‹ Back to projects

Argus

A desktop app for understanding a market before betting on it.

Argus combines Google Research's TimesFM 2.5 temporal forecasting model with a multi-agent pipeline of language models that debate the qualitative analysis, and a VectorBT backtester that checks every decision against mathematical evidence before it goes live.

On top of the analysis engine sits a full Portfolio Manager, integrated with CCXT to generate and execute orders on derivatives exchanges, with an institutional-grade money management framework and a fully autonomous auto-trading scheduler.

Today the trading automation covers BTC only: the data layer is multi-asset, but nothing else is traded autonomously yet — a limitation the project's own documentation states plainly.

  • Platform
    Desktop — Windows, macOS, Linux
  • Engine
    TimesFM 2.5 + multi-agent pipeline
  • Interface
    Six panels in a single window
  • Scope
    Automated trading on BTC only
  • License
    AGPL v3 (commercial license available)
  • Starts or stops the automated trading cycle, with a countdown to the next candle and a run log
  • Manages spot and futures positions via CCXT: leverage, P&L, stop loss and take profit as ROI%
  • Runs Pattern Matching (KNN) against BTC history, with a confidence score and target price
  • Runs a TimesFM 2.5 temporal forecast on the same historical series
  • Runs the multi-agent LLM pipeline and shows the full debate behind every decision
  • Exports backtests and analysis to Excel