WebBetas

SocialPilot

WebApp

SocialPilot

InvestIQ is a self-hosted investment intelligence platform that combines live market data with local AI to surface trading opportunities in real time. Built on PHP, MySQL, and Ollama (running locally to eliminate API costs), the platform continuously monitors hundreds of stocks across Yahoo Finance screeners and delivers actionable signals, alternative picks, spike predictions, and detailed deep-scan analyses through a dark trading terminal interface. The platform generates BUY/WATCH/AVOID signals with confidence scores, entry zones, price targets, and stop-loss levels. It tracks spike predictions using 1-minute candle analysis — detecting Bollinger Band squeezes, volume buildups, and RSI extremes before a move happens. A Top Picks system generates one curated recommendation per strategy type (best opportunity, safest trade, aggressive, options play, crypto, ETF, short-term, swing). All analysis runs on a local Ollama instance, keeping costs at zero regardless of usage volume.

  • Domain - invest.webbetas.com

  • Year - 2026

  • Type - WebApp

  • Industry - Technology

Challenge

Most retail investors either rely on expensive subscription services with generic recommendations, or spend hours manually scanning charts. The goal was to build something that felt like having a quant analyst running 24/7 — scanning the entire market, identifying setups the user wouldn't know to look for, and presenting them clearly enough to act on immediately. The core technical challenges were: keeping latency low while Ollama generates multi-minute analyses, ensuring live prices are always accurate (Yahoo Finance returns split-adjusted historical data that diverges from current prices), preventing duplicate signals across runs, and building a polling architecture that lets the browser show results as they arrive without holding HTTP connections open long enough for nginx to time them out.

Solution

All AI jobs are launched via nohup as fully detached background processes — completely decoupled from the HTTP request lifecycle. The frontend fires a fire-and-forget fetch to run.php, then polls the database every 5 seconds for new results, showing them as they arrive. This means nginx timeouts are irrelevant; the process runs until completion regardless of browser state. Live prices are resolved client-side: the same fetchChartData() call that draws the chart also returns Yahoo's regularMarketPrice, which immediately overwrites any stale AI-generated price on the card. Charts and prices always match because they come from the same source. Signal deduplication prevents back-to-back runs from re-analysing the same stocks — each run excludes symbols already signalled in the last two hours and moves to fresher opportunities. The spike predictor cross-references recent spike events and skips symbols that already moved, keeping predictions genuinely forward-looking. The result is a platform that scans 150+ live market symbols every cycle, generates technically-grounded signals with specific indicator references, and presents everything in a terminal-style UI designed for fast decision-making — running entirely on self-hosted infrastructure at zero ongoing cost.
Projects Projects
  • ⌘ It is a long established fact that a reader will distracted by the readable content of a page when looking an its readable ⌘
  • ⌘ It is a long established fact that a reader will distracted by the readable content of a page when looking an its readable ⌘
  • ⌘ It is a long established fact that a reader will distracted by the readable content of a page when looking an its readable ⌘
  • ⌘ It is a long established fact that a reader will distracted by the readable content of a page when looking an its readable ⌘
2024 2024

J.V. Hilliard

Author

2022 2022

GuildsVR

Social Network

2014 2014

BookPressed

Social Network