Natural language to a shell script/command in a flash right in your terminal

4 months ago 7

English | 中文

you is a command-line tool that translates natural language instructions into executable shell commands, making command-line operations more accessible and intuitive. It's designed especially for newcomers to command-line interfaces, but also helps experienced users by reducing cognitive load and documentation searches.

  • Talk in Plain Language: Just tell the tool what you want to do in regular words, and it turns that into commands the computer understands
  • Chat Mode: Have a back-and-forth conversation where you can ask for multiple things while the tool remembers what you talked about before
  • Command Review: The tool handles the hard technical stuff but lets you decide when to run commands
  • Save for Later: Keep useful commands to use again without having to ask the tool each time

Setup Script (Linux & macOS)

Download and run the setup script:

curl -O https://raw.githubusercontent.com/AspadaX/you/main/setup.sh && chmod +x ./setup.sh && ./setup.sh && rm ./setup.sh

To update:

curl -O https://raw.githubusercontent.com/AspadaX/you/main/update.sh && chmod +x ./update.sh && ./update.sh && rm ./update.sh

To uninstall:

curl -O https://raw.githubusercontent.com/AspadaX/you/main/uninstall.sh && chmod +x ./uninstall.sh && ./uninstall.sh && rm ./uninstall.sh

If you have Rust installed:

Run a command described in natural language:

you run "find the largest file in my downloads directory"

Get an explanation of what a command does:

you explain "find . -type f -name '*.txt' -size +10M"

Start a conversational session to run multiple related commands:

List all cached scripts:

you list # or use the alias you ls

Remove a specific cached script:

you remove <script_name> # or use the alias you rm <script_name>

Configure your preferred CLI

You may want to use fd over find, or prefer using a different CLI rather than letting the LLM guess. In this case, you may update the configuration file located at ~/.you/configurations.json. Below is an example:

{ "preferred_clis": [ { "name": "fd", "preferred_for": "search files. and replace find" } ] }

Now, you will use fd over find when you issue commands relevant to searching files.

If you would like to enable cache feature, you may also want to enable it in the ~/.you/configurations.json:

{ "enable_cache": true, "preferred_clis": [ { "name": "fd", "preferred_for": "search files. and replace find" } ] }
# Find files you've modified in the last week you run "show me files I've modified in the last 7 days" # Get system information you run "how many CPU cores and how much RAM does this system have?" # Complex tasks made simple you run "compress all JPG images in the current directory and save them to a new folder" # Remote operations you run "connect to my server at 192.168.*.* and check disk space"

you works with various LLMs:

  • Works well with small models like smollm2
  • Compatible with OpenAI compatible APIs, such as DeepSeek.
  • Compatible with ollama for using any open-source model for free
  • Configure your preferred model for the best balance of performance and accuracy
  1. Type your request in natural language
  2. Review the suggested command(s) and explanation
  3. Type 'y' to execute or provide additional guidance
  4. If there's an error, the AI automatically suggests a corrected command
  5. Optionally save useful command sequences for future reuse
  • Reduce Documentation Searches: Get the right command without extensive searching
  • Learning Tool: See how natural language translates to actual commands
  • Productivity Boost: Accomplish complex tasks with simple instructions
  • Safe Command Execution: Review commands before execution
  • Error Recovery: Get help when commands fail
  • Context Retention: In interactive mode, the AI remembers previous commands

MIT

Created by Xinyu Bao

This project would not be possible without these amazing libraries:

  • anyhow: Error handling made simple and flexible
  • async-openai: API client for interacting with OpenAI's language models
  • cchain: Command chaining functionality for shell operations
  • chrono: Date and time handling with precision
  • clap: Command-line argument parsing with a beautiful interface
  • console: Terminal text styling and utilities
  • indicatif: Progress indicators for command-line applications
  • serde/serde_json: Powerful serialization and deserialization framework
  • sysinfo: System information gathering across platforms
  • tokio: Asynchronous runtime for efficient operations
  • surfing: Parse JSON out of plain texts

A big thank you to all the developers who maintain these open-source libraries!


you - because command lines should understand you, not the other way around.

Read Entire Article