AssistMe

This is just a simple CLI tool written in python for basic interaction with LLM's. Currently supports OPenAI's GPT-3/4 chat models as well as Googles PALM model.

---

Setup

Installation is possible directly from github using pip.

pip install 'git+https://github.com/GandalfsDad/assist-me.git'

An OpenAI API key is required and should be in the environment variable OPENAI_API_KEY. A Google API key is required to use the PALM model and should be in the environment variable GOOGLE_API_KEY.

export OPENAI_API_KEY=<your key here>
export GOOGLE_API_KEY=<your key here>

---

Usage

There are two supported interaction subcommands: chat and simple. Simple is a one response version of chat where you can ask a question and get a single response. Chat is a more interactive version where you can ask multiple questions and get multiple responses.

Simple

assistme simple --input "What is the meaning of life?"
>>> Response: The meaning of life is to live a life of meaning.

If the input flag is not provided the user will be prompted for one.

assistme simple
>>>Input: 

Chat

assistme chat
>>>Input: 

The chat interface will await a user input and then respond and await another input until you kill the process or type exit,q, or quit.

There are also several supporting subcommands related to managing profiles and checking versions.

Profile Commands

save-profile - saves a profile to file

assistme save-profile --profile "Gandalf" --system "You are Gandalf the wizard from Tolkien's Novels. You respond only as he would"

show-profile - shows the details of a profile

assistme show-profile --profile "Gandalf"
>>> Profile: Gandalf
>>> System: You are Gandalf the wizard from Tolkien's Novels. You respond only as he would

show-profiles - shows all saved profiles

assistme show-profiles
Found 1 profiles
>>> Profile: Gandalf
>>> Prompt: You are Gandalf the wizard from Tolkien's Novels. You respond only as he would

Version

assistme version
>>> AssistMe version: 0.0.4

---

Options

--name / -n [name]

The name option can be used to specify the name of the Assistant. This is mostly only relevant for chat but possible to use with simple as well.

assistme simple --name "Gandalf"

--model / -m [model]

The model option can be used to specify the model to use. Currently only gpt3,gpt4 and palm are possible with gpt3 being the default. Values are not case sensitive.

assistme simple --model gpt4

--system / -s [system]

The system option can be used to specify the system message. For PALM this will be the context message,

assistme simple --system "You are Gandalf the wizard from Tolkien's Novels. You respond only as he would"

--profile / -p [profile]

The profile option can be used to specify the profile to use. If not provided the default profile will be used. If a system option is provided it will also save / overwrite the profile with the provided system message. If no system option is provided and profile is provided it will prompt the user for a system message.

assistme simple --profile "Fred" --input "What is the height of the Eiffel Tower?"
>>>System Prompt:

--max-tokens [max-tokens]

Sets the max tokens for the API call. This is the maximum number of tokens that will be returned in the response. Default is 750.
This is only relevant for OPENAI models. For PALM the max tokens are inherited from the model class. Something to look into when the PALM API is properly released.

assistme chat --max-tokens 100

--temperature [temperature]

Sets the temperature for the API call. This is the randomness of the response. The higher the number the more random the response. Default is 0.5.

assistme chat --temperature 0.2

--input / -i [input]

This is an option for the simple command only. If it is omitted you will be prompted for input.

---