GREB by Cheetah AI

How this started - We were working on a vector based RAG system for our Coding agent Cheetah AI (we were trying to make a context engine like augment code) , There we found that Indexing the entire codebase and retreving chunks is a pain in the ass and other approach that agents take with grep, the problem is that half time agent was taking on greps while working on large codebases. So we decided to take a different approach with greb - We tried to run multiple parallel grep calls and paired that with our Read and Retreival pipeline accelerated by GPU to simulate an agents entire file finding process in just a few seconds (Which normally takes a minute in large repositories). We ended up with something pretty cool here. PS - While this is a good version of our product, it can still be better with yout support, What we are working on is still too niche and we are afraid of wasting our time on something that apparently no one cares about . This is a good version but still we can use faster GPU inference and better trained Models for our use case. So if you find this useful for you we would love some response and validation so that we can get the confidence to go all in and make it the best we can. Thank you

GREB looks like a promising product can't wait to see it's impact. congratss onn the launch guys. Goodd luck!