NeuroTide

Autonomous AI Torrent Engine

Comment

Inspiration

Traditional torrent clients are often slow, require full downloads before playback, and leave residual files that compromise privacy. We wanted to create a solution that not only streamlines streaming but also uses AI to make autonomous decisions, protects user privacy, and allows seamless multi-device experiences. The idea came from combining our interest in agentic AI, P2P networks, and privacy-focused systems.

What it does

NeuroTide is an AI-powered torrent engine that transforms file sharing into an intelligent, adaptive, and private experience. It: Streams video or audio in real-time, without waiting for full downloads. Uses a reinforcement learning AI agent to autonomously prioritize torrent pieces for smooth playback. Offers zero-disk RAM-only streaming to ensure complete privacy. Supports multi-device handoff, letting users continue sessions seamlessly on PC or mobile. Includes a plugin ecosystem for subtitles, summarization, virus scanning, and more.

How we built it

We began by developing the core asynchronous torrent engine in Python using asyncio, enabling high-performance streaming. Then, we implemented an AI reinforcement learning scheduler to dynamically prioritize pieces based on network conditions and swarm health, creating an autonomous streaming experience.

For privacy, we designed a RAM-only mode, ensuring ephemeral sessions with zero residual data. We also built a multi-device handoff system using QR codes and NFC so users can seamlessly switch devices without losing progress. To extend functionality, we created a plugin framework, allowing features like auto-subtitles, virus scanning, and content summarization.

Throughout development, we iterated on the architecture, ran performance tests, and refined the AI scheduler to balance efficiency, autonomy, and usability.

Challenges we ran into

Designing the reinforcement learning scheduler to adapt in real time to varying network conditions. Implementing zero-disk streaming while maintaining smooth playback. Ensuring multi-device session continuity without interrupting downloads or AI decisions. Balancing asynchronous performance with AI computations and plugin extensions.

Accomplishments that we're proud of

Built a fully autonomous AI agent capable of intelligent torrent piece selection. Implemented privacy-first, zero-disk streaming for ephemeral sessions. Enabled seamless multi-device handoff, a feature rarely seen in torrent clients. Created a flexible plugin ecosystem for extensibility and future growth.

What we learned

How to apply reinforcement learning in real-time decentralized networks. Best practices for asynchronous Python programming in high-performance streaming applications. Techniques for designing privacy-aware P2P systems with ephemeral sessions. The importance of clear documentation and UX when building complex AI systems.

What's next for NeuroTide

Train and enhance the AI prefetcher for more predictive and efficient piece selection. Implement encrypted swarm communication for enhanced security. Develop a mobile companion app for full multi-device support. Expand the plugin marketplace for community contributions and additional functionality.

Built With

  • aiohttp
  • asyncio
  • bitstring
  • dht
  • fast-bencode
  • httpstreaming
  • magnetlinks
  • onxx
  • pyclamd
  • pytest
  • python-3.9+
  • python-qrcode
  • pytorch
  • requests
  • rich
  • starlette
  • udp
  • uvicorn
  • webrtc
Share this project:

Updates