traca

AI-powered trading analyst that combines real-time market intelligence, behavioural coaching, and social media automation.

Powered by Deriv API, traca serves both retail and prop firm traders through a unified interface that turns complex market data into actionable understanding.

MVP Link: https://frontend-adnancantfunctions-projects.vercel.app?_vercel_share=0WLCBmlikBqEotQV5J9MVCRryghRWuEK

🚀 Overview

traca is a split-view web platform featuring dashboard panels alongside a conversational AI analyst. It provides:

• Instant Market Intelligence: Plain-language explanations of price movements and technical patterns.
• Behavioural Coaching: Detection of emotional or impulsive trading patterns (FOMO, revenge trading) with timely nudges.
• Social Automation: Autonomous AI personas that generate and draft market updates for LinkedIn and X.

✨ Core Features (MVP)

• Real-time Market Insights: Integration with Deriv API across Forex, Crypto, and Stock markets.
• Conversational AI: Unified chat handling market queries, behavioural feedback, and social content creation in one place.
• Split-View Dashboard: Side-by-side view of live price charts, sentiment gauges, and the AI analyst chat.
• Behavioural Pattern Detection: Identification of win/loss streaks and risk escalation with habit-building reinforcement.
• Social Content Drafting: Platform-appropriate content generation (Professional for LinkedIn, concise for X).

🏗️ Architecture

traca uses a Modular Monolith architecture optimized for real-time data flow and AI inference:

• Frontend: React (Vite) + TailwindCSS + shadcn/ui + Zustand.
• Backend: FastAPI (Python) + Uvicorn + WebSockets.
• AI Engine: Mistral API (cloud LLM).
• Data Source: Deriv API for real-time pricing and trade history.
• Persistence: SQLite for trade history, session-based chat memory, and content drafts.

🛠️ Tech Stack

Layer	Technology
Backend	FastAPI, Uvicorn, websockets, deriv_api, pandas-ta
Frontend	React 19, Vite, TanStack Query, Zustand, Recharts, Lightweight Charts
AI/ML	Mistral API, Pydantic v2
Styling	TailwindCSS 4, shadcn/ui

🚦 Quickstart

Prerequisites

• Python 3.11+
• Node.js & npm
• Mistral API key
• Deriv API Token

1. Backend Setup

cd backend
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
# Set DERIV_API_TOKEN in .env
uvicorn main:app --reload

For hosted frontends (e.g. Vercel), set CORS origins in backend/.env:

ALLOWED_ORIGINS=https://your-vercel-app.vercel.app

Use ALLOWED_ORIGINS=* for a public demo.

Set a JWT secret in backend/.env for auth:

JWT_SECRET=change_me
JWT_ALGORITHM=HS256
JWT_EXPIRE_MINUTES=1440

WebSocket now requires auth. The frontend sends the token automatically once you log in or register.

2. LLM Setup

Set these environment variables (see backend/.env.example):

MISTRAL_API_KEY=your_mistral_api_key_here
MISTRAL_MODEL=mistral-small-latest
MISTRAL_BASE_URL=https://api.mistral.ai/v1

3. Frontend Setup

cd frontend
npm install
npm run dev

If the frontend runs on a different host/port than the backend, set:

VITE_WS_URL=ws://127.0.0.1:8000
VITE_API_URL=http://127.0.0.1:8000

See frontend/.env.example.

⚠️ Key Pitfalls & Prevention

• API Rate Limits: Deriv API has rate limits; subscription model and error handling are managed to avoid being blocked.
• LLM Hallucinations: Numbers are parsed programmatically from Deriv API; the LLM is used only for narrative explanation.
• Render Performance: High-frequency price updates are managed via Zustand to avoid React Context re-render storms.

🗺️ Roadmap

• Persistent cross-session AI memory.
• Multi-modal input (Analyze chart screenshots).
• Automated regulatory compliance filters for social posts.
• Mobile-native responsive application.

---

Built with ❤️ using Deriv API.