futures	IBKR	ES, NQ, CL, GC	Contract multipliers, margin
Options	IBKR, Schwab	SPY calls/puts	Full chain, Greeks
CFDs	Pepperstone	Indices, Commodities	Leveraged products

10. Trading Modes

Three trading modes with increasing levels of risk and required verification.

flowchart LR subgraph Mock["MOCK MODE"] M1["Simulated $100k"] M2["No broker connection"] M3["Zero risk"] M4["Default for all users"] end subgraph Paper["PAPER MODE"] P1["Realistic simulation"] P2["Almgren-Chriss slippage"] P3["Bid-ask spread modeling"] P4["Fill delays"] end subgraph Live["LIVE MODE"] L1["Real broker execution"] L2["MFA verification required"] L3["Risk acknowledgement"] L4["Real capital at risk"] end Mock -->|"Opt-in"| Paper Paper -->|"Graduation gates + MFA"| Live

Mode	Description	Risk Level	Requirements
Mock	Simulated $100k portfolio, no broker connection, instant fills	None	Default - no setup needed
Paper	Realistic simulation with slippage, spread, and fill delay modeling	None	Opt-in via settings
Live	Real broker execution with actual capital	Real capital at risk	MFA + Risk Ack + Graduation Gates

Paper Trading Features

Realistic Simulation: Paper mode uses institutional-grade simulation:

Almgren-Chriss Model: Square-root market impact based on trade size and volatility
Bid-Ask Spread: Asset-class specific spreads (tighter for liquid equities, wider for crypto)
Fill Delays: Simulated latency based on market conditions
Slippage Calculation: Persistent tracking in algorithm_trades.slippage_bps

11. Multi-Symbol Execution

The executor processes each symbol in the strategy's universe with shared resource management.

flowchart TB subgraph Strategy["Strategy AST"] Universe["universe: ['AAPL', 'MSFT', 'NVDA', 'GOOGL']"] Entry["entry: { when: rsi(14) < 30, size: kelly(0.5) }"] Exit["exit: { when: rsi(14) > 70 }"] end subgraph Executor["Multi-Symbol Executor"] FetchPrices["Fetch prices for all symbols"] Validate["Price validation (bounds, spikes)"] subgraph PerSymbol["For Each Symbol"] BuildCtx["Build EvalContext"] EvalEntry{"entry.when?"} EvalExit{"exit.when?"} CalcSize["Calculate position size"] end subgraph Shared["Shared Resources"] CashPool["Shared Cash Pool"] RiskBudget["Portfolio Risk Budget"] CorrMatrix["Correlation Matrix"] Concentration["Concentration Limits"] end end subgraph Output["Trade Execution"] Trades["Generated Trades"] Positions["Updated Positions"] Telemetry["Equity Telemetry"] end Universe --> FetchPrices FetchPrices --> Validate --> PerSymbol PerSymbol --> Shared Shared --> Trades --> Positions --> Telemetry

Multi-Symbol Features

Feature	Description
Per-Symbol Decisions	Independent entry/exit evaluation for each symbol in universe
Shared Cash Pool	All positions draw from common cash balance, preventing over-allocation
Portfolio Risk Management	Aggregate risk limits applied across all positions
Correlation Monitoring	Tracks pairwise correlations, applies haircuts when correlated
Concentration Limits	Max single position, sector limits, asset class limits

12. Multi-Broker Integration

Unified broker adapter interface with credential encryption and auth-specific flows.

