This project provides a sophisticated system for generating, mining, and submitting alpha signals to the WorldQuant platform. It targets quantitative researchers and developers seeking to automate and accelerate their alpha discovery process. The primary benefit is a significant improvement in generation speed and a reduction in operational costs by leveraging local LLMs and GPU acceleration.

How It Works

The recommended "Naive-Ollama" approach utilizes local Large Language Models (LLMs) served via Ollama, such as Llama 3.2:3b or Llama 2:7b, to ideate alpha expressions. This is coupled with NVIDIA GPU acceleration for faster inference. The system orchestrates a workflow involving continuous alpha generation, intelligent mining of promising expressions for variations, and compliant submission to WorldQuant Brain, all managed through a Flask-based web dashboard for real-time monitoring and control.

Quick Start & Requirements

• Primary install/run command: docker-compose -f docker-compose.gpu.yml up -d (for GPU support).
• Prerequisites: WorldQuant Brain credentials (saved in naive-ollama/credential.txt), Docker, NVIDIA GPU with CUDA support.
• Access: Web Dashboard at http://localhost:5000, Ollama WebUI at http://localhost:3000.
• Links: Discord: https://discord.gg/3B2TmHQw, YouTube Tutorials: https://www.youtube.com/watch?v=xwr9atsulSA, https://www.youtube.com/watch?v=EAeujBRrKiI.

Highlighted Details

• Achieves 3-5x faster alpha generation compared to previous Kimi API integrations.
• Employs local LLM processing via Ollama, eliminating external API costs and enhancing data privacy.
• Features a real-time web dashboard for comprehensive monitoring of GPU, Ollama, orchestrator, and WorldQuant API statuses, along with manual controls.
• Supports fully automated 24/7 operation for continuous alpha discovery and submission.
• Includes Docker support for streamlined deployment and management of services.

Maintenance & Community

The project encourages community contributions via GitHub issues and pull requests. A Discord server is available for community chat and support. Several features are marked as "TODO" or "Incoming," indicating ongoing development.

Licensing & Compatibility

This project is licensed under the Apache License 2.0. It also integrates components from Dify (agent-dify-api and agent-dify-web), which are also licensed under the Apache License 2.0, ensuring compatibility for commercial use and closed-source linking under the terms of the license.

Limitations & Caveats

Some tutorials are marked as "WIP" (Work In Progress). The successful_alpha_submitter.py script is currently not recommended for use as it submits alphas in a single batch rather than adhering to WorldQuant's daily limits. Certain older components like "Pre-Consultant" have limitations on concurrent simulations and available operators/data fields.