Actionable Template-driven Research: Operational Meta-Scientific Arts & Crafts

All right. Hello, welcome everyone. This is ActiveInference Stream number 16.1. It's December 2nd, 2025. I'm going to start the stream with a GitHub push. Everything is on the GitHub repo doxology template. We will reload that and we're live. Looking forward to your comments and any ideas or questions in the chat. Here is where we are going today. We are going to a test-driven development, configurable, end-to-end test suite operation, data analysis and visualization, validation steps, and document rendering, including with citations and clickable links and equations formatted as a scientific document, as well as a variety of other fun formats, including HTML files that can be opened for the web and slides for conversion to slides. That's the document rendering functionality. There is a LLM review output section that provides review and improvements and suggestions for the manuscript in English. There's a translation output using LLMs locally to translate. Then there are literature database search through APIs and document download and local LLM summarization, which we're going to use. So without getting too far ahead of ourselves, let's go to the agenda. Okay. The title of the stream is Actionable Template Driven Research, which is going to get at the core of what this template repo is. Operational Meta Scientific Arts and Crafts, which I guess is like the genre. The stream is live on YouTube on the Active Inference Institute channel. This will be a live stream interactive walkthrough, and we can also do some improvements and some other random stuff. See where it goes. Greetings Upcycle. Good to see you. Tools. I'm using a Mac computer. Cursor 2.146 with mostly the auto and the opus recently released, available at a discount, opus 4.5. And locally, Ollama will be using Llama 3 and Quan type models. Tools. Okay. Here in this, we can look at this on GitHub. There's a few big themes that are going to come up and I'm sure many more. This is just a first step. There's going to be a big focus on end-to-end configurability and ease of reproducibility. There's going to be the usage of a lot of thin orchestrators and articulated reference, which means that the modular flexible methods are underlying in a kind of single composable toolkit of the tool. And then, there's going to be a set of affordances and orchestration happens separated from those tools, which gives good ability to express different workflows and swap out different tool implementations. There's going to be a two-layer model to this repo overall, which I might as well describe here. The project folder is self-contained and is intended to be at the project level, like the specific research paper that's going to get generated. And this is going to be a test-driven package alongside a manuscript expressed in these modular markdown files, which are undergoing this internal coherence test suite, which could be zero tests or you could use it extensively, with the SRC code orchestrated through tests and scripts, which in the context of a certain formatting for the manuscripts in markdown, now pop out of the project, is going to be utilized by the template repo using the infrastructure modules and the scripts in the infrastructure and ultimately this top level run.sh at the top level, which is going to pop up this main menu where we can run the tests, we could render the manuscript, we could run the scientific review, we'll just start that one up now. So here we already, in this, I have these tabs here is one process going with the local LOM summarizing papers that were downloaded from a literature search. We'll look at that step. Here is the output of the translation into Chinese, Hindi, Russian. Here is the manuscript rendering, which includes the full test suite, test suite within the project level of the SRC here with respect to these tests within the project, and then the infrastructure level tests, which are in the top level test folder. So double test suite, then running the scientific…