flowchart TB subgraph Adapters["BrokerAdapter Interface"] Interface["placeOrder() | getPositions() | getBalance() | closeAll()"] end subgraph Alpaca["Alpaca"] AlpacaAuth["API Key Auth"] AlpacaAssets["Stocks, ETFs, Crypto"] AlpacaFeatures["Commission-free\n$100k paper balance"] end subgraph Binance["Binance"] BinanceAuth["API Key Auth"] BinanceAssets["Crypto spot + futures"] BinanceFeatures["Testnet for paper\n24/7 trading"] end subgraph IBKR["Interactive Brokers"] IBKRAuth["Session Auth\n(Gateway required)"] IBKRAssets["All asset classes"] IBKRFeatures["Gateway at localhost:5000\n55s keepalive ping"] end subgraph Pepper["Pepperstone"] PepperAuth["OAuth 2.0 + PKCE"] PepperAssets["Forex, CFDs, indices"] PepperFeatures["30-day token\n6-hour refresh job"] end subgraph Schwab["Charles Schwab"] SchwabAuth["OAuth 2.0 + PKCE"] SchwabAssets["US stocks, ETFs, options"] SchwabFeatures["7-day token limit\nWeekly re-auth required"] end Interface --> Alpaca Interface --> Binance Interface --> IBKR Interface --> Pepper Interface --> Schwab

Broker	Auth Type	Asset Classes	Key Notes
Alpaca	API Key	Stocks, ETFs, Crypto	Commission-free, $100k paper balance
Binance	API Key	Crypto spot + futures	Testnet for paper trading
IBKR	Session	All (stocks, options, futures, forex, bonds)	Requires IBKR Gateway on localhost:5000
Pepperstone	OAuth 2.0	Forex, CFDs, indices, commodities	30-day token, auto-refresh scheduler
Schwab	OAuth 2.0	US stocks, ETFs, options	7-day hard token limit - weekly re-auth

Security: All broker credentials are encrypted with AES-256-GCM. API responses only return ****last4 for sensitive fields. Live trading requires MFA verification.

13. NAV Calculation

Real-time Net Asset Value calculation tracking fund performance.

flowchart LR subgraph Inputs["Inputs"] Positions["Open Positions\n(qty × current_price)"] Cash["Cash Balance\n(current_capital)"] Shares["Shares Outstanding"] end subgraph Calculation["NAV Calculation"] PosValue["Sum Position Values"] TotalEquity["Total Equity =\nPositions + Cash"] NAVShare["NAV/Share =\nTotal Equity / Shares"] end subgraph Metrics["Performance Metrics"] DailyRet["Daily Return =\n(NAV - prevNAV) / prevNAV"] CumRet["Cumulative Return =\n(NAV - initialNAV) / initialNAV"] Drawdown["Drawdown =\n(peak - equity) / peak"] HWM["High Water Mark"] end Positions --> PosValue Cash --> TotalEquity PosValue --> TotalEquity Shares --> NAVShare TotalEquity --> NAVShare NAVShare --> DailyRet NAVShare --> CumRet NAVShare --> Drawdown

NAV Formula

function calculateNAV(fund, positions) {
  // 1. Sum position market values
  const positionsValue = positions.reduce((sum, p) => sum + p.market_value, 0);

  // 2. Add cash balance
  const totalEquity = positionsValue + fund.currentCapital;

  // 3. Calculate NAV per share
  const navPerShare = totalEquity / fund.sharesOutstanding;

  // 4. Calculate returns
  const dailyReturn = (navPerShare - previousNav) / previousNav;
  const cumulativeReturn = (navPerShare - initialNav) / initialNav;

  // 5. Calculate drawdown
  const drawdown = (peakEquity - totalEquity) / peakEquity;

  return { navPerShare, totalEquity, dailyReturn, cumulativeReturn, drawdown };
}

14. Capital Allocation Strategies

Three allocation strategies for distributing capital across strategies.

flowchart TB subgraph Strategies["Available Strategies"] S1["Strategy A\nSharpe: 1.5\nDD: 8%"] S2["Strategy B\nSharpe: 1.2\nDD: 12%"] S3["Strategy C\nSharpe: 0.9\nDD: 15%"] end subgraph EqualWeight["Equal Weight"] EW["Each gets 33.3%"] end subgraph SharpeWeight["Sharpe Weighted"] SW["A: 41.7%\nB: 33.3%\nC: 25.0%"] end subgraph RiskParity["Risk Parity"] RP["Inverse volatility:\nA: 43.5%\nB: 29.0%\nC: 27.5%"] end Strategies --> EqualWeight Strategies --> SharpeWeight Strategies --> RiskParity

