Jeeves assistant

• Description
• Features
• Dependencies
• Getting started
  • Start Ollama
  • Start OpenAI

Description

Jeeves is a virtual assistant based on a LLM and running as a Telegram bot.

The bot has two main commands:

• new to clear the conversation history and start chatting again;
• start/help/info to show the info menu;
• shutdown to shut down the host machine (to gracefully shut down if you remotely control the device power);

All sent messages are replied by the assistant.

Features

It can work with:

• a local model from Ollama;
• a model from an external OpenAI-compliant provider;

Make sure to create the .env file with the variable you need to run one or another.

Dependencies

Before getting started, make sure you have obtained your Telegram bot token and set it in the .env file.

Getting started

To get started place in the base project folder and type:

make build

Depending on the service you want, go to start ollama or start openai

---

Start Ollama

To start the Ollama based version, type:

PROFILE=ollama make up

The profile will also start the local ollama service to host the model.

Only the first time you will also have to download the model, so you will also have to type:

make ollama-first-setup

---

Start OpenAI

To start the OpenAI compliant based version, just type:

make up

---

Once services are up, to tail on logs, type:

make tail

Logs will also be persisted at stack/logs.

To stop the services, type:

make down