How Gemini CLI, AI Input Method, and Obsidian Create a Seamless Workflow

The article walks through a hands‑on workflow that integrates Gemini CLI, the Zhipu AI input method (referred to as "Xiao Ao"), and Obsidian for knowledge management and content creation. After a brief introduction, the author lists the workflow nodes that will be covered.

Tool Selection

Obsidian and Gemini CLI are presented as essential tools for knowledge management. The author compares two categories of AI input methods: one based on real‑time translation with low latency and privacy, and another offering richer agent capabilities for complex scenarios. After testing, the Zhipu AI input method Xiao Ao is chosen for its superior recognition, scene handling, and customizable style output.

Workflow Nodes

1、用嘴在Obsidian中启动Gemini Cli
2、查看当前文件夹磁盘占用，并显示所有Python相关的进程
3、看下当前文件夹有多少张图片，并做图片格式转换
4、视频格式转GIF，且控制大小
5、Obsidian写作：从爆款标题到文案创作到局部优化改写一整套工作流
7、编程做插件开发场景
8、评论区回复
9、懒人计算器

Voice‑Activated Gemini CLI

Instead of typing gemini in the Obsidian terminal, the user holds the fn key and says “Help me start Gemini”. The AI input method interprets the spoken command and executes the preset Gemini CLI instruction, eliminating typing errors and speeding up startup.

Terminal Monitoring

Normally the user would manually check disk usage and running Python processes. Now a single spoken request—“Show current folder disk usage and list all Python‑related processes”—lets the AI input method generate the appropriate command, which Gemini CLI runs instantly.

File Management and Media Conversion

Managing numerous files in Obsidian can be tedious. By saying “How many images are in the current folder and convert their formats”, the AI input method produces the necessary commands. For video‑to‑GIF conversion with size limits, the user says “Convert demo.mp4 to a GIF of the first 3 seconds, under 10 MB”, and the AI returns a complete ffmpeg command that Gemini CLI executes.

Obsidian Writing

The workflow supports note‑taking and content creation. For public‑account articles, the AI can generate titles, draft copy, perform local optimizations, and format output (e.g., converting text to charts). The author notes current limitations, such as lack of image insertion and tool‑calling capabilities.

Cross‑Platform Content Generation

Using a custom plugin, the author can transform text into a Xiaohongshu (Little Red Book) post by speaking “Convert to Xiaohongshu copy”. The conversion is limited by input length constraints.

Plugin Development

Previously, Gemini CLI assisted with Obsidian plugin development. With the AI input method, commands that previously required manual typing can now be issued by voice, making the development loop smoother.

Comment Reply Automation

To address delayed fan replies, the author uses the AI input method to draft responses, reducing the time spent on manual typing.

Calculator Shortcut

Simple calculations that would normally require opening a calculator can be performed by voice, with the AI input method returning the result instantly.

Underlying Model

The Zhipu AI input method is built on the open‑source GLM‑ASR‑Nano‑2512 model (1.5 B parameters), which achieves state‑of‑the‑art performance in speech recognition, enabling the described workflow.

Conclusion

The author reflects that the most efficient workflow strips away mouse clicks and window switching, relying only on a spoken command and an immediate result, emphasizing that the spark of creativity remains the most valuable asset even in an AI‑driven era.