LLM Agents

Agents use LLMs to take action or provide curated reports.

News Bot

The news.py script will use an OpenAI API based LLM to process raw weather payloads, stock prices and top news items into a curated news report.

This can be set up as a scheduled cron job to send a personalized news update (see news-cron.sh). You will need to set up the msmtp mail service to send email. See instructions https://wiki.archlinux.org/title/Msmtp.

echo "Gathering the news for you..."

# Update these configurations
export ABOUT_ME="I'm a 35 year old woman who lives in Los Angeles. I work at Acme in technology and AI. I have 2 boys."
export COMPANY="MyCompany"
export ALPHA_KEY="alpha-key" # Get Alpha Advantage API Key for Stock Prices - https://www.alphavantage.co/
export OPENAI_API_BASE="http://localhost:4000/v1"
export OPENAI_API_KEY="sk-3-laws-safe"
export LLM_MODEL="llama"
export CITY="Los Angeles"
export CITY_WEEKEND="Ventura"
export EMAIL_FORMAT=true

echo ""
python3 ./news.py

Movie Bot

The movie.py script uses an LLM to make a movie suggestion. It records previous recommendations to prevent repeating suggestions. Outputs a single line recommendation and writes a full detailed recommendation to MESSAGE_FILE. This can be set up as a scheduled cron job to send you a text or email you a recommendation.

echo "Thinking about a movie recommendation for you..."

# Update these configurations
export OPENAI_API_BASE="http://localhost:4000/v1"
export OPENAI_API_KEY="sk-3-laws-safe"
export LLM_MODEL="llama"
export ABOUT_ME="We love movies! Action, adventure, sci-fi and feel good movies are our favorites."
export DATABASE="./movie.db"
export MESSAGE_FILE="./message.txt"

python3 ./movie.py

Conversational Agent

The conversational.py script uses an LLM (or two LLMs) to create a back and forth conversation. This script sets up a teacher and student identities. The teacher and student LLMs take turns responding to each other. The conversation continues until a stop prompt is given or a maximum number of rounds is reached. The teacher LLM then provides a summary of the conversation and an evaluation of the student.

python3 conversational.py