AI Command Line Tools

Welcome to the susi_shell repository, which is part of the susi_api project aimed at providing AI capabilities in a command-line based environment. This suite of tools enables users to interact with AI services directly from their terminal.

Some of these tools currently require a working connection to the OpenAI API. Future versions of this command line tool collection should use local self-hosted services. We implement those function in the susi_api project.

Available Commands (all to be implemented)

• run <model-name> - Run a model with the given name and chat.
• ask <model-name> - Provides an answer to a submitted question.
• context - Select, view or create a context for the ask command.
• complete - Takes a piece of text and uses AI to generate a contextually relevant completion.
• listen - record audio and optionally pass it to the transscript command
• transscript - Listens to audio input and transcribes it into text.
• translate - Translates a block of text from one language to another.
• say - Converts text to speech. (Note: This command is currently available on macOS. by default)

Usage

You can use susi_shell in one of two ways:

• interactive: run ./susi_shell.sh run <model-name> to use the susi_shell in a similar way as ollama
• batch: run echo "hello world" | ./susi_shell.sh ask <model-name> to use stdin as input for a prompt to the <model-name>