ai-gap-trading-forecaster
RL Forecasting Engine & Adversarial UI Testbed
Overview
gap-trader is a dual-purpose engineering project that bridges financial time-series forecasting with ML-powered quality systems. On the surface, it is a gap trading assistant that scans for price gaps, manages watchlists, and tracks trades over time.
Beneath the surface, it functions as an Adversarial UI Testbed. The frontend is intentionally volatile — DOM structures, CSS classes, and component hierarchies are designed to mutate over time. This controlled chaos simulates the kind of UI drift seen in fast-moving product teams and provides a realistic environment to train and validate the self-healing Playwright locators and LLM+RAG test generation pipeline implemented inai-test-gen.
Core Objectives
Time-Series Gap Forecasting
Scan for gap-up / gap-down patterns and experiment with RL-style decision policies over financial time-series.
Adversarial DOM Mutation
Programmatically shift locators, component hierarchies, and CSS classes to simulate UI drift in real products.
ML Pipeline Validation
Provide a closed-loop environment to measure recovery rate, latency, and stability of the ai-test-gen self-healing test suite.
Human-in-the-Loop Trading Sandbox
Allow manual oversight of RL-style signals while capturing rich telemetry for both trading and QA experiments.