lask 0.3.0

Ask your llm from the command line

lask

Ask LLMs right from the terminal.

Features

• CLI for multiple LLM providers (OpenAI, Anthropic, AWS Bedrock, Azure)
• Customizable models and parameters
• Minimal dependencies (requests, boto3 for AWS)
• Streaming responses
• Repl mode for interactive use, with temporary chat history
• Pipe input support
• Smart command mode: describe what you want in plain English and lask translates it into a shell command, with confirmation before execution

Installation

pip install lask

Usage

With OPENAI_API_KEY in your environment or in ~/.lask-config:

lask What movie is this quote from\? \"that still only counts as one\"

Or as a repl:

lask

==== Lask REPL Mode ====
Using provider: openai
Smart command: auto

> What movie is this quote from? "that still only counts as one"
LLM response here...
> When was that movie released?

Smart command mode also works in the REPL. When smart_command is auto or true, the REPL automatically detects command requests and handles them the same way — propose, confirm, execute:

> show disk usage for the current directory

  du -sh .

Press Enter to run, or any other key to abort.

> explain what du does
The `du` command estimates file space usage...

REPL commands:

Command	Description
!help	Show available REPL commands
!clear	Clear the screen
!history	Show prompt history
!vi	Switch to Vi editing mode
!emacs	Switch to Emacs editing mode
exit / quit	Exit the REPL
Ctrl+C	Interrupt a running command or response
Ctrl+D	Exit the REPL

Or via pipe:

echo "What movie is this quote from? \"that still only counts as one\"" | lask

Smart Command Mode

By default, lask automatically detects if your prompt is asking for a shell command. When it is, it translates your natural language into a command, shows it to you, and waits for confirmation:

$ lask show diff between current branch and main

  git diff HEAD..main

Press Enter to run, or any other key to abort.

Press Enter to execute, or any other key to abort. The executed command is added to your shell history, so you can press ↑ to recall and re-run or edit it.

The behaviour is controlled by the smart_command setting (default: auto):

Value	Behaviour
auto	LLM decides if the prompt is a command request or a general question
true	Always treat the prompt as a command request
false	Never use smart command mode, always answer normally

In REPL mode, smart command keeps a rolling history of your interactions so you can reference them naturally ("do the same but with -v", "now compress that", etc.). Two settings control what is included:

Setting	Default	Description
smart_context_commands	true	Include previous prompts and commands in context
smart_context_output	false	Also capture and include command output

Setup

1. Get API keys from your provider:

   • OpenAI
   • Anthropic
   • AWS Bedrock (uses AWS credentials)

2. Create ~/.lask-config:

   [default]
   provider = openai  # openai, anthropic, aws, azure
   system_prompt = Always answer questions concisely.
   
   [openai]
   api_key = your-api-key-here
   model = gpt-4.1
   
   [anthropic]
   api_key = your-api-key-here
   model = claude-3-opus-20240229
   
   [aws]
   model_id = anthropic.claude-3-sonnet-20240229-v1:0
   region = us-east-1
   
   [azure]
   api_key = your-azure-api-key
   resource_name = your-resource-name
   deployment_id = your-deployment-id
   

   See examples/example.lask-config for a full example.

3. Or use environment variables:

   export OPENAI_API_KEY='your-api-key-here'
   export ANTHROPIC_API_KEY='your-api-key-here'
   # AWS uses standard AWS credentials
   export AZURE_OPENAI_API_KEY='your-api-key-here'
   

Configuration Options

Provider Selection

[default]
provider = openai  # openai, anthropic, aws, azure

Streaming

[openai]
streaming = false  # Disable streaming (true by default)

System Prompts

[default]
system_prompt = Always answer questions concisely.

[openai]
system_prompt = You are a helpful AI assistant.  # Provider-specific

Smart Command

[default]
smart_command = auto              # auto, true, or false
smart_context_commands = true      # include command history in context
smart_context_output = false       # include command output in context

Provider-Specific Settings

Each provider supports model, temperature, max_tokens, and other parameters.

See examples/example.lask-config for all options.

Development

This repo uses uv:

# Install dependencies
uv sync

# Build
uv build

# Install for development
pip install -e .

# For AWS Bedrock
pip install boto3

License

GNU General Public License v3.0 (GPL-3.0)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.