I'm working on how a Langchain agent executor can basically idle until either a scheduled event occurs that it needs to run or a new task is added to its queue. Right now I'm in the planning stage but my thinking is I would build a framework that the agent executor runs which allows it to learn about new tasks and schedules. Basically like an event loop in programming. Each task or scheduled event would be an agent with its own set of tools that the LLM could use to execute the task. Each task would log all of its steps as well as the outcome so a log reader agent could be invoked to show me information relevant to the task that was executed. I'm thinking of using markdown for outcome display running behind a simple web server. I'm also planning on building a simple notification system that could display on my local Mac or send a message through the Pushover system so I can see anything critical on my iPhone or iPad when I'm not at my desk. #

More to come.#

It's been a long while since my last post! Since writing last I have been going down the AI/ChatGPT rabbit hole with gusto! My intention going forward is to document what I am learning and how I am using the knowledge. Even if no one else reads this I will have a record of my trials and tribulations to keep me warm at night! 😂#