A/B/C Agent Builder Club - Plan, then Act - Multi Agents


Details
Going to take a look at DERA, but for improving tests.
DERA: Enhancing Large Language Model Completions with Dialog-Enabled Resolving Agents
https://arxiv.org/abs/2303.17071 (no need to read the paper)
The idea is
I'll briefly present the concept of the agent
Short look over the code
Discussion / Hacking as you like
For those who are just starting out I can help with some foundation tasks like connecting to models, setting up libraries, parsing output etc.
Idea is to start without langchain, but we might lean on some libraries later on. Primarily Python but JS folks should be able to play along if it is their daily driver.
Ideas for other sessions include:
DERA style actor critic (this one)
Web research agent
RAG
Document summarization
Memory (checklist, tickets, done)
Reflexion (done)
Chatting alongside other activity
Cooperative (Planner-Actor, done)
YouTube video summarization / precis (done)
Long running agents (done)

A/B/C Agent Builder Club - Plan, then Act - Multi Agents