Transcript: So, today I was able to make the burrito deployable, or at least slightly more easy to deploy, in the sense that there's a docker-compose file now, and that docker-compose file sets up a ChromaDB instance for your vector store, as well as running the burrito server itself. So that's pretty good. It was a lot of back and forth this morning on exactly what to do or how to deploy this, particularly in the sense of some of me wanting to make it extremely automated so I can have it be self-service. That is still an item on the list to do, is a self-service burrito. But we're not there yet, and I don't know if that's super critical for this moment. Let me just write that real fast, self-service, and then we want to rename. Okay. Yeah, and it was a bit interesting just going through Docker and being like, huh, how do I do this? And then also doing some research of like, is docker-compose used to deploy bigger services? And kind of the answer is no, basically, which isn't that surprising. I mean, I'm building in a very strange way in the sense like it's meant to run on an individual person's machine as opposed to wanting to be run on, you know, like a real-ass computer. So that's a bit different than most, I would say. And yeah, I think it's just impacted how these things get deployed. It's not like I just have a database and I have a blah, blah, blah. It's like, oh no, I'm trying to run like a hundred of the same service but with a different configuration, potentially. Like if I'm going to run it for friends or whatever and everyone has their own. It's just like a different way of running and building things, quite frankly. So yeah, it's interesting. I think some of these components like the vector store, I don't know, that's probably eventually just going to become part of the main binary. I've already written part of a vector database myself, and the performance has got awful. But it wouldn't be that hard, I think, to write a fast SIMD version or something like that, or something that's like CUDA accelerated or M1 accelerated. I think that's probably the next step is just like a vector database which is tuned to the specific application and purpose, rather than trying to use something off the shelf. But that's like a much further down the line problem. For now we can run the microservice of Chroma. It's just so much extra memory for very little gain, I think. And the performance is unknown. It's much easier to just try a bunch of different cosine algorithms over a test set that you have and be able to just inspect it. And that's what I did when I wrote it. This algorithm clearly works best for this data, so that's the algorithm that I'm going to use for this vector store. So yeah, there's still lots of abstractions and stuff internally to be made, and making it more developer-friendly and all of this. But we're on our way. We're on our way.
I finished making breakfast tacos and have a day planned with reading, meditation, and possibly writing. My focus for the day is improving the main burrito website to make it interactive and pondering the challenge of making it easily deployable for others without the pain of maintaining a separate instance. I'm contemplating whether to maintain the existing system or create a new one, maybe using Docker to bundle the components. The decision is complicated due to the unique architecture of building decentralized applications and requires further thought.
84.76% similar
Today marked a significant advancement in the burrito project, where the image pipeline, established the previous day, became fully functional and integrated into a webpage, complete with an effective querying system. The visual aspect of the project, particularly the image embeddings, was both intriguing and aesthetically pleasing, although its effectiveness is still under review. The project is now at a stage where the creator is keen to move beyond personal experiments to sharing the results with others, with the immediate goal being to encourage a small group of individuals to test the developments. The focus for the week has shifted to actual user engagement through getting people to sign up and provide feedback, driven by the enthusiasm of witnessing the project's imagery features come to life.
The realization of the value in this application lies in its ability to perform searches quickly, efficiently, and accurately. There are multiple approaches to enhance its functionality, with a focus on both data storage and the improvement of search capabilities, which is currently the most critical yet challenging aspect. Concerns exist about the app's method of aggregating all processed data, which feels inherently flawed, though it's being temporarily accepted for the valuable data it provides. This tension between a recognized need for development against the reluctance to proceed with an imperfect solution underscores the complexity of the problem at hand.
I had a really nice day focused on improving and optimizing our API, dealing with various issues along the way. I managed to create and deploy Docker images, enhancing the overall performance and deployability of our system, which I'm quite excited and satisfied with. Despite it taking more time than anticipated and pushing back other projects, I'm optimistic about continuing this work. Additionally, I spent quality time with friends, preparing and enjoying meals together, which was a warming end to the day. Although feeling slightly under the weather, I'm hopeful for tomorrow and plan an early start for a highlining adventure, pending how I feel in the morning.
82.64% similar
I'm starting a 15-minute meditation before diving into my project tasks. My goal is to simplify the deployment of containers on our own hardware, making it more efficient to manage resources without external assistance. The tasks ahead include publishing a Docker image, adding backup functionalities to my API, incorporating speech capabilities into a library, and crafting the API to be serverless. Furthermore, I plan to utilize Cloudflare's key-value store and repeatedly summon "Hano" for reasons unspecified, concluding my note with gratitude.