twentyEyesLLM

Use LLM to generate a daily activity report on computer. You can run it using a inference server locally or remotely with Ollama and OpenAI.

This small program basically take a screenshot between a time-interval, submit it to the inference server along a system prompt and process the result, aggregating them all into a local history file. Then you can generate Markdown reports about your local computer activities in a specific day.

Requirements

• Python 3.11.8
• Ollama
• At least an i7 CPU (or equivalent), 16gb ram and a dedicated GPU. This program was tested on Macbook Pro early-2018, Macbook Air M2 and Steam Deck (with Arch-based SteamOS).

Usage

Running with Ollama

$ python -m twentyEyesLLM --ollama --interval 7

Running with Openai

$ OPENAI_API_KEY=...
$ python -m twentyEyesLLM --openai --interval 7

DISLAIMER (1): During the first running using Ollama, this program will download the necessary models to perform the image inference, which since the current date is LLAVA3.

DISCLAIMER (2): During the first running you will be asked to grant Screen Recording permissions to the terminal. It is necessary in order to allow this program to take screenshots and submit to the inference server, remember, you can choose between use a local inference server (Ollama) or a remote server (OpenAI).

Show local history and generate report with selected history

$ python -m twentyEyesLLM history --list
24_06_02
24_05_04
24_06_07
24_06_06

$ python -m twentyEyesLLM report --date 24_06_07 > my_report.md
$ open my_report.md

Here you can see how a report looks like.

Clear local history

$ python -m twentyEyesLLM history --clear

Get the full arguments descriptions

$ python -m twentyEyesLLM --help

About the name

"Twenty Eyes" is one of my Misfits favorite songs from their debut studio album "Walk Among Us". From my perspective, the song talks about the spider capability to see everything with twenty eyes.