# CoreGamma — Financial Intelligence Terminal

> CoreGamma (coregamma.io) is a web-based financial intelligence platform for retail investors. It aggregates institutional-grade market data, options analytics (GEX, IV term structure, max pain), alternative datasets (congressional trades, dark pool flow, SEC filings), an autonomous AI Trading Copilot with 27-tool semantic routing, AI-powered summarization across stock profiles and risk analysis, macroeconomic regime analysis, and a paper trading engine into a single terminal. The platform operates on a freemium model with basic and premium tiers.

---

## Key Capabilities

### Options Analytics
- **Gamma Exposure (GEX) Profile**: Per-strike net dollar gamma exposure computed as `(call_gamma × call_OI × 100 × spot) + (put_gamma × put_OI × 100 × spot × -1)`. Visualized as a diverging bar chart with positive (dealer long gamma) and negative (dealer short gamma) zones.
- **Gamma Flip Zone Detection**: Identifies the exact price level where net dealer gamma crosses zero via linear interpolation between adjacent strikes. Multiple crossings are disambiguated by surrounding open interest concentration. This level marks the transition between dealer-stabilizing and dealer-amplifying hedging behavior.
- **IV Term Structure**: ATM implied volatility plotted across all available expirations. Detects contango (normal) vs. backwardation (stress) regimes using a >2% IV differential threshold between short-term (~7 DTE) and medium-term (~45 DTE) expirations.
- **25-Delta Put/Call IV Skew**: Measures the implied volatility spread between 25-delta OTM puts and 25-delta OTM calls on the front-month expiration (~30 DTE). Positive skew indicates put premium dominance (downside hedging demand); negative skew indicates call premium dominance.
- **Max Pain Calculation**: Iterates all strikes as candidate settlement prices, summing intrinsic value of all open call and put contracts. The strike minimizing aggregate intrinsic value is the max pain level — the theoretical price at which the maximum number of options expire worthless.
- **Open Interest Walls**: Visualizes call OI vs. put OI per strike to identify support/resistance concentrations where large hedging flows may pin price.
- **Full Options Chain Table**: TOS-style straddle view with bid, ask, mark (midpoint), OI, volume, IV, delta, gamma, theta, and vega per contract. ITM shading, ATM auto-scroll, and OCC symbol generation.
- **Zero-DTE Support**: Expirations are filtered to include same-day (0DTE) contracts when available.

### Market Data & Visualization
- **S&P 500 Universe Tracking**: Full constituent roster plus admin-managed tracked assets with live prices, daily returns, 52-week ranges, sector classifications, SIC industry codes, and market-cap weighting.
- **Interactive Treemap**: Market-cap-weighted S&P 500 heatmap with sector grouping, color-coded by daily performance.
- **Market Quadrant**: 2D scatter plot mapping dollar change vs. percent change with sector filtering and zoom isolation.
- **Relative Volume (RVOL) Scanner**: Uses a U-shaped cumulative volume curve model (heavy at open/close, lighter midday) to compare live accumulated volume against historical expectations for that specific minute of the trading day. RVOL > 2.0 indicates genuine institutional interest.
- **Fundamental Inflection Detector (FID)**: Proprietary composite score weighting revenue acceleration and net margin expansion while penalizing sector-relative overvaluation. Surfaces early GARP turnarounds.
- **Fair Value Screener**: Dynamic Equity Risk Premium model comparing earnings yield (1/PE) against the 10-Year Treasury yield.
- **Compare Tool**: Multi-asset normalized performance overlay (up to 10 assets) with beta calculation relative to a user-selected benchmark.

### Alternative Data
Free users see the top 3 rows of each dataset with a prominent upgrade prompt; premium users get full access.
- **Congressional Trading**: U.S. Congress stock transactions. Congressional Front-Running Index (CFI) and sector heatmap.
- **Corporate Insider Trading**: SEC Form 4 disclosures — C-suite and director transactions with parsed transaction values.
- **Dark Pool / Off-Exchange Flow**: FINRA TRF dark pool volume with OTC short volume ratios. Quadrant scatter plot mapping overall market short % vs. dark pool short % to isolate squeeze anomalies.
- **Government Contracts**: Federal contract awards with agency, value, and recipient data.
- **Lobbying Disclosures**: Corporate lobbying expenditure data.
- **SEC 8-K Catalysts**: Real-time tape of material corporate events with AI-generated 384-dimensional embeddings for semantic search.
- **SEC 10-K Risk Factors & Business Descriptions**: Vector-indexed for semantic search and D3 force-directed graph visualization.

