Skip to content

Redhound Product Roadmap — Trader's Heaven

Turn Redhound from a single-session stock analyzer into a full-stack trading intelligence platform that alerts traders to opportunities, manages their portfolio, and validates strategies with historical data.

Vision

A trader opens Redhound in the morning. Their dashboard shows overnight alerts — NVDA hit a key support level, AAPL's Piotroski score just improved, sector rotation signals are flashing for energy. They drill into the NVDA alert, see the full multi-agent analysis already ran automatically, and decide to add to their position. Their portfolio view updates with the new position, shows real-time P&L, and the risk engine recalculates their overall exposure. Later, before deploying a new mean-reversion strategy, they backtest it across 2 years of data and compare it against their current approach.

What We Have Today

Core Analysis Engine - 10-specialist analyst team: 5 core analysts (Technical, Fundamentals, Sentiment, News, Market Context) running in parallel + 5 optional edge analysts (Sector, Insider, Options Flow, Short Interest, Earnings Revisions) dynamically weighted when data is available (40–50% faster than sequential) - Confidence-gated debate: HIGH confidence (>80%) skips debate entirely; MEDIUM runs 1 round; LOW runs 2 rounds — saving 30–70K tokens per session - Deterministic Risk Overlay: Zero-LLM position sizing, stop-loss (ATR/Bollinger), and regime-aware adjustments - Signal accuracy system: Entropy-based confidence, VIX linear interpolation, RSI divergence detection, OBV volume confirmation, multi-timeframe confirmation, sector rotation signal, earnings proximity discounts, and regime-specific BUY/SELL thresholds

Adaptive Intelligence - Adaptive analyst weights: Weekly accuracy tracking drives per-analyst weight updates via WeightService - Vector memory: pgvector-powered historical context retrieval for consistent decision making - Walk-forward backtesting: Out-of-sample validation with MAE/MFE per-trade analysis

Web Application (redhound.vercel.app) - Sessions: Full session history, real-time agent reasoning feed, debate visualization - Stock profiles: Live price, fundamentals, indicators, and signals for any ticker - Screener: Filter S&P 500 / NASDAQ 100 by technical and fundamental conditions in real-time - Portfolio: Track holdings, P&L, allocations, and risk exposure - Watchlists & Alerts: Multiple watchlists; price, technical, fundamental, and signal alerts - Background scanner: Continuous market scanning with configurable conditions and auto-analysis - Opportunities feed: Detected trading opportunities with confidence scoring - Backtest UI: Strategy builder, equity curves, trade log, monthly returns heatmap

Infrastructure - Real-time price feed: WebSocket-based live price broadcasting during market hours - Usage dashboard: Token and cost tracking per session and analyst - Observability: Prometheus/Grafana stack, structured logging, health checks

Phase 1: Portfolio Hub ✅

The core screen traders will live in. Track positions, see P&L, and understand exposure at a glance.

1.1 Portfolio Dashboard (/portfolio)

  • Holdings table: Ticker, shares, avg cost, current price, P&L ($ and %), weight in portfolio
  • Portfolio summary bar: Total value, day change, total P&L, cash balance
  • Allocation donut chart: By sector, by position size
  • Risk heatmap: Per-position risk score (using existing RiskScorer system), concentration risk, correlation warnings
  • Quick actions: "Analyze" button per holding (launches session), "Add position" modal

1.2 Position Management

  • Manual position entry: Ticker, shares, entry price, entry date, notes
  • CSV import: Upload brokerage export (support common formats: Schwab, Fidelity, IBKR, generic)
  • Edit/close positions: Update shares, record partial/full exits with exit price
  • Transaction history: Log of all adds, reduces, closes with timestamps
  • P&L tracking: Realized vs unrealized, per-position and aggregate

1.3 Portfolio Analytics

  • Correlation matrix: Between portfolio holdings (backend API exists, need frontend)
  • Volatility analysis: Per-holding and portfolio-level (backend API exists)
  • Beta exposure: Portfolio beta vs SPY (backend API exists)
  • Sector breakdown: Auto-classify holdings by GICS sector
  • Concentration alerts: Warn when single position >20% or sector >40%

Data Model

portfolios: id, user_id, name, created_at
positions: id, portfolio_id, symbol, shares, avg_cost, status (open/closed), opened_at, closed_at, notes
transactions: id, position_id, type (buy/sell), shares, price, date, fees

