The ell command-line interface for LLMs written in Bash allows users to interact with LLMs from their terminal, making it pipe-friendly and easy to ask questions. It supports function calling and more via templates. Users need bash, jq, curl, and perl to run ell. Configuration involves setting variables for different LLM models like gemini-1.5-flash and gpt-4o-mini. Usage examples include asking questions, specifying models, and recording terminal input and output for context. The name “ell” comes from a combination of shell and LLM, and the project is written in Bash for simplicity. It is lightweight, easy to install, and can be extended and modified easily. Contributions are welcome, and the project is licensed under the MIT License. Controversial information includes the use of perl and the exclusion of other programming languages for development.
https://github.com/simonmysun/ell