Awww. I don’t do social media. I’ll make my wife go in my stead. :rofl:

I started that one in 2015 or early 2016 @PandaPi
Time flys!

Right on! I live vicariously through my better-half too we been married 28 years now. I’m over 50 :winking_face_with_tongue:

Im very public lol. I ran conventions and stuff and my experiment made it where anyone can have a very long conversation about me with Google I built that as a proof of concept lol..

I’m a little confused here. Are you talking about chatgpt or the API, and if it’s ChatGPT why are you using a project with source files explaining the project and giving it the resources it needs?

I have not had this problem since the projects function was created.

No I am talking about python and a full repo. It’s not ChatGPT it’s not any model in particular it is a program.. Do you know how to run code @windysoliloquy ?
Out of hearts.

It’s standalone Python program, not ChatGPT Projects or any specific API. The repository is a local knowledge system that stores project context in HTML, indexes it, and retrieves the relevant information when needed. You can connect it to the OpenAI API, another model API, or a local model, but that’s optional. The repository itself is model-agnostic.

The point is that when a new runtime starts, it doesn’t rely on chat memory. My program rebuilds the project’s context from the local knowledge base and passes only the relevant information to whatever model you’re using. That’s why I called it a solution for runtime or session discontinuity, not a ChatGPT feature.

So the API isn’t the project—it’s just one possible inference engine that can sit on top of it.

MPLPB's answer to runtime discontinuity is that continuity lives in the artifacts. A fresh session with no memory does not need the human to re-explain the project; it needs to be pointed at a structure that explains itself. The public build achieves this with published, cross-linked documents recovered through search.

I will be 53 in 9 days. I’d like to say I still feel 30, but for all the exercise and martial arts I did, my knees tell me otherwise. :rofl:

@mitchell_d00 got this curiosity, what’s the magnitude range of files that are indexed that you are talking about in the project? Traversing files shouldn’t be a biggie in modern hardware. Eventually why not manage that second context through a vector store locally and corresponding local LLM?

Right i understand all of that…
But i’ve never had a problem switching sessions with the proper context files already embedded in what I’m using…

[@windysoliloquy
Right, and that’s exactly the distinction I’m making. You’re talking about ChatGPT Projects with embedded context files. I’m talking about a standalone Python knowledge system that works independently of any chat platform. If I switch models, APIs, or even run locally, I don’t lose continuity because the context is rebuilt from my indexed corpus rather than being stored inside the chat. They’re solving similar problems in different layers of the stack.

@DysTopia
The global/public version is huge because it spans years of cross-platform material and uses the broader web as its retrieval substrate. The local version I’m discussing here is much smaller and intentionally bounded, so raw file traversal is not the challenge.

The purpose of the local crawler is to preserve the corpus structure: document identity, metadata, scope, relationships, revision status, and which version is current. And yes, a local vector store with a local LLM is completely compatible with it. That would be one deployment configuration.

I just don’t make vectors or a particular LLM mandatory. The same source corpus can sit behind SQLite full-text search, a hybrid keyword/vector store, or another index, with either a local model or an API on top. The local HTML web is the source of truth; the index and inference engine are replaceable layers. The larger global version is where scale becomes a more serious concern.

Well, I was talking about chatgpt and the API

Now that I understand that you’re talking about switching models and companies, I get why you did what you did.

Sorry don’t see many posts here that are for things outside of OpenAi projects.

No I’m so happy you chatted @windysoliloquy . It’s just a project, that I’m inviting the community. I’m fascinated by continuance (continuance intention) and corrigibility. Your questions are wonderful and most invited

I’m out of hearts

It tickles me pink folks want to chat about it!
Thank you to everyone who is here and will be :infinity::heart:

Infograph of the local version.