Backend

  • POST/GET/PUT/DELETE /api/v1/portfolios/ — CRUD
  • POST/GET/PUT /api/v1/portfolios/{id}/positions/ — position management
  • GET /api/v1/portfolios/{id}/analytics — returns correlation, volatility, beta, allocation
  • GET /api/v1/portfolios/{id}/performance — time-series P&L (daily/weekly/monthly)
  • Price refresh: Periodic vendor fetch for current prices (cache 1 min during market hours, 1 hour after)

Phase 2: Watchlists & Alerts ✅

Traders need to track stocks they're interested in and get notified when conditions are met.

2.1 Watchlist (/watchlist)

  • Multiple watchlists: "Earnings This Week", "Breakout Candidates", "Value Picks"
  • Watchlist table: Ticker, last price, day change %, 5-day trend sparkline, last signal (from sessions), notes
  • Quick add: Ticker search with autocomplete (existing component)
  • Bulk analyze: Select multiple tickers → run batch analysis
  • Drag-and-drop reorder and color-coded tags

2.2 Alert System

  • Price alerts: Above/below threshold, % change from current, 52-week high/low proximity
  • Technical alerts: RSI oversold (<30) or overbought (>70), MACD crossover, Bollinger Band squeeze/breakout, SMA crossover (golden cross/death cross)
  • Fundamental alerts: Earnings surprise (>5% beat/miss), Piotroski score change, significant insider buying
  • Signal alerts: When a Redhound session produces a BUY/SELL signal with confidence above threshold
  • Delivery channels:
  • In-app notification center (bell icon in header, dropdown feed)
  • Email digest (configurable: real-time, hourly, daily)
  • Push notifications (PWA / service worker)
  • Webhook URL (for power users who want Slack/Discord/Telegram integration)

2.3 Alert Engine (Backend)

  • Scheduler: APScheduler background jobs checking alert conditions every 1–15 minutes
  • Price checks: Every 1 minute during market hours
  • Technical checks: Every 15 minutes
  • Fundamental checks: Daily after market close
  • Alert evaluation pipeline: Fetch latest data → evaluate condition → if triggered, create notification + respect cooldown (don't spam same alert)
  • Alert history: Log of all triggered alerts with timestamps

Data Model

watchlists: id, user_id, name, color, sort_order, created_at
watchlist_items: id, watchlist_id, symbol, notes, tags[], added_at, sort_order
alerts: id, user_id, symbol, type (price/technical/fundamental/signal), condition JSONB,
        status (active/triggered/paused/expired), cooldown_minutes, last_triggered_at, created_at
notifications: id, user_id, alert_id, title, body, read, channel (in_app/email/push/webhook), created_at

Phase 3: Continuous Background Monitoring ✅

The killer feature. Redhound runs autonomously, scanning the market for opportunities and surfacing them proactively.

3.1 Market Scanner

  • Screener presets: Oversold bounce candidates, momentum breakouts, value accumulation zones, earnings catalysts
  • Custom screens: Combine conditions (RSI < 30 AND Piotroski > 6 AND price near 52-week low)
  • Universe selection: S&P 500, NASDAQ 100, custom list, watchlist, portfolio holdings
  • Scan frequency: Configurable (every 15min / hourly / daily)
  • Scan results page (/scanner): Table of matches with key metrics, click to drill into full analysis

3.2 Opportunity Detection Pipeline

The heart of the system. A lightweight, fast pre-screen that decides which tickers deserve a full multi-agent analysis.

[Scheduler] → [Universe Filter] → [Quick Screen (deterministic)]
  → RSI, price vs SMA, volume spike, earnings proximity, news sentiment
  → Score each ticker (0-100 opportunity score)
  → Top N tickers above threshold → [Full Multi-Agent Analysis]
  → Results → [Notification Engine] → Alert user
  • Quick screen runs without LLM calls — pure technical/fundamental checks using existing analyst tools
  • Full analysis only triggered for high-scoring opportunities (configurable threshold)
  • Dedup logic: Don't re-analyze ticker if analyzed in last 24 hours unless significant price movement (>3%)
  • Rate limiting: Max N concurrent analyses, queue overflow tickers for next cycle

