Skillget

Titans---Learning-to-Memorize-at-Test-Time

agent
Security Audit
Pass
Health Pass
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Community trust — 69 GitHub stars
Code Pass
  • Code scan — Scanned 10 files during light audit, no dangerous patterns found
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

Titans - Learning to Memorize at Test Time

README.md

Titans Demonstration Platform

image

This project demonstrates ideas from the paper Titans: Learning to Memorize at Test Time using seven specialized AI agents and a native desktop interface.

Paper: https://arxiv.org/pdf/2501.00663v1

What It Runs

The primary app is now a native desktop UI:

  • Entry point: main.py
  • Windows executable: titans.exe
  • Launcher: titans.bat
  • UI: agent selection, demonstration runs, collaborative insights, live metrics, and real-time visual interaction

Agents

  1. OpenAI Agent (Neural Memory Module)
  2. Anthropic Agent (Memory as Context)
  3. Mistral Agent (Memory as Gate)
  4. Groq Agent (Memory as Layer)
  5. Gemini Agent (Experimental Validation)
  6. Cohere Agent (Innovations)
  7. Emergence Agent (Analysis)

Setup

  1. Install dependencies:
pip install -r requirements.txt

Copy .env.sample to .env and fill in the API keys you plan to use.

  1. Configure optional provider keys in .env:
  • OPENAI_API_KEY
  • ANTHROPIC_API_KEY
  • MISTRAL_API_KEY
  • GROQ_API_KEY
  • GOOGLE_API_KEY
  • COHERE_API_KEY
  • EMERGENCE_API_KEY

Note: The desktop app starts even if some providers are unavailable. Unavailable agents are shown with status/details in the UI.

Run

Recommended:

titans.exe

Alternative:

titans.bat

Other option:

python main.py

Important: run from the project folder, or use titans.bat (it handles path setup automatically).

Desktop Features

  • Native Tkinter desktop interface
  • Real-time runtime telemetry and visual panel
  • Numeric-series extraction from demonstration output
  • Play/scrub chart interaction for time-based behavior
  • Output console + agent details pane
  • Adjustable split-pane layout with remembered last position

Troubleshooting

  • If the app closes quickly, run with titans.bat and read the terminal output.
  • If python main.py fails from D:\GitHub\Titans, switch to the project directory first.
  • google.generativeai deprecation warnings are non-fatal warnings.

Project Files

  • main.py: Desktop application
  • titans.exe: Packaged Windows executable for desktop launch
  • titans.bat: Robust launcher for Windows
  • main.pygui: GUI Pie metadata file (not the runtime entrypoint)
  • agents/: Provider-specific agent implementations

Reviews (0)

No results found