temm1e-labs/temm1e
One of the ground zero for AGI, maybe, probably. Free and Open Source forever and ever!
What it solves
Temm1e is designed to move beyond one-off AI interactions by creating a persistent AI companion and agent runtime. It solves the problem of "episodic intelligence"—where AI forgets everything between sessions—by providing an environment where an agent can maintain memory, personality, and the ability to return to scheduled work over time.
How it works
Written in Rust, Temm1e acts as a harness for various LLMs (Anthropic, OpenAI, Gemini, etc.). It manages the agent loop through a five-step process: understanding requests, executing tools, managing context, recording outcomes, and releasing resources.
Key technical components include:
- Memory Systems: Uses $\lambda$-Memory for continuous attention curves (decaying importance over time) and Engram for long-term facts.
- Computer Use: Features "Prowl" for authenticated browser interaction and "Gaze" for desktop screenshots and input actions.
- Coding Layer: "Tem-Code" allows the agent to inspect files, edit code, and run shell/Git operations directly in a repository.
- Coordination: Supports a swarm model ("Many Tems") for splitting work into dependency-aware tasks and "TemDOS" for maintaining specialist cores (e.g., research or debugging).
- Persistence: The "Perpetuum" system enables scheduled activity and background monitoring.
Who it’s for
Developers and power users who want a persistent, autonomous agent that can operate across their terminal, browser, and desktop while retaining a level of personality and long-term memory.
Highlights
- Persistent Identity: Maintains a consistent personality and memory across different conversations and sessions.
- Computer Use: Native browser sessions and desktop control for interacting with apps without APIs.
- Mathematical Rigor: Uses specific equations for memory decay, task selection, and evidence ranking to make agent behavior inspectable.
- Multi-Agent Coordination: Ability to spawn worker swarms to handle complex, parallelizable tasks.
- Cross-Platform Interface: Accessible via a TUI, CLI, or messaging gateways.
관련
- 프로젝트
- 프로젝트
- 프로젝트
- 프로젝트
- 프로젝트