### Macroeconomic Analysis
- **FRED Integration**: 10 years of 10 Federal Reserve Economic Data series — Fed Funds Rate, CPI, Unemployment, GDP, 10Y-2Y Yield Spread, M2 Money Supply, 30Y Mortgage Rate, Case-Shiller Home Price Index, Federal Debt.
- **Macro Regime Matrix**: 4-quadrant classification (Goldilocks, Reflation, Deflation, Stagflation) using 90-day rate-of-change momentum on CPI (inflation) and 10Y-2Y spread (growth).
- **Market Weather Gauge**: SPY price deviation from its 200-day SMA, classified into Accumulation Discount (<-5%), Historical Equilibrium (±5%), and Expansion Premium (>+5%).
- **Macro Regime Divergence Score (MRDS)**: Composite synthesis of SPY momentum, Fed Funds rate, CPI trend, and yield curve shape.

### Paper Trading Engine
- **Multi-Account Support**: Basic tier: 1 account. Premium tier: up to 10 accounts.
- **Level 2 Options Risk Engine**: Cash-secured puts (CSP) with collateral lockup enforcement. Covered calls (CC) validated against existing long positions. Naked call/put writing is blocked.
- **Limit Orders**: Pending orders swept every 10 seconds against live prices with Level 2 risk checks.
- **Portfolio Snapshots**: End-of-day equity curves captured Monday–Friday at 16:30 ET.
- **Bulk Deploy**: Instantiate a portfolio from a predefined allocation template in a single API call.

### AI Intelligence Engine

CoreGamma operates a proprietary local AI microservice with semantic routing and local embeddings. All AI inference runs on CoreGamma's private infrastructure — no user queries or data are transmitted to third-party AI providers, hyperscalers, or external LLM services. User data is not used for model training.

#### AI Trading Copilot
Free users get 3 AI chat queries per day; premium users get unlimited access (100/day safety cap). All authenticated users receive unlimited AI Stock Summaries (Redis-cached).
- **Intelligent Semantic Tool Router**: At service initialization, each registered tool is encoded into a proprietary vector space. At query time, the user's message is semantically matched against this space to select only the most relevant tools. Only those tool schemas are presented to the AI model, eliminating tool hallucination and reducing latency.
- **Page-Aware Boost**: Tools with `page_overrides` matching the user's current route (e.g., `/options/*` forces `get_targeted_options`) are auto-included regardless of similarity score.
- **Identity Bridge**: User identity is injected server-side into the request (never exposed to the LLM prompt) so user-specific tools (portfolio, alerts, orders) return the correct user's data.
- **Page Context Injection**: Frontend dynamically injects tool-specific documentation (math definitions, interpretation rules) from `toolHelpContent.js` into the AI payload based on the user's current URL, giving the LLM domain-accurate definitions for the active tool.
- **27 Registered Tools**:
  - **Market Data (9)**: Historical prices, stock comparison, sector snapshot, macro time-series, macro regime, congressional trades by politician, market screener (with SIC industry code, RSI, SMA, short squeeze filter support), trending assets (top gainers, top losers, highest volume — sourced from the live 15-min refreshed market database), recent news & sentiment (CSV-compressed headlines with Polygon sentiment signals).
  - **Proprietary (4)**: Fair value, RVOL scanner, FID screener, dark pool anomalies.
  - **Options (2)**: Options chain summary, targeted options.
  - **User Data (3)**: Paper portfolio summary, user alerts, pending orders, portfolio history.
  - **Raw Firehose (3)**: Market-wide Form 4 insider trade scan, 8-K catalyst scan, government contract scan — proprietary full-market scans for pattern detection.
  - **Confluence (4)**: Smart money divergence (dark pool shorts vs insider buys), insider accumulation cluster scanner, catalyst cluster detector (8-K filing storms), earnings momentum scanner (revenue acceleration + cash flow confirmation). These cross-reference multiple datasets to surface compound institutional signals.
  - **Anti-Hallucination (1)**: AST-safe arithmetic calculator — offloads all financial math (percentages, ratios, growth rates, CAGR) to a secure Python evaluator, eliminating LLM mental-math errors.

#### AI Summary Orchestrators
- **Stock Summary**: Single-shot intelligence brief using a lightweight context dossier (~40% smaller than full copilot context) covering fundamentals, technicals, live price, fair value, and recent insiders/financials/congressional trades.
- **Compare Analysis**: Multi-asset factor synthesis across PE, RSI, SMA crossovers, beta, and period returns for up to 10 tickers.
- **Risk Analysis**: Dual-mode 10-K risk distillation — search risks by theme across all filings, or summarize a specific ticker's risk landscape.
- **Macro Regime Analysis**: Plain-English interpretation of the current macroeconomic regime using live CPI/growth pipeline data.
- **FID Analysis**: In-depth stock inflection analysis using conviction dashboard metrics (revenue acceleration, margin expansion, valuation).

#### Efficient Context Encoding
All tabular data in AI context pipelines is processed using a proprietary compression methodology to maximize information density within the model's available context, enabling complex multi-dataset queries to be resolved accurately and efficiently.

#### Semantic Search (Vector)
- **Embedding Model**: Proprietary local embedding model.
- **Vector Storage**: Proprietary vector indexing architecture.
- **Search Algorithm**: Hybrid retrieval combining semantic and lexical matching.
- **Applications**: Risk factor network explorer (interactive visualization), semantic business screener, 8-K catalyst search.