3.3 Opportunity Feed (/opportunities)

  • Real-time feed: Cards showing detected opportunities with:
  • Ticker, current price, opportunity type (oversold bounce, breakout, etc.)
  • Key metrics that triggered detection
  • Confidence score, timestamp
  • "View Full Analysis" link (if auto-analyzed)
  • "Analyze Now" button (if only pre-screened)
  • Filters: By opportunity type, confidence, sector, watchlist membership
  • Time-based view: "Today", "This Week", "This Month"

3.4 Monitoring Dashboard (/monitor)

  • Active scans status: Which scanners are running, last run time, next run time
  • Scan history: Results per scan cycle (found N candidates, triggered M analyses)
  • Resource usage: API calls consumed, rate limit status per data vendor
  • Pause/resume controls: Per scanner

Backend Architecture

  • Scheduler: APScheduler in-process cron jobs, no external worker required
  • Scanner: Runs as background FastAPI task, picks up due scanner configs every minute
  • Scan job: Symbol list + conditions → evaluate → results
  • Auto-analysis job: Full RedhoundGraph execution for top candidates
  • Config: Max concurrent scans, cool-down periods, API budget limits

Phase 4: Historical Backtesting ✅

Validate strategies before committing capital. Build on the existing BacktestEngine.

4.1 Backtest Engine

  • Real historical OHLCV data from the vendor API for entry/exit simulation
  • Trade simulation: Entry at signal price, exit at stop-loss or take-profit or holding period
  • Slippage and commissions (configurable)
  • MAE/MFE per trade: Maximum Adverse/Favorable Excursion analysis with optimal entry delay
  • Walk-forward validation: Sliding train/test windows for out-of-sample performance assessment

4.2 Backtesting UI (/backtest)

  • Strategy builder:
  • Select analysts to include and their weights
  • Set confidence thresholds
  • Choose position sizing method (fixed, Kelly, risk-parity)
  • Set holding period, stop-loss %, take-profit %
  • Universe & date range:
  • Single ticker or list
  • Start/end date picker
  • Split into in-sample/out-of-sample periods
  • Run backtest: Progress bar, cancel button, estimated time

4.3 Backtest Results (/backtest/{id})

  • Performance summary: Total return, annualized return, Sharpe ratio, Sortino ratio, max drawdown, win rate, profit factor, avg trade duration
  • Equity curve chart: Portfolio value over time with drawdown overlay
  • Trade log table: Entry/exit date, ticker, direction, P&L, holding period
  • Monthly returns heatmap: Green/red grid by month and year
  • Comparison mode: Run two strategies side-by-side, overlay equity curves
  • Benchmark comparison: vs SPY buy-and-hold over same period

4.4 Weight Optimizer

  • Use existing BacktestEngine.optimize_weights() with real data
  • UI: Slider for each analyst weight, "Auto-optimize" button
  • Optimization target: Sharpe, Sortino, max return, min drawdown (selectable)
  • Results: Heatmap showing Sharpe by weight combination, recommended weights
  • Apply to live: One-click to update signal aggregation weights in settings

4.5 Strategy Library

  • Save strategies: Name, description, parameters (weights, thresholds, sizing)
  • Share strategies: Public/private toggle, shareable link
  • Leaderboard: Rank strategies by Sharpe/return over standardized test period

Data Model

backtest_runs: id, user_id, config JSONB (weights, universe, dates, sizing),
               status (pending/running/completed/failed), results JSONB, created_at
backtest_trades: id, run_id, symbol, direction, entry_date, entry_price,
                 exit_date, exit_price, pnl, holding_days
strategies: id, user_id, name, description, config JSONB, is_public, created_at

Phase 5: Signal Intelligence & Accuracy Tracking 🔄

Close the loop. Track whether signals were right, learn from mistakes, and improve over time.

