Skip to content

Home

Jump to bottom
Kingler16 edited this page Apr 20, 2026 · 3 revisions

Velora Wiki

Velora is a self-hosted AI wealth advisor that runs on a Raspberry Pi or a small Linux box and talks to you through a web dashboard, a floating AI chat, and a Telegram bot. It uses Claude Code CLI under the hood, so there are zero API costs for the AI — an existing Claude Pro/Max subscription is enough.

Highlights

  • Web Dashboard with 8 pages (Dashboard, Portfolio, Analysis, Markets, Briefings, Recommendations, Chat, Settings) built on FastAPI, Jinja2, HTMX and vanilla CSS.
  • Liquid Glass UI v2 — cyan/indigo/violet theme, grainy blur blobs, auto dark/light via prefers-color-scheme, Inter + Geist Mono fonts.
  • AI Chat everywhere — full-page /chat plus a floating widget on every page, powered by an MCP server with 16 tools, streaming responses and a confirmation flow for write actions.
  • Command Palette (Cmd+K) for navigation and quick actions (log trade, refresh, new chat).
  • Multi-step Trade Modal (action → amount → preview) with live total, currency toggle and toast feedback.
  • Interactive Charts — ApexCharts donuts, range-area, heatmap and treemap plus TradingView Lightweight Charts for net-worth history; inline SVG sparklines in holdings tables.
  • Briefings — AI-generated daily/weekly reports delivered to Telegram and the web, with Markdown and syntax-highlighted code blocks.
  • Recommendations Engine with BUY/SELL/WATCH/HOLD calls, automatic outcome detection and hit-rate tracking.
  • Multi-Account Portfolio across Trade Republic, Erste Bank, Interactive Brokers, Flatex and Scalable Capital, with EUR/USD conversion and tax-loss-harvesting hints.
  • System Self-Update — one-click git fetch + reset --hard + systemctl restart from the settings page; portfolio data is preserved.
  • i18n — German and English, auto language detection for chat, briefings and UI.

Pages

Quick Start

git clone https://github.com/Kingler16/Velora.git
cd Velora
python3 setup.py
python3 -m src.main web

Then open http://localhost:8080.

Tech Stack

  • Python 3.12 runtime
  • FastAPI + Jinja2 + HTMX for the web layer
  • Vanilla CSS — the Liquid Glass design system, no framework
  • ApexCharts + TradingView Lightweight Charts + highlight.js for visuals and code rendering
  • MCP Server exposing 16 tools that the in-browser chat calls
  • Claude Code CLI as the AI backend (OAuth via Pro/Max, no API key required)
  • Data sources: yfinance, FRED, ECB, Brave Search, Bloomberg RSS, Finnhub
  • Self-hosted on Raspberry Pi / RockPi, around 90 MB RAM at rest

Screenshots

Current screenshots live on the landing page: https://kingler16.github.io/Velora/

Quick Links

License

MIT

Clone this wiki locally