註冊並分享邀請連結,可獲得影片播放與邀請獎勵。

Michael Livs
@micLivs
Deep in the agentic stack.
加入 April 2016
146 正在關注    4.1K 粉絲
@PrimeIntellect did PI-pi, so here i am doing pi-PI. introducing pi-rlm prime intellect did a thing i wanted to do for a long time, they fused a stateful jupyter kernel with an agent. pi-rlm is that idea on top of bun and typescript. it takes over all of your available tools and swallows them into a single tool called `execute`, which runs code in a persistent evaluator. a single-tool pi that does everything. shell is Bun.$, pi's own tools are mounted inside as tools.*, subagents are a function call that returns a handle. @badlogicgames's pi stays untouched until you ask for it: pi install npm:@shift-labs/pi-rlm pi --rlm (btw variables survive across turns, sessions, and restarts, have fun)
顯示更多
0
25
358
23
轉發到社區