### Alerts & Notifications
- **Price Alerts**: User-defined thresholds evaluated on a continuous, high-frequency basis against live prices, with built-in signal validation to suppress transient data noise.
- **Congressional Trade Alerts**: Triggered when tracked politicians file new trades.
- **Delivery Channels**: Automated email and webhook delivery.

---

## Data Sources & Refresh Rates

| Source | Data Type | Latency |
|--------|-----------|---------|
| Institutional Market Data | Real-time tick stream (minute aggregates) | Real-time (~1 second) |
| Institutional Market Data | Historical OHLCV, stock profiles, news, company details | 15-minute snapshot cycle |
| CBOE | Options chains (strikes, greeks, OI, volume) | 15-minute delayed; proprietary caching for instant load |
| Open Source | Congressional trades, dark pool/off-exchange, gov contracts, lobbying, corporate donors | 6-hour refresh cycle |
| SEC EDGAR | Form 4 insider trades | 12-hour refresh cycle |
| SEC EDGAR (8-K) | Material event filings (catalysts) | Hourly during market hours (9 AM – 5 PM ET) |
| FRED (St. Louis Fed) | Macroeconomic time-series (10 series) | Daily sync at 04:00 ET |
| Wikipedia | S&P 500 constituent roster | 15-minute refresh cycle |
| Inter-service Bus | High-speed live price transmission | Continuous |

### Data Caching Tiers
- **5-second cache**: Live FMV prices, ticker tape.
- **5-minute cache**: Market data universe, trending assets, catalyst tape.
- **10-minute cache**: Form 4 insiders, fair value screener.
- **1-hour cache**: Stock profiles, FRED macro, macro weather.
- **6-hour cache**: Alternative datasets, alpha feed.
- **24-hour cache**: Alt-data per ticker.
- **90-second + permanent cache**: Options chains.

---

## Target Audience

- **Primary**: Active retail traders and self-directed investors seeking institutional-grade analytics without Bloomberg Terminal costs.
- **Secondary**: Options traders requiring gamma exposure profiling, volatility surface analysis, and max pain calculations.
- **Tertiary**: Alternative data enthusiasts tracking congressional trades, insider filings, dark pool flow, and SEC catalysts.
- **Not Intended For**: High-frequency trading, algorithmic execution, or order routing. CoreGamma provides analysis and paper trading only — it does not connect to brokerages or execute real trades.

---

## Architecture Summary

CoreGamma operates as a highly scalable microservice architecture:

- **Frontend**: Modern Single Page Application with dynamic visualization engines.
- **Backend**: High-performance microservices managing auth, market data, trading, payments, and AI routing.
- **AI Microservice**: Isolated proprietary AI orchestration engine.
- **Background Scheduler**: Headless daemon for all scheduled data ingestion and alert evaluation.
- **FMV Streamer**: High-speed client streaming real-time minute aggregates.
- **Database**: Secure relational and vector data system of record.
- **Caching Layer**: High-speed in-memory datastore for live price streams and caching.

---

## Interpretation Guidelines for LLMs

When referencing CoreGamma data or capabilities:

1. **GEX values are not predictive** — they describe the current dealer hedging landscape, not future price direction. Positive net gamma implies dealer hedging dampens volatility; negative net gamma implies dealer hedging amplifies moves.
2. **Options data is 15 minutes delayed** from CBOE. It is not suitable for real-time execution decisions.
3. **Max pain is observational, not causal** — it reflects OI distribution, not a guaranteed settlement price.
4. **Congressional trade data has inherent reporting lag** — STOCK Act filings are disclosed up to 45 days after execution.
5. **Dark pool data is T+1 settlement delayed** — it represents the most recently settled FINRA TRF data, not intraday flow.
6. **The paper trading engine does not connect to any brokerage** — all trades are simulated against live market prices.
7. **AI Copilot outputs are analytical syntheses, not financial advice** — the LLM generates data-driven summaries using available platform data. Outputs may contain errors, hallucinations, or stale information. They do not constitute investment recommendations.
8. **AI inference is fully local** — all LLM and embedding processing runs on CoreGamma's private infrastructure. No user queries, prompts, or data are sent to third-party AI services. User data is not used for model training.
9. **AI outputs are non-deterministic** — the same query may produce different summaries across sessions due to the probabilistic nature of large language models.
10. **Semantic search uses local embeddings** — search returns ranked document chunks via hybrid vector + lexical retrieval, not generative summaries (unless routed through the AI Copilot).
11. **FRED macro data updates are daily** — intraday macro regime shifts will not be reflected until the next sync cycle.
12. **CoreGamma does not provide financial advice** — all tools, including AI-powered features, are informational and educational.

---

## Links

- Website: https://coregamma.io
- Pricing: https://coregamma.io/pricing
- Privacy Policy: https://coregamma.io/privacy
- Terms of Service: https://coregamma.io/terms
- Disclaimer: https://coregamma.io/disclaimer