ask

ask is a Go CLI to query LLMs from your terminal with SDK-free HTTP clients.

Build

go build -o ask .

Quick Start

1. Set a provider key. If missing, config.json and config.template.json are created automatically on first config-using command:

ask key set openai --value sk-...

2. Set default provider and model:

ask provider set openai
ask models select --provider openai --search mini

3. Ask:

ask "command to remove a commit from git"

Or start without a question and type it interactively:

ask
# ask> command to remove a commit from git

You can also pipe the question via stdin:

echo "command to remove a commit from git" | ask

Command Prefill UX

If the model returns command, ask opens an editable terminal prompt with the command prefilled.

• Enter: run command
• Edit text first: run edited command
• Ctrl+C: copy suggested command to clipboard and exit prompt
• Ctrl+D: exit prompt without running

Core Commands

ask "question" [options]
ask models list|select|set|current
ask provider list|current|set|show|add|remove
ask key set|show|clear
ask config show|path|template
ask markdown on|off|status
ask help ask|models|provider|key|config|markdown

Ask Options

• -p, --provider <name>
• -m, --model <id>
• --timeout <dur|sec> (default: 90s)
• --no-markdown
• --no-run
• --json

If your question starts with -, use:

ask -- "-question that starts with dash"

Providers

Built-in providers:

• openai
• anthropic
• gemini
• ollama
• openrouter

Add a custom OpenAI-compatible provider:

ask provider add myproxy \
  --base-url https://llm.example.com/v1 \
  --api-key-env MYPROXY_API_KEY

Config

Default config path:

• macOS/Linux: ~/.ask/config.json
• Windows: %USERPROFILE%\.ask\config.json

Overrides:

• ASK_CONFIG=/path/to/config.json
• ASK_CONFIG_DIR=/path/to/config/dir
• ask --config /path/to/config.json ...

Security defaults:

• config directory mode: 0700
• config file mode: 0600

API key resolution order:

1. Environment variable from api_key_env (or built-in default env var)
2. api_key in config.json

Show active paths:

ask config path
ask config template

Show raw config content:

ask config show

Config Template (Example)

{
  "version": 1,
  "current_provider": "",
  "providers": {
    "openai": {
      "api_key": "",
      "model": "gpt-5-nano",
      "api_key_env": "OPENAI_API_KEY"
    },
    "anthropic": {
      "api_key": "",
      "model": "",
      "api_key_env": "ANTHROPIC_API_KEY"
    },
    "gemini": {
      "api_key": "",
      "model": "",
      "api_key_env": "GEMINI_API_KEY"
    },
    "ollama": {
      "api_key": "",
      "model": "",
      "base_url": "http://127.0.0.1:11434"
    },
    "openrouter": {
      "api_key": "",
      "model": "",
      "api_key_env": "OPENROUTER_API_KEY"
    }
  },
  "custom_providers": {
    "myproxy": {
      "base_url": "https://llm.example.com/v1",
      "api_key": "",
      "model": "",
      "api_key_env": "MYPROXY_API_KEY",
      "headers": {
        "X-Client-Name": "ask"
      }
    }
  },
  "render_markdown": true
}

Notes

• Model lists are fetched from provider APIs
• Responses are requested in structured JSON (answer, command) with fallback parsing.
• Markdown rendering uses charmbracelet/glamour.

Development

go test ./...
go vet ./...
go build ./...