Summary

OpenLumara is a modular, token-efficient AI agent framework built from scratch in Python, prioritizing a local-first, lightweight, and secure design. It targets users seeking optimized performance for local AI backends like llamacpp and koboldcpp, significantly reducing token usage and offering a highly customizable experience. The framework is particularly beneficial for life management tasks and individuals managing executive dysfunction.

How It Works

The core framework is hand-coded, with AI assisting in identifying improvements and debugging, ensuring all integrated code is manually audited. Non-core components, such as channels, were largely AI-generated but also underwent manual review. This AI-assisted approach allows rapid development while maintaining control. OpenLumara's architecture is fundamentally modular, enabling users to enable or disable specific functionalities—from memory and scheduling to shell access—to tailor the agent precisely to their needs and minimize resource footprint. Its design emphasizes extreme token efficiency, making it cost-effective and performant.

Quick Start & Requirements

Installation involves cloning the repository (git clone https://github.com/Rose22/openlumara) and running the provided scripts (run.sh for Linux, run.bat for Windows). Post-startup, users configure API connections via the Web UI. Updates are managed via update.sh or update.bat for git-cloned instances. It requires an OpenAI API-compatible backend (local or cloud) and optionally Docker/Podman for its sandboxed shell module.

Highlighted Details

• Supports any OpenAI API-compatible backend, including local options like llamacpp, ollama, and koboldcpp.
• Offers a fully private, self-hosted deployment option.
• Highly modular design allows granular control over features like shell access, memory, scheduler, and awareness modules.
• Features a scheduler system for automated tasks and a memory system using efficient MessagePack storage.
• Prioritizes token efficiency, with tools (/status, /context) to monitor usage.
• Includes optional modules for character personas (as a Character.AI alternative) and a secure, sandboxed shell environment.
• Supports multiple communication channels: WebUI, CLI, Telegram, Discord, and Matrix (with encryption).

Maintenance & Community

No specific details regarding maintainers, community channels (e.g., Discord, Slack), or project roadmap were provided in the README.

Licensing & Compatibility

The README does not specify a software license. This absence creates ambiguity regarding usage rights, modification, and distribution, particularly for commercial applications.

Limitations & Caveats

The project acknowledges AI assistance in its development, with non-core components being largely AI-generated and manually audited. A plugin downloading system is planned for future implementation. The lack of explicit licensing information is a significant caveat for adoption.