Strategy	Formula	Best For
Equal Weight	`allocation = 1 / numStrategies`	Simple diversification, no performance history
Sharpe Weighted	`allocation = max(0, sharpe) / sum(sharpes)`	Allocate more to risk-adjusted performers
Risk Parity	`allocation = (1/vol) / sum(1/vols)`	Balance risk contribution across strategies

15. Automatic Rebalancing

Fund automatically rebalances when allocations drift or on schedule.

flowchart TB subgraph Triggers["Rebalancing Triggers"] Scheduled["Scheduled\n(rebalance_interval_days)"] Drift["Allocation Drift\n(> rebalance_threshold_pct)"] Manual["Manual\n(User request)"] Risk["Risk Event\n(Post-crisis recovery)"] end subgraph Process["Rebalancing Process"] Calc["Calculate target allocations"] Compare["Compare to current allocations"] Delta["Compute required trades"] Execute["Execute rebalancing trades"] Record["Record to fund_rebalancing_events"] end subgraph Outcome["Outcome"] Updated["Updated position sizes"] Audit["Audit trail entry"] NAVUpdate["NAV recalculated"] end Triggers --> Process --> Outcome

Rebalancing Settings

Setting	Description	Default
`auto_rebalance_enabled`	Enable automatic rebalancing	false
`rebalance_threshold_pct`	Drift percentage to trigger rebalance	5%
`rebalance_interval_days`	Scheduled rebalancing interval	30 days

16. Risk Management

Multi-layer risk management with automatic fund pausing.

flowchart TB subgraph RiskChecks["Risk Checks (Every Tick)"] MaxDD["Max Drawdown Check"] DailyLoss["Daily Loss Check"] Concentration["Position Concentration"] Correlation["Strategy Correlation"] end subgraph Thresholds["Risk Presets"] Conservative["Conservative\nDD: 10%, Daily: 2%, Pos: 15%"] Moderate["Moderate\nDD: 20%, Daily: 5%, Pos: 25%"] Aggressive["Aggressive\nDD: 30%, Daily: 8%, Pos: 40%"] end subgraph Actions["Breach Actions"] Warn["Warning logged"] Pause["Fund auto-pauses"] Notify["User notified"] Audit["Audit trail entry"] end RiskChecks --> |"Breach"| Actions Thresholds --> RiskChecks

Risk Presets

Level	Max Drawdown	Max Daily Loss	Max Position
Conservative	10%	2%	15%
Moderate	20%	5%	25%
Aggressive	30%	8%	40%

Auto-Pause: When a fund breaches max drawdown or daily loss limits, it automatically pauses and stops all trading. The user is notified and must manually restart after reviewing the situation.

Database Schema

Core tables for fund management (Migration 228).

erDiagram funds ||--o{ fund_strategies : has funds ||--o{ fund_positions : holds funds ||--o{ fund_trades : executes funds ||--o{ fund_nav_history : tracks funds ||--o{ fund_rebalancing_events : rebalances funds ||--o{ fund_audit_log : logs funds { uuid id PK uuid user_id FK string name string status numeric initial_capital numeric current_capital numeric total_equity numeric shares_outstanding numeric nav_per_share numeric high_water_mark string risk_level string allocation_strategy numeric max_drawdown_pct numeric max_daily_loss_pct numeric max_position_pct boolean stock_hunter_enabled numeric stock_hunter_min_confidence boolean auto_rebalance_enabled numeric rebalance_threshold_pct jsonb regime_thresholds boolean regime_auto_switch_enabled timestamp last_regime_switch_at } fund_strategies { uuid id PK uuid fund_id FK uuid algorithm_id FK string strategy_type numeric target_allocation_pct numeric current_allocation_pct boolean enabled uuid algorithm_run_id } fund_positions { uuid id PK uuid fund_id FK string symbol uuid strategy_id FK string status numeric quantity numeric entry_price numeric current_price numeric market_value numeric unrealized_pnl numeric realized_pnl } fund_trades { uuid id PK uuid fund_id FK uuid position_id FK string symbol string side numeric quantity numeric price numeric total_value string source_type } fund_nav_history { uuid id PK uuid fund_id FK date snapshot_date numeric nav_per_share numeric total_equity numeric daily_return_pct numeric cumulative_return_pct numeric drawdown_pct }

