April Agents Bring May Simulations: OpenManus, RxInfer.jl, and AgentLaboratory

All right, welcome everyone. It is April 2nd, 2025, and this is Active Inference Stream number 11.1. And I'll begin with a GitHub push. Looking forward to everyone's updates and ideas and questions in the chat. This should be an interesting stream. So, the title is April Agents Bring May Simulations. OpenManusRxInfer.jl and Agent Laboratory. I have a few things prepared, a few software implementations we can play around with. Also, some topics to explore. Not sure how any of the demos will go live. However, it's the hope that a few things will be pretty interesting and we can fix and explore along the way. So, April Agents Bring May Simulations. Or as Alexandra just pointed out, April Agents May Bring Simulations. There are many exciting developments and advancements in the multi-agent synthetic intelligence systems space right now. In a pretty broad sense. Short-term, medium-term, all the terms. Let's explore a few articulating, complementary, interesting kinds of multi-agent frameworks. Autonomous frameworks. Look at different ways of looking at what makes agentic. What kinds of systems engineering are being done within and among the agents. All these kinds of orchestration questions. Looking where we're at instantaneously in April 2025. And more broadly. So, there will be some jumping around since some of the pieces take different amounts of time to load or play out. So, again, I'm not sure what exactly will come through. That'll be part of the fun though. On all of these branches, I've been having some interesting times. There are three cursor windows open. So, starting with the two LLM type models. Right now is running in this terminal. The work of Samuel Schmidgall, Agent Laboratory. I'll read the overview. Agent Laboratory is an end-to-end autonomous research workflow meant to assist you as the human researcher towards implementing your research ideas. Agent Laboratory consists of specialized agents driven by large language models to support you through the entire research workflow. From conducting literature reviews and formulating plans to executing experiments and writing comprehensive reports. This system is not designed to replace your creativity, but to complement it. Enabling you to focus on ideation and critical thinking while automating repetitive and time-intensive tasks like coding and documentation. By accommodating varying levels of computational resources and human involvement, Agent Laboratory aims to accelerate scientific discovery and optimize your research productivity. This has an associated paper describing more. So, I got it to work initially. Output a really awesome 17-page paper, images, equations. I have the document around. Then I started modifying it, trying to get more logging, get more access to it. And kind of went too far. And now I've just pulled back to the main branch that they've provided. Just to show this experiment while this one's running. And then we'll see how long it takes. Hopefully it'll finish within the stream. This YAML file, which they provide some initial examples of. I made a variant that has the following sections. Copying it over with control-I, write a new YAML compliant file that has this semantics. The goal is to develop a partially observable Markov decision process, POMDP, implementing active inference that is in a thermal homeostatic setting. So, kind of the classic air conditioner, do-nothing heater temperature. The API key for OpenAI is put here. And the model is provided for the LLM backend and the review, I guess, separately. Language. Here's different stages. So, I found that when larger numbers were used, there were some interruptions. I don't know if that was because of stepping away from the computer. So, these are relatively low levels of steps, but it'd be conceivable and interesting to see what happens when literature review is huge numbers or what happens there. How to use a structured knowledge base for literature review. It's also possible to run…