Case Study · Agentic AI Architecture

Aura: The Agentic AI Swing-Trade Co-Pilot

Solving the emotional trading problem for retail investors with a hyper-logical, RAG-powered quantitative analyst built with Angular and Go.

2026 Gemini 3.1 Pro Integrated Vercel Monorepo
Live
RapidAPI Yahoo Finance Data
Strict
JSON AI Output Enforcement
<2s
Go Serverless Execution

Background

Retail stock investors often fall victim to emotional trading, particularly when executing short-term swing trades or BSJP (Beli Sore Jual Pagi) strategies on the Indonesian Stock Exchange (IDX). They rely on scattered, unstructured chat-bot outputs which fuel intuition over logic.

I built Aura to be a hyper-logical, risk-averse Lead System Architect & Quantitative Analyst. She speaks in technical certainties and outputs structured data, completely removing conversational fluff and emotional bias from trading.

"The goal was to force an LLM to output structured quantitative setups rather than conversational text, forcing the user into a logical trading framework."

Challenges: Combating Emotional Trading

🧠

Emotional Bias

Investors make irrational decisions based on market panic or hype. They need cold, hard quantitative limits (Entry, Take Profit, Stop Loss).

📉

Unstructured Data

Standard LLMs output paragraphs of text which are hard to parse quickly during fast-moving trading hours.

⏱️

Delayed Analysis

Flipping between Yahoo Finance, charting tools, and ChatGPT takes too long for BSJP strategies where minutes matter.


The Solution: Agentic RAG Workflow

Aura is built on an agentic RAG workflow. Instead of acting as a standard chatbot, she pulls live market data via RapidAPI behind the scenes, analyzes it using Google's Gemini 3.1 Pro, and enforces a strict JSON output schema.

💻
User QueryAngular UI
Go Serverless APIFetch RapidAPI Live Data
🧠
Gemini 3.1 ProAgentic Analysis
📊
JSON Strategy CardRendered on UI
System Preview
Dark Mode Quantitative Strategy UI
A highly logical interface showing exact Entry, Take Profit (Green), and Stop Loss (Red) values generated by the AI based on live data.
Aura Trading Copilot Screenshot

Architecture & Technologies

Aura uses a Vercel-optimized Monorepo to maintain security and performance, bridging a frontend SPA with serverless backend functions.

Angular 18
Go (Golang)
Vercel Serverless Functions
Gemini 3.1 Pro API
Tailwind CSS (Slate/Neon)
RapidAPI (Yahoo Finance)