Octopus Trading Platform

AI-powered trading platform with real-time analytics and 11 orchestrated agents

Features · Agents · Installation · Docs

---

Overview

Octopus is an AI-powered trading system that combines real-time market data, analytics, ML models, and automated trading in a single interface. The backend coordinates 11 AI agents (M1–M11) for data collection, strategy, risk, sentiment, and reporting.

---

System architecture

flowchart LR
    subgraph You["🖥️ You"]
        UI[Next.js]
    end

    subgraph Octopus["🐙 Octopus"]
        API[FastAPI]
        Agents[11 Agents]
        API --> Agents
    end

    subgraph Store["💾 Store"]
        DB[(PostgreSQL)]
        Cache[(Redis)]
    end

    UI <-->|REST · WS| API
    Agents --> DB
    Agents --> Cache

Loading

Data in → Agents think → You decide.

---

AI Agents

The platform uses 11 orchestrated agents (M1–M11) with distinct roles and personas. Each has a character name used across the UI (Command Center, Risk, Reports, etc.).

ID	Character	Role	Responsibility
M1	Nexus	Data Collection	Market data, news, alternative data pipelines
M2	Vault	Data Warehouse	Storage, validation, historical datasets
M3	Pulse	Real-time Processor	Streaming data, live analytics, alerts
M4	Atlas	Strategy Agent	Signals, strategy execution, backtesting
M5	Neuron	ML Models	Prediction, classification, deep learning
M6	Guardian	Risk Management	VaR, position sizing, compliance
M7	Oracle	Price Prediction	Time-series and price forecasting
M8	Shadow	Paper Trading	Simulated execution, paper portfolio
M9	Echo	Market Sentiment	News and social sentiment analysis
M10	Chronicle	Backtesting	Historical testing, strategy validation
M11	Lens	Visualization	Charts, dashboards, report insights

Pipeline flow: Data (M1, M3, M9) → ML & prediction (M5, M7) → Risk & strategy (M6, M4) → Backtest & viz (M10, M11). The orchestrator routes tasks via submit_task() and runs full pipelines via coordinate_pipeline().

---

Features

• Dashboard – Portfolio overview, market watchlists, live data
• Command Center – Order entry, positions, bots, options
• Options – Options chain and strategies
• Portfolio & Risk – Multi-asset tracking, VaR, stress tests
• Strategies & Backtesting – Strategy builder and historical backtests
• AI Models – Training, predictions, insights
• Reports & Visualization – AI-powered reports and charts
• Platform search – Search anything (pages, commands) via ⌘K

---

• Repo layout: docs/PROJECT_STRUCTURE.md
• Development log (zero to now): docs/DEVELOPMENT-LOG.md

Installation

Prerequisites

• Node.js 18+, Python 3.10+, PostgreSQL 14+, Redis (optional)

Quick start

# Clone
git clone https://github.com/massoudsh/Findash.git
cd Findash

# Backend
python -m venv venv
source venv/bin/activate   # Windows: venv\Scripts\activate
pip install -r requirements/requirements.txt

# Frontend
cd frontend-nextjs
npm install

Run

# Terminal 1 – backend
python3 start.py --reload

# Terminal 2 – frontend
cd frontend-nextjs && npm run dev

• Frontend: http://localhost:3000
• API docs: http://localhost:8000/docs

Run with Docker (core stack)

The core stack runs API, frontend, PostgreSQL, Redis, Celery worker/beat, Prometheus, and Grafana. Redis is exposed on port 6380 by default to avoid conflict with a local Redis on 6379.

# Core only (no LLM services)
docker compose -f docker-compose-core.yml up -d

# Optional: run a smoke test after bring-up
./scripts/healthcheck-core.sh

# Or use the interactive service manager:
./scripts/start-services.sh

• API: http://localhost:8011 (mapped from 8000 in container)
• Frontend: http://localhost:3000
• Grafana: http://localhost:3001

Run with Docker + LLM profile (optional)

LLM services (Falcon TGI, FinGPT inference) are under the llm profile. Use them for report generation; see docs/llm-report-models.md.

# Core + LLM services (TGI Falcon, FinGPT inference)
docker compose -f docker-compose-core.yml --profile llm up -d

Set in your env (or .env): FALCON_TGI_URL=http://localhost:8080, FINGPT_LOCAL_URL=http://localhost:8081, and optionally HF_TOKEN for HuggingFace. See config/env.example for all LLM variables.

For production, use a docker-compose.override.yml (or env file) to set ENVIRONMENT=production, secure secrets, and correct database/Redis hosts.

---

Documentation

• API: Swagger at /docs, ReDoc at /redoc
• Architecture: See docs/ARCHITECTURE_DIAGRAMS.md and docs/orchestrator-architecture.md for detailed diagrams
• Agent–user decision workflow: From market updates to reports — docs/AGENT_USER_DECISION_WORKFLOW.md (Aladdin-style)

---

Tech stack

Layer	Stack
Frontend	Next.js 14, TypeScript, Tailwind CSS, Shadcn UI
Backend	FastAPI, Python 3.10+
Data	PostgreSQL (TimescaleDB), Redis
ML/AI	PyTorch, scikit-learn, Celery workers

---

License

MIT – see LICENSE.