Account Information
My Account
-
Phone is optional while password-only mode is active. Operational alerts will use Slack after setup.
SMS OTP login is mandatory in this version and cannot be disabled.
Change Email
New email must first be one of the two allowed emails.
Change Password
System Overview
Trading Execution Policy
No real buy/sell order is placed by this website. The system only collects data, trains/evaluates models, monitors signals, sends alerts, and shows charts. Humans place all trades manually.
Health
-
Data Policy
-
Storage Usage
-
Company Option Lab
Company Search
Browse the full 500-symbol research universe. Search ranks ticker/company names that start with your text first, then other matches.
Multi-company Option Lab
Compare four companies side by side. Each chart can use a different ticker and contract. Auto refresh reloads the latest GCS data as the 1-minute collector saves new files. Y-axis is option mid price ($); contract value is roughly mid × 100.
-
-
-
-
Storage Manager
Retention Policy
-
Delete Preview Only
This does not delete data. Real web deletion is disabled by default.
-
Data Explorer & Export Center
Filters
Excel export is for filtered data only. Large Top 500 archives should stay in GCS as parquet.
Statistics Summary
-
Preview
Actual Option Price Chart
Uses collected actual data when available. Prediction overlay will appear after model module is added.
Trading Desk
Option Trade / Mid Price Monitor
Trader-style page: the large chart shows collected option trade/mid prices now. Model overlay will appear only after trained prediction files exist. Lower cards show HEAD(3) from live datasets.
Latest option rows
A quick table from the same collected OPRA option rows used by the Trading Desk chart.
Live Source HEAD(3)
Each card shows the first 3 readable rows from the most recent saved GCS files for that category/source. This makes it easy to confirm what is being collected right now.
Trading desk technical metadata
-
Quant Data Lake
Collected Data Browser
Click a collector card to inspect what is actually saved in GCS. Preview rows show in an Excel-like table, with quick numeric summaries and a simple chart when columns are available.
Selected data
Export uses the selected category, ticker, date, and row limit. Excel includes a Summary sheet, Numeric Summary sheet, and Data sheet.
Quick charts & numeric summary
For Option Features / IV Greeks, the dashboard shows five fixed metric charts: mid, implied volatility, delta, theoretical price, and spread percent.
Technical metadata
-
Text-to-Feature Engine
Convert language data from news, SEC filings, earnings, macro events, and provider feeds into model-ready numeric features. This creates point-in-time text event rows and ticker/window aggregate rows that can be merged with options, price, IV, volume, and macro data.
Model-ready aggregate features
Normalized text event rows
Text feature technical metadata
-
Delete data by category/date
Preview works by default. Real delete requires ALLOW_WEB_DELETE=true and exact confirmation.
-
API Key Manager
Live Provider Health
Keys are stored backend-only in Secret Manager. This page now separates key registration, live tests, collector/data status, and renewal guidance.
Due soon details
-
Save / Activate provider key
ThetaData note: select ThetaData, paste your API token, then click Activate key. Scheduled option collectors will try ThetaData first and automatically fall back to yfinance if ThetaData is unavailable.
If expires_at is blank, collection keeps using this key until you activate a replacement. If expires_at is set, collection stops after that date.
Provider notes / renewal guidance
Obtained: FRED, Alpha Vantage, Finnhub, EIA, BLS, Nasdaq Data Link, FMP, Tiingo, Marketstack, NewsAPI.
SEC: no API key. Use SEC EDGAR User-Agent provider and Activate/Test with blank value.
Marketstack: Basic 10,000 requests/month.
BLS: registered v2.0, 500 queries/day.
Rule: Activate = backend-only storage + scheduled collectors may use it. Renewal guidance is shown in the status table; if the provider does not publish fixed key expiry, this app shows a safe review cadence.
Data Health Monitor
Pipeline status
Verify that OPRA, feature tables, underlying bars, market, macro, news, and support datasets are actually arriving on schedule. Freshness is session-aware, so overnight/weekend gaps are not automatically treated as failures.
Click a dataset row to inspect its latest files, collector log, scheduler, Cloud Run execution, and warnings. Row counts come from parquet metadata when available; full OPRA files are not loaded into memory just for monitoring.
Selected dataset diagnosis
Alert Center
Slack Alerts
Twilio/SMS is not required. Paste a Slack Incoming Webhook URL below, save it backend-only, then test here. Alerts are for collector failures, fallback use, data-stall warnings, Options empty warnings, and daily summaries.
This URL is stored backend-only in Secret Manager as slack_webhook. It will not be displayed again after saving.
-
Alert rules prepared
- Fallback provider used → Slack warning and retry on next scheduled run.
- Options data remains empty → Slack warning.
- Collector/API errors → Slack warning after backend detects failure.
- Daily summary can be enabled later through Scheduler.
Scheduler-first Automation
Option Data Collection Control
ACTIVATE arms the backend. Existing Cloud Scheduler option jobs will keep collecting at their scheduled times until you deactivate. DEACTIVATE uses admin password confirmation while Emergency Password Mode is ON. Twilio is no longer required; operational alerts move to Slack.
-
Manual run once for testing only
Real production collection should run from Google Cloud Scheduler even when you are traveling.
-
Deploy scheduler script
scripts/powershell/06_Create_Quant_Schedulers_v20.ps1
This creates Cloud Scheduler jobs for options, market data, news windows, macro, earnings, and model review SMS.
Admin Settings
Allowed Signup Emails
Only these two emails can create or use accounts. Owner only can update the list.
-
Security Audit Logs
Backend/GCS record of account changes, login attempts, OTP events, password resets, and allowlist changes. Passwords and OTP codes are never stored.
Company Model Report
Model Validation Console
Train a real research-only model from the data currently available, verify t+5 labels, candidate training, OOS prediction, actual resolution, graph, and report. Validation models can never replace a production Champion.
Time-ordered final holdout only. No fabricated points.
Full E2E validation record
-
Production Training Console
After sufficient history is available, queue strict production training from the website. During regular market hours the request waits for the low-cost after-market trainer. Existing Champions remain active until a candidate passes the promotion gate.
Model Training & Readiness
Training completion is not the same as production readiness. A model is counted as READY only when a validated champion is available.
| Ticker | Workflow status | Progress | Production champion | Version | Message |
|---|---|---|---|---|---|
| Loading… | |||||
Dual Retraining Framework
Intraday monitoring detects material prediction divergence and queues work for after-market processing. Saturday governance handles normal weekly review. Both routes preserve the current champion until a challenger passes validation.
Retraining Decision Board
Governance recommendations for the 500-company universe. Each list shows roughly 15 rows at once and scrolls internally for the remaining models.
Retraining Required 0
| Ticker | Decision | Direction | Drift | Overfit | Primary reason | |
|---|---|---|---|---|---|---|
| Loading… | ||||||
No Scheduled Retraining 0
| Ticker | Decision | Direction | Drift | Overfit | Primary reason | |
|---|---|---|---|---|---|---|
| Loading… | ||||||
Company-Specific Model Governance
Choose one company to review its accuracy, uncertainty, feature selection, retraining reason, model architecture, and recent prediction-vs-market behavior.
5-Trading-Day Actual vs Model Prediction
Select a company. The chart uses resolved point-in-time evaluation rows and never fabricates predictions.
Company-Specific Feature Governance Policy
The data lake may collect many sources, but no company model is forced to use all of them. Features and model families are admitted ticker-by-ticker using point-in-time out-of-sample validation. Candidate changes are tested before champion promotion.
-