Dont know how to do something in your terminal?

howto print my system information

And you'll be told!

╭────────────────────────────────────────────────────────────────────────────────╮
│ You can use the `uname -o` command to display your operating system.           │
│                                                                                │
│                                                                                │
│  uname -o                                                                      │
│                                                                                │
│                                                                                │
│ This command works with various Unix-like systems, including Linux and macOS.  │
│ If you're running a 64-bit kernel on an Apple M1 or ARM-based hardware, it     │
│ will return "Darwin" (the underlying OS name for macOS).                       │
╰────────────────────────────────────────────────────────────────────────────────╯

Installing

howto is available on pypi and can be installed in any manner of ways

pip install howto-ai

pipx install howto-ai

uvx --from howto-ai howto

Note: pre 1.0.0 release are assumed to be unstable and might break or do unexpected things, use wisely

Usage

λ: howto --help

 Usage: howto [OPTIONS] [QUERY]...

╭─ Arguments ────────────────────────────────────────────────────────────────────╮
│   query      [QUERY]...  This is what you'd like to ask as a question. Empty   │
│                          queries will open a prompt.                           │
│                          [default: None]                                       │
╰────────────────────────────────────────────────────────────────────────────────╯
╭─ Options ──────────────────────────────────────────────────────────────────────╮
│ --debug                 --no-debug            Make no requests to llms         │
│                                               [default: no-debug]              │
│ --dry-run               --no-dry-run          Make no requests to llms         │
│                                               [default: no-dry-run]            │
│ --config-path           --no-config-path      Print the default config path    │
│                                               and exit.                        │
│                                               [default: no-config-path]        │
│ --config-show           --no-config-show      Print the config and exit.       │
│                                               [default: no-config-show]        │
│ --install-completion                          Install completion for the       │
│                                               current shell.                   │
│ --show-completion                             Show completion for the current  │
│                                               shell, to copy it or customize   │
│                                               the installation.                │
│ --help                                        Show this message and exit.      │
╰────────────────────────────────────────────────────────────────────────────────╯

Backends

howto uses 3rd party services as backends. By default, It's configured to use ollama which is expected to be running locally (make sure you read their docs). If it is, your howto commands will work.

howto can also be used against popular local, private and public AI APIs including ChatGPT, Gemini and others. To configure this, you need to edit the config file or use the commands below. Locating the config file can be done with:

howto --config-path

Edit this in your favourite text editor. If this is difficult, please open an Issue and we can help.

General Directions

Set the model you want by using the --set-model flag or updating the config file at .config/howto-ai/config.toml. We use liteLLM for the proxy service, so any supported model will work.

howto --set-model '<provider>/<model>'

or

model = "<provider>/<model>"
# for example
model = "huggingface/facebook/blenderbot-400M-distill" # https://huggingface.co/facebook/blenderbot-400M-distill
# ...

Here are some examples of popular providers and their models:

Ollama

howto --set-model='ollama/deepseek-r1:32b'
howto --set-model='ollama/deepseek-r1:32b'
howto --set-model='ollama/deepseek-r1:32b'
howto --set-model='ollama/deepseek-r1:32b'

Open AI

howto --set-model 'gpt-4o-mini' # https://platform.openai.com/docs/models
howto --set-model 'gpt-4o'
...

then export your OPENAI_API_KEY for that session (in your .envrc or .bashrc or wherever)

export OPENAI_API_KEY="<your key here>" # https://platform.openai.com/api-keys

Hugging Face

howto --set-model 'huggingface/Qwen/Qwen2.5-Coder-32B-Instruct' # https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct
howto --set-model 'huggingface/facebook/blenderbot-400M-distill'  # https://huggingface.co/facebook/blenderbot-400M-distill
...

then export your HUGGINGFACE_API_KEY for that session (in your .envrc or .bashrc or wherever)

export HUGGINGFACE_API_KEY="<your key here>" # https://huggingface.co/docs/hub/security-tokens

and you should be good to go

Anthropic

howto --set-model 'claude-3-opus-latest' # https://docs.anthropic.com/en/docs/about-claude/models
howto --set-model 'claude-3-5-sonnet-latest'
howto --set-model 'claude-3-5-haiku-latest'
...

then export your ANTHROPIC_API_KEY for that session (in your .envrc or .bashrc or wherever)

export ANTHROPIC_API_KEY="<your key here>" 

and you should be good to go

Remember AI can produce rubbish code, remember to research the commands howto spits out