Linux command
mods 命令
网络
复制后可按需替换文件名、目录或参数。
常用示例
Ask a one-off question
mods "[explain what SIGPIPE does]"
Pipe file contents into a prompt
cat [file.py] | mods "[review this code for bugs]"
Select a specific model
mods -m [gpt-4o] "[summarise this]"
Continue the last conversation
mods -C "[and how would I test that?]"
Resume a named conversation
mods -c [refactor] "[next step]"
Format the response as Markdown
mods -f "[write release notes from these commits]"
List saved conversations
mods -l
Apply a custom role / system prompt
mods --role [shell] "[find large files in /var]"
说明
mods is a command-line AI companion that turns STDIN and arguments into a prompt, sends it to a configured LLM, and streams the response back to the terminal. It supports OpenAI, Anthropic, Cohere, Groq, Google Gemini, Azure, and local providers such as Ollama and LocalAI, selected per-request via --model or in the YAML configuration. Conversations are cached to disk so they can be resumed by name or continued from the last turn, and Markdown rendering via Glamour lets replies display with syntax highlighting and headings when --format is set.
参数
- -m, --model _NAME_
- Use the named model (for example `gpt-4o`, `claude-3-5-sonnet`, a local Ollama model).
- -M, --ask-model
- Prompt interactively to choose a model.
- -f, --format
- Ask the model for formatted (Markdown) output and render it in the terminal.
- --format-as _FORMAT_
- Specify the output format (for example `markdown`, `json`).
- -r, --raw
- Print the raw, unformatted response.
- -q, --quiet
- Suppress non-error output.
- -P, --prompt
- Include the prompt from the command arguments and STDIN in the response.
- -p, --prompt-args
- Include the CLI prompt arguments in the response.
- --max-tokens _N_
- Limit the response length.
- --no-limit
- Do not restrict the response length.
- --word-wrap _WIDTH_
- Wrap output at the given column (default: 80).
- -t, --title _NAME_
- Title the current conversation for later retrieval.
- -l, --list
- List saved conversations.
- -c, --continue _NAME_
- Resume a saved conversation by name.
- -C, --continue-last
- Resume the most recent conversation.
- -s, --show _NAME_
- Print a saved conversation.
- --no-cache
- Do not persist the conversation to the cache.
- --role _NAME_
- Apply a custom role / system prompt defined in the config.
- --temp _FLOAT_
- Sampling temperature.
- --topp _FLOAT_, --topk _INT_
- Top-p / top-k sampling parameters.
- --theme _NAME_
- UI theme (`charm`, `catppuccin`, `dracula`, `base16`).
- -x, --http-proxy _URL_
- Route requests through an HTTP proxy.
- --settings
- Open the settings file in `$EDITOR`.
FAQ
What is the mods command used for?
mods is a command-line AI companion that turns STDIN and arguments into a prompt, sends it to a configured LLM, and streams the response back to the terminal. It supports OpenAI, Anthropic, Cohere, Groq, Google Gemini, Azure, and local providers such as Ollama and LocalAI, selected per-request via --model or in the YAML configuration. Conversations are cached to disk so they can be resumed by name or continued from the last turn, and Markdown rendering via Glamour lets replies display with syntax highlighting and headings when --format is set.
How do I run a basic mods example?
Run `mods "[explain what SIGPIPE does]"` in a terminal, then adjust file names, paths, flags, or remote targets for your system.
What does -m, --model _NAME_ do in mods?
Use the named model (for example `gpt-4o`, `claude-3-5-sonnet`, a local Ollama model).