
Experience the cutting-edge in code search with bloop, a GPT-4-powered engine that transforms how developers interact with and explore codebases. In this video, we'll provide a comprehensive overview of how you can use natural language to ask questions about any GitHub repository, such as the newly released Twitter algorithm. With bloop's innovative features, including conversational search, regex search, and precise code navigation, you'll gain unprecedented insight into your code. Sync seamlessly with GitHub, utilize advanced query filters, and benefit from the seamless integration of the powerful GPT-4 AI model. Bloop stands as a shining example of the power of open-source development, leveraging the Rust ecosystem, Tantivy, Qdrant, and Tauri for exceptional performance and versatility. Not only will we guide you through the process of using bloop to explore the Twitter algorithm, but we'll also demonstrate the tool's potential with a variety of codebases, showcasing how bloop can revolutionize your approach to coding challenges. Whether you're a seasoned developer or just starting, bloop's natural language search capabilities make it an invaluable asset in navigating complex codebases, improving efficiency, and enhancing your overall understanding of projects. Don't miss this opportunity to dive into the next generation of code search and exploration. Join us as we unveil the power of bloop, the GPT-4-powered code search engine that's reshaping the way developers tackle their work. Experience first-hand the synergy between GPT-4, open-source projects, and bloop's unique features, as we embark on a journey to unlock the full potential of code navigation.
Weekly deep dives on AI agents, coding tools, and building with LLMs - delivered to your inbox.
Free forever. No spam.
Subscribe FreeNew tutorials, open-source projects, and deep dives on coding agents - delivered weekly.
--- type: transcript date: 2023-04-03 youtube_id: MF_Trk46vs8 --- # Transcript: Bloop: The GPT-4 Based Code Search-Engine all right in this video I'm going to be showing you bloop so bloop is a code search engine that uses gpt4 to answer questions about your code so the way that it works is you can link up a local repo or a GitHub repo now these have to be within your repository so for instance if you want to look and have a conversation with someone else's repo you'll first have to Fork it and actually have it either locally or Linked In Your GitHub repository now once you have it you'll be able to see the repo here and sort of interesting timing with making this video is the Twitter algorithm was just released so I thought it'd be interesting to take a look at the algorithm so I forked a copy of it it's within my GitHub repository now and now once you install bloop and I can show you where to to do this if you just read through their their getting started here if you go to download the app there's a way you can do it through the command line as well if you'd like to go that route and you if you just look at the system that you're on so I have an Intel based Mac and if you just search that on this page here you'll see that there's um the version that correlates to it just just below there so once you have that installed you'll simply have to link your GitHub repo or look at it look at the repo locally like you see here and then once you have that you have the option of searching the repo with regex or natural language so I'm going to mainly focus on showing you the natural language functionality and like I mentioned so the Twitter algorithm a lot of people have a lot of questions about how it works or you know particular you know interesting tidbits here and there about how a major social media platform ranks and orders things so if we had particular questions about this now one approach could be people could read the source code as many have I've seen lots of threads on people looking through about how it's interesting you know the weighting of how they're waiting you know retweets versus likes versus all a host of things but this is sort of a different approach on how you can look at it so considering how popular you know gpt4 and Chachi BT is this is an interesting way to explore something like this so if I just go here and you can let's start off with a simple question so what does this repo do now if you just query it like this it's going to have a similar interface to chat gbt but the unique part about it is it will also give you the results here within the files so you can imagine how this could be useful where say in a different context where maybe you're onboarding to a company let's say and you're tasked with building a particular feature but you don't know where necessarily all that feature might tie in so maybe you you know there's a um a similar component let's say of the feature that you have to build and by simply querying it with natural language here it could scan that repo and it could answer questions like you know consider linking it here and reference it in this API endpoint and you know it's using graphql for you know uh the the the certain part of the feature that you're trying to build out now that's just an example but here so if we just ask broadly okay what does this repo do so we see this repo contains the CR mixer candidate generation service for Twitter so it's designed to speed up the iteration and development the candidate generation et cetera et cetera so it's pretty um technical it's not like here's how the algorithm works and and you know the the plain English for it but if you start to ask questions uh to the repo that might be more pertinent to so you know we could say something like how how does the ranking system work in Twitter and let's just see what we got now the one thing to note with this is it it is pretty fast that's the thing that first strike me is now once you actually download the repo it does take a little bit of time to index it and I'm not sure exactly what it does under the hood whether it sets up some sort of embeddings and you know searches based on vector and the relatedness of the the queries that you put in but once you have it it's a pretty interesting thing to play around with so I'm not going to spend too much time here showing you examples but I just thought I'd introduce this it's a great project I've really enjoyed it using it just like the past several days and I just thought it'd be useful for others to check it out so obviously a multitude of use cases here you know we can look at you know the impression based fatigue you know the our answer here it ranks candidates based on the given weights of each algorithm while preserving like and then just as we'll do a couple more here let's just say how much does a re -tweet weight on a post virality okay so retweet based on top tweet scoring function so this calculates the score on retweets uh counts and recency it normalizes the reached week count between 0 and 20. so okay it gives you a range and it gives you a bit of a calculation on how they Factor it so like I said not going to give too too many examples here but hopefully you enjoyed this if you do like comment subscribe and until the next one
Technical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.