Skip to content

GeneralTradingSarl/computer_Agent

Repository files navigation

computer_Agent

Automation and trading experimentation platform built around open-source computer-use AI agents.

Python Docker Open Source

Overview

computer_Agent is an experimentation and integration project for AI-powered desktop and cloud automation. It is built on top of Agent S (Simular AI), an open-source, compositional generalist-specialist framework for computer-use agents, combined with:

  • Perplexity_agent, a local, self-hosted search and knowledge engine based on Perplexica, integrated as a knowledge module
  • Trading automation hooks for MetaTrader and custom Python trading bots
  • Evaluation and setup scripts for benchmarking on OSWorld-style tasks

What's original here vs. upstream

The underlying computer-use agent architecture, GUI grounding, and the benchmark results below (OSWorld, WindowsAgentArena, AndroidWorld) come from the Agent S research framework. Full credit goes to the Simular AI team; see their repository and papers for details. This repository's own contribution is the integration layer: wiring the agent framework to a trading and automation use case, bundling a Perplexica-based knowledge engine, and providing setup scripts for evaluation.

Reference benchmark results (Agent S2, upstream)

Benchmark Score Previous SOTA at time of release
OSWorld (15 step) 27.0% 22.7% (UI-TARS)
OSWorld (50 step) 34.5% 32.6% (OpenAI CUA)
WindowsAgentArena 29.8% 19.5% (NAVI)
AndroidWorld 54.3% 46.8% (UI-TARS)

Results as published by the Agent S2 paper; not independently reproduced in this repository. See the upstream project for the latest Agent S3 results, which surpass these numbers.

Use cases explored in this repo

  • Automating trading strategies and risk management workflows
  • AI assistants for research, finance, and productivity, grounded with a local knowledge engine
  • LLM-driven document analysis and workflow automation
  • GUI and API automation for desktop and cloud environments

Setup

See the upstream Agent S installation guide for the base framework, plus the docs and scripts in this repository for the trading and knowledge-engine integrations.

Tags

ai-agents automation llm computer-use trading-bot python open-source agent-s

Contact

About

Automation and trading experiments built on Agent S2 (simular-ai), an open-source computer-use agent framework: MetaTrader bot integration, Perplexica-based knowledge engine, OSWorld evaluation setup.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages