azc 1.2.0

Command-line tool for interacting with LLMs

azc - LLM Chat from the command-line

A command-line tool for interacting with LLMs.

Why should you use this?

• If you're a command-line junkie, you don't need to switch to another tool to chat with your LLM
• One tool, multiple LLMs (OpenAI, Anthropic, Ollama, Gemini...) - Why settle for one when you can have them all?
• Pay-as-you-go pricing for LLM providers (Cheaper in many cases)

Features

• Multi-provider support
  • Ollama
  • OpenAI
  • Anthropic
  • Gemini
  • Grok
• Streaming responses (see response as it is being generated)
• Persistent command-line history (use up and down arrows to navigate)
• Chat history & reset (full discussion, start new chat)
• Switch provider and model (compare models and providers)
• Markdown output (nicely formatted headings, lists, tables, etc.)
• Command-line parameters (first prompt)

Possible future features

• Support more LLM providers (Perplexity, Groq...)
• Support RTL languages
• Save output to file
• Patterns library (a-la fabric)
• Upload files
• Aggregate responses from multiple providers
• Support more modes (image generation, transcription, etc.)
• Track total cost of API calls
• Add to homebrew

Installation

pip install azc

if you want to make it available system-wide (any folder, without having to activate your python environment), you can install it system-wide:

echo 'alias azc="$HOME/projects/azc/env/bin/azc"' >> ~/.bashrc

• Replace ~/.bashrc with the shell you are using (zsh, bash, etc.)
• Replace $HOME/projects/azc/env/bin/azc with the path to the azc executable
• On Windows, you can add the path to your PATH environment variable (ask your nearest LLM for more details)

Running

% azc
azc> how tall is the eiffel tower?
                                  ollama:llama3.1:latest (1st message)
  The Eiffel Tower stands at an impressive height of:

  • 324 meters (1,063 feet) tall, including its antenna.
  • 302.9 meters (994.7 feet) tall, excluding its antenna.

  It was the world's tallest man-made structure when it was first built for the 1889 World's Fair in
  Paris.

azc> q
👋 Bye
%

You can specify the first prompt as a command-line argument:

% azc -b "What is the capital of Panama?"
The capital of Panama is Panama City.
%

Command-line parameters

Parameter	Description
(default)	First prompt
-d / --double-enter	Press enter twice to submit - This is useful for those who want to use multi-line prompts without pressing ctrl-j to add new line.
-b / --batch	Exit after the first response
-v / --verbose	Print verbose output
-p / --provider	The provider to use (e.g. openai or ollama). Abbreviations allowed, like op for openai
-m /--model	The model to use. Abbreviations allowed, like tu for gpt-3.5-turbo

Commands

Command	Description
q or exit	Exit the program
h or ?	Show help
l	List models
n	Start new chat
p	Change provider. p - (p followed by a space) trigger auto-completion menu
m	Change model
ctrl-n	New line

Setup

You will need to configure at least one LLM API.

You should create a .env file which contains your API Key/s. The .env file should be located in home directory, either directly under $HOME or under $HOME/.config/, the latter taking precedence. See .env.sample for a sample .env file, with the expected environment variables' names. Remove those that are not relevant to you.

Here are the links to the API sign-up pages (or download in case of Ollama):

• OpenAI
• Anthropic
• Ollama
• Gemini
• Grok

You can configure the default models you want to use in azc_config.json. This file is expected to be found under ~/.config/azc_config.json or ~/.azc_config.json if you don't have a ~/.config folder.

Limitations

• Streaming updates are limited to screen height (after that it displays ellipsis and will update the display only when the response is complete)
• No support for RTL languages

Contributing

Contributions are welcome! Please feel free to submit a PR.

To run in development:

% python -m az.az

License

MIT