API Endpoints

17 API routes for fund management.

Method	Path	Description
POST	`/api/predict/v1/fund-manager`	Create new fund
GET	`/api/predict/v1/fund-manager`	List user's funds
GET	`/api/predict/v1/fund-manager/:id`	Get fund details
PUT	`/api/predict/v1/fund-manager/:id`	Update fund config
DELETE	`/api/predict/v1/fund-manager/:id`	Delete fund
POST	`/api/predict/v1/fund-manager/:id/start`	Start/resume fund
POST	`/api/predict/v1/fund-manager/:id/pause`	Pause fund
POST	`/api/predict/v1/fund-manager/:id/stop`	Stop fund & close positions
POST	`/api/predict/v1/fund-manager/:id/deposit`	Deposit capital
POST	`/api/predict/v1/fund-manager/:id/withdraw`	Withdraw capital
POST	`/api/predict/v1/fund-manager/:id/strategies`	Add strategy
DELETE	`/api/predict/v1/fund-manager/:id/strategies/:sid`	Remove strategy
GET	`/api/predict/v1/fund-manager/:id/performance`	Get performance metrics
GET	`/api/predict/v1/fund-manager/:id/nav-history`	Get NAV history
GET	`/api/predict/v1/fund-manager/:id/positions`	Get open positions
GET	`/api/predict/v1/fund-manager/:id/trades`	Get recent trades
POST	`/api/predict/v1/fund-manager/:id/rebalance`	Trigger manual rebalance

File Locations

Key backend and frontend files for the AI Fund Manager.

Backend Files

File	Purpose
`packages/be/src/fund-manager/types.ts`	Type definitions for Fund, FundStrategy, FundPosition, FundTrade
`packages/be/src/fund-manager/repository.ts`	Database CRUD operations
`packages/be/src/fund-manager/service.ts`	Business logic: create/update/delete, NAV, rebalancing
`packages/be/src/fund-manager/orchestrator.ts`	Main tick loop and fund coordination
`packages/be/src/fund-manager/scheduler.ts`	Scheduler job definitions
`packages/be/src/fund-manager/regime-strategy-selector.ts`	Regime detection and strategy selection
`packages/be/src/api/predict/v1/fund-manager/handlers.ts`	API handlers

Frontend Files

File	Purpose
`packages/fe/src/components/fund-manager/types.ts`	UI types
`packages/fe/src/components/fund-manager/FundCard.tsx`	Summary card for fund list
`packages/fe/src/components/fund-manager/FundCreator.tsx`	5-step creation wizard
`packages/fe/src/components/fund-manager/FundDashboard.tsx`	Main fund dashboard
`packages/fe/src/components/fund-manager/RegimeStrategyPanel.tsx`	Regime strategy controls
`apps/web/src/app/fund-manager/page.tsx`	Fund list page
`apps/web/src/app/fund-manager/[id]/page.tsx`	Fund detail page

Document	Description
Regime Strategy Selector	Detailed regime detection and strategy switching documentation
Algorithm Trading System	DSL primitives, backtest engine, paper executor
PINN Architecture	Physics-informed neural network for regime prediction
`docs/internal/trading/68-ai-fund-manager.md`	Full markdown documentation
`docs/internal/trading/82-regime-strategy-selector.md`	Regime strategy selector documentation

System Statistics