Interesting. I have been looking at classifying model tool calls for a session replay pipeline. My goal is to know which reach external systems and would need to be mocked out. I currently have a small deterministic pipeline + fallback to a small model call.
Interesting. I have been looking at classifying model tool calls for a session replay pipeline. My goal is to know which reach external systems and would need to be mocked out. I currently have a small deterministic pipeline + fallback to a small model call.
Ultra fast shell command classification for coding agents, basically.