rajudandigam

agent-inspect

rajudandigam

Local execution trees for TypeScript AI agents. agent-inspect helps you understand what happened inside an AI agent run — locally. It turns manual steps, tool calls, LLM calls, structured logs, failures, durations, and run metadata into readable execution trees you can inspect from the terminal. It is built for TypeScript/Node.js developers..

AI 简介

agent-inspect 是一个面向 TypeScript/Node.js 开发者的本地 AI 代理可观测性工具,用于在不上传云端的前提下,将 AI 代理运行过程中的步骤、工具调用、LLM 调用、失败信息、耗时及元数据结构化为可读的执行树,并通过终端 CLI 查看与对比。其核心特点是纯本地运行、零依赖云服务、支持手动埋点与结构化日志集成、兼容 OpenTelemetry 标准导出,且无需账号或仪表盘。适用于本地开发调试、快速迭代验证、离线环境下的 AI 代理行为分析,以及作为生产级可观测平台(如 LangSmith/Langfuse)的前置轻量层。

TypeScript
MIT License
206
Stars
85
Forks
70
Watchers
15
Issues

Star 增长

今日0
近 7 天0
近 30 天0
综合评分45.8
默认分支main