展示HN:LogWhisperer – 本地AI驱动的系统日志摘要工具(无需API密钥)
我正在开发一个名为 LogWhisperer 的开源工具。它是一个基于命令行的日志摘要工具,使用本地的 LLM(通过 Ollama)来解析和总结 Linux 系统日志,如 journalctl 或 /var/log/syslog,完全不依赖互联网或 OpenAI API。
它的用途包括:
- 快速处理大量日志
- 识别 systemd 故障或启动问题
- 从原始日志行获取人类可读的摘要
在底层,它通过 Ollama 使用 mistral 或 phi,并支持:
- 文件或 journalctl 源
- 便于定制的命令行选项(--entries、--model 等)
- Markdown 格式的输出报告
- 进度指示器/加载动画,以避免命令行“卡住”的混淆
* 完全离线工作
* 轻量级,可以通过一行命令安装
* 无需云调用或令牌依赖
GitHub: [https://github.com/binary-knight/logwhisperer](https://github.com/binary-knight/logwhisperer)
欢迎反馈、问题或功能建议!
查看原文
I've been working on an open-source tool called LogWhisperer. It's a CLI-based log summarizer that uses a local LLM (via Ollama) to interpret and summarize Linux system logs like journalctl or /var/log/syslog — with zero internet or OpenAI API dependency.<p>It's useful for:<p>Quickly triaging large volumes of logs<p>Identifying systemd failures or boot issues<p>Getting human-readable summaries from raw log lines<p>Under the hood, it uses mistral or phi via Ollama, and supports:<p>File or journalctl sources<p>CLI flags for easy customization (--entries, --model, etc.)<p>Markdown output reports<p>Spinner/progress UX to avoid CLI "hang" confusion<p>* Works entirely offline
* Lightweight, installable with a one-liner
* No cloud calls or token dependencies<p>GitHub: <a href="https://github.com/binary-knight/logwhisperer">https://github.com/binary-knight/logwhisperer</a>
Feedback, issues, or feature ideas welcome!