5.1 Signal Performance Dashboard (/signals)

  • Accuracy metrics: Overall hit rate, by ticker, by action (BUY/SELL/HOLD), by confidence bucket
  • P&L attribution: Which analyst contributes most to winning trades?
  • Confidence calibration chart: Are 80% confidence signals right 80% of the time?
  • Time-based accuracy: Has accuracy improved over time? (agent memory should help)
  • Filters: Date range, ticker, action type, confidence range

5.2 Automatic Outcome Tracking

  • For portfolio positions that were analyzed: Automatically track price movement after signal
  • Outcome windows: 1 day, 5 days, 20 days, 60 days post-signal
  • Auto-classify outcome: Correct (price moved in predicted direction by >X%), Wrong, Neutral
  • Feed outcomes to agent memory: Wire reflect_and_remember() into automatic flow

5.3 Analyst Leaderboard

  • Rank all analysts by contribution to accuracy
  • Show which analyst's signal diverges most from final decision (and whether that divergence helps or hurts)
  • Suggest weight adjustments based on historical accuracy data

Phase 6: Enhanced Market Intelligence

Broader market context and richer data to make the platform indispensable.

6.1 Market Overview Page (/market)

  • Market indices: S&P 500, NASDAQ, DOW, Russell 2000 with intraday charts
  • Sector heatmap: S&P sectors colored by daily performance (like finviz map)
  • Market breadth: Advance/decline ratio, new highs/lows, VIX gauge
  • Economic calendar: Upcoming Fed meetings, CPI, jobs data, earnings (FRED API integration exists)
  • Fear & Greed indicator: Composite of VIX, put/call ratio, breadth, momentum

6.2 Ticker Deep Dive Enhancement

  • Earnings history: EPS surprise chart, revenue trends
  • Institutional ownership: 13F data tracking (quarterly)
  • Options flow: Unusual options activity (if data source available)
  • Comparison mode: Side-by-side two tickers with all analyst metrics

6.3 News & Sentiment Hub

  • Real-time news feed: Aggregated from multiple sources, filtered by portfolio/watchlist
  • Sentiment trend chart: Per-ticker sentiment over time
  • Event impact analysis: How did past earnings/FDA approvals/etc. affect price?
  • AI news digest: Daily summary of market-moving news relevant to user's holdings

Technical Enablers (Cross-Cutting)

Background Job System ✅

  • Technology: APScheduler (in-process, no external worker required)
  • Jobs: Alert checks, scan cycles, price refreshes, adaptive weight updates, daily digests

Real-Time Price Service ✅

  • Technology: PriceBroadcastService — bulk vendor API fetch every 30 seconds during market hours
  • Delivery: WebSocket channel broadcast to all subscribed frontend clients

Batch Analysis Engine ✅

  • Concurrency: asyncio.Semaphore (max 3 concurrent scans)
  • Queue: Per-scanner dedup (24h cooldown unless >3% price move)
  • Auto-analysis cap: 5 per scan cycle

Notification Infrastructure ✅ (in-app) / Pending (email/push/webhook)

  • In-app: WebSocket push to connected clients; persisted for offline access
  • Email / push / webhook: Planned for future iteration

Mobile-Responsive Design

  • Ensure all pages work on mobile (traders check on the go)
  • Key mobile priorities: Portfolio overview, alert notifications, opportunity feed

Status Summary

Phase Status Notes
Phase 1: Portfolio Hub ✅ Implemented Portfolio CRUD, P&L, analytics
Phase 2: Watchlists & Alerts ✅ Implemented Watchlists, alert engine, in-app notifications
Phase 3: Background Monitoring ✅ Implemented Scanner, screener, opportunity detection
Phase 4: Historical Backtesting ✅ Implemented Walk-forward, MAE/MFE, weight optimizer
Phase 5: Signal Intelligence 🔄 In Progress Adaptive weights done; full dashboard UI pending
Phase 6: Market Intelligence Pending Market overview, sector heatmap, options flow

Success Metrics

  • Daily active users: Traders returning daily (goal: >60% DAU/MAU)
  • Sessions per user per week: Are they analyzing more? (goal: >10)
  • Alert-to-action rate: % of alerts that lead to analysis or trade (goal: >30%)
  • Signal accuracy: Tracked BUY/SELL signals that were correct at 5-day window (goal: >55%)
  • Portfolio coverage: % of portfolio holdings with recent (<7 day) analysis (goal: >80%)
  • Time to opportunity: Minutes from market event to user notification (goal: <15 min)

Non-Goals (For Now)

  • Automated trading / order execution: We inform, the trader decides
  • Options / futures / crypto: Focus on US equities first
  • Social features: No forums, no copy-trading, no public profiles
  • Brokerage integration: No direct broker API connections (manual position entry only)
  • Mobile native app: PWA first, native later if warranted