4 ms·

I had never heard of this "x" utility before, but I really dislike its philosophy. A lot of its modules seem downright harmful, like, `x install`, which uses a

by rafram 2y ago

I had never heard of this "x" utility before, but I really dislike its philosophy. A lot of its modules seem downright harmful, like, `x install`, which uses a crowdsourced database of `curl "https://<url> https://<url>" | sh -c`-style commands [1]. New submissions are reviewed via GitHub, but it's still a terrible idea. There's a reason people use package managers with central build script repositories and checksum verification. Oh, and it does AI, or something like that [2]. This just feels like a BusyBox that doesn't know what it wants to be. [1] https://www.x-cmd.com/install/ https://www.x-cmd.com/install/ [2] https://www.x-cmd.com/start/ai-llm https://www.x-cmd.com/start/ai-llm

lijunhao 2y ago
X-CMD design philosophy is all about freedom to choose. Take genact as example (see: https://www.x-cmd.com/install/genact https://www.x-cmd.com/install/genact). 1. X-cmd aggregates all known installation methods, sourced from the community and official documentation ( including apt, dnf, ... curl), and presents them in an interactive list. 2. It is the decision for the user to CHOOSE the installation method that best suits the situation. Whether it's apt, dnf, or any other method, x-cmd always puts user in control. As a former operation engineer, I often encountered situations where installing a simple tool, like jq, on a client's server required unnecessary sudo privileges. Even worse, some software installations demanded extensive dependency upgrades, potentially destabilizing the entire system. To address this, I developed x-cmd/pkg, a tool management system that prioritizes portable tools. This approach eliminates the need for the unecessary privileged access and minimizes the impact on the original system. X-cmd also offers other methods for installing portable software from the community: like asdf, cosmo. It is all user's decision. About the LLM, there won't be a code to eval without user's confirmation. X-CMD won't be a powerful framework like Langchain; it's just a simple command-line interface (CLI) that uses CURL. Large language models (LLMs) may be used to generate commands, but no code will be executed without explicit user confirmation. User safety and control are top priorities. We prioritize transparency by open-sourcing our code and providing best-effort documentation and demo. When operating in the field, you're not executing opaque binary code, but rather human-readable POSIX shell scripts. This allows you to fully understand the actions being performed when necessary.

by rafram 2y ago

lijunhao 2y ago

X-CMD design philosophy is all about freedom to choose. Take genact as example (see: https://www.x-cmd.com/install/genact https://www.x-cmd.com/install/genact). 1. X-cmd aggregates all known installation methods, sourced from the community and official documentation ( including apt, dnf, ... curl), and presents them in an interactive list. 2. It is the decision for the user to CHOOSE the installation method that best suits the situation. Whether it's apt, dnf, or any other method, x-cmd always puts user in control. As a former operation engineer, I often encountered situations where installing a simple tool, like jq, on a client's server required unnecessary sudo privileges. Even worse, some software installations demanded extensive dependency upgrades, potentially destabilizing the entire system. To address this, I developed x-cmd/pkg, a tool management system that prioritizes portable tools. This approach eliminates the need for the unecessary privileged access and minimizes the impact on the original system. X-cmd also offers other methods for installing portable software from the community: like asdf, cosmo. It is all user's decision. About the LLM, there won't be a code to eval without user's confirmation. X-CMD won't be a powerful framework like Langchain; it's just a simple command-line interface (CLI) that uses CURL. Large language models (LLMs) may be used to generate commands, but no code will be executed without explicit user confirmation. User safety and control are top priorities. We prioritize transparency by open-sourcing our code and providing best-effort documentation and demo. When operating in the field, you're not executing opaque binary code, but rather human-readable POSIX shell scripts. This allows you to fully understand the actions being performed when necessary.