Screen-Reading Automated Trading Agent

I need an AI-driven assistant that watches my live market charts, recognises when my predefined rules are met, and then executes the corresponding order on a web-based broker platform without asking for confirmation. Here is the flow I have in mind: the agent captures the chart window in real time, applies computer-vision/OCR techniques to read price, indicators and any on-screen markers I specify, checks these inputs against the entry, exit and risk parameters I will hand over in a rule set, and, once conditions line up, submits the exact order (size, direction, stop, target) directly through the broker’s web interface or API. After entry it should also be able to modify or close positions automatically if the same rule set demands it. Robust logging of every action and a simple dashboard or log file for audit are essential so I can review what the agent is doing while I’m away. I will provide: • The full trading rule book, indicator descriptions and screenshots for calibration • Credentials or sandbox access to the broker’s web platform • A test script of historical scenarios the agent must pass before we go live Acceptance criteria: 1. 100 % correct detection of the indicator and price conditions in my test scenarios 2. Order submission, modification and closure on the web platform in under one second after the trigger fires 3. Clear error handling—for example, if the broker rejects an order, the agent retries or notifies me in the log Code can be in Python (TensorFlow, OpenCV, Selenium, or a lightweight headless browser driver—whatever achieves stability and speed), but it must run on Windows and be straightforward for me to launch each day. A concise README and recorded demo will complete the hand-off.