[{"data":1,"prerenderedAt":295},["ShallowReactive",2],{"project-38368570":3},{"id":4,"name":5,"line":6,"line_source":7,"title":8,"url":9,"domain":10,"maker":11,"posted_at":12,"year":13,"points":14,"comments":15,"rank_in_year":16,"shown_in_year":17,"top":18,"tier":19,"category":20,"subcategory":21,"format":22,"keywords":23,"reception":29,"sentiment":30,"stance_mix":31,"scored_comments":35,"aspects":36,"strip":43,"post_text":45,"what":46,"said":83,"maker_in_thread":175,"shape":187,"grid":191,"attempts":258,"same_week":259,"maker_shows":40,"maker_hidden":288,"reader_version":289,"card":290},"38368570","Neum AI","Open-source large-scale RAG framework","title","Show HN: Neum AI – Open-source large-scale RAG framework","https:\u002F\u002Fgithub.com\u002FNeumTry\u002FNeumAI","github.com","picohen",1700594429,2023,155,30,392,3109,13,"notable","ai","rag framework","library",[24,25,26,27,28],"rag","vector embeddings","data pipelines","vector database","llm","curious",-0.25,{"positive":32,"neutral":33,"mixed":32,"critical":34},0,14,5,19,{"execution":37,"privacy":39,"idea":41,"performance":42},{"positive":32,"critical":38},1,{"positive":32,"critical":40},2,{"positive":32,"critical":40},{"positive":32,"critical":32},[38,44,38,44,44,38,38,32,32,32,44,32,44,44,44,38,32,32,44,44,44,32,32,32,44,44,32,32,44,44],3,"Over the last couple months we have been supporting developers in building large-scale RAG pipelines to process millions of pieces of data.\n\nWe documented our architecture and approach in an HN post (https:\u002F\u002Fnews.ycombinator.com\u002Fitem?id=37824547) a couple weeks ago. Today, we are open sourcing the framework we have developed.\n\nThe framework focuses on RAG data pipelines and provides scale, reliability, and data synchronization capabilities out of the box.\n\nFor those newer to RAG, it is a technique to provide context to Large Language Models. It consists of grabbing pieces of information (i.e. ",{"problem":47,"audience":53,"pricing":57,"motivation":58,"time_to_build":57,"stack":62,"openness":80,"stage":81,"team_size":82},{"value":48,"source":49,"evidence":50},"Building scalable, reliable RAG data pipelines with sync and metadata","maker",[51],{"id":52},"post",{"value":54,"source":49,"evidence":55},"Developers building RAG applications on large datasets",[56],{"id":52},null,{"value":59,"source":49,"evidence":60},"Supporting developers building large-scale RAG pipelines for customers",[61],{"id":52},[63,71],{"value":64,"source":65,"evidence":66},"Weaviate (vector database)","commenter",[67],{"id":68,"by":69,"text":70},"38370300","eigenvalue","Cool. Do you do any of the relevance calculations directly, or is that all handled by Weaviate? If so, is there any way to influence that part of it, or is it something of a…",{"value":72,"source":49,"evidence":73},"Postgres (via Supabase)",[74,77],{"id":75,"by":11,"text":76},"38376160","Postgres is already available :)",{"id":78,"by":11,"text":79},"38376164","In fact, we will be releasing a blog post in the next few days of how we do real-time syncing for a RAG application with postgres hosted on supabase","open_source","launched","company",{"praise":84,"objections":95,"requests":125,"questions":140,"alternatives":155},[85],{"value":86,"source":65,"evidence":87},"Semantic chunking and metadata handling seen as useful ideas",[88,92],{"id":89,"by":90,"text":91},"38371961","ddematheu","Yeah, we were playing around with doing some semantic chunking. Works okay for some use cases. We have some ideas to go further on that. Generally we have found that recursive…",{"id":93,"by":90,"text":94},"38385405","LlamaIndex is pretty awesome. There are a couple areas where we think we are driving some differentiation. 1. The management of metadata as a first class citizen. This includes…",[96,107,117],{"value":97,"source":65,"evidence":98},"Frameworks are thin ETL wrappers; direct engineering is as easy",[99,103],{"id":100,"by":101,"text":102},"38378204","ramoz","Very bearish on these frameworks and abstractions. Yes, obviously useful for prototyping and creating hype articles & tweets with fun examples. However any engineer is capable of…",{"id":104,"by":105,"text":106},"38378500","hobs","The problem is that business people have been just rubbing together libraries for decades making money (though maybe not the MOST) and will see these frameworks as a \"simple\" way…",{"value":108,"source":65,"evidence":109},"Semantic chunking runs LLM-generated code without sanitizing",[110,114],{"id":111,"by":112,"text":113},"38372156","hrpnk","Don't you find it dangerous to just run the code w\u002Fo any sanitizing? Why not capture a few strategies that the LLM returns as code that can be properly audited (and ran locally…",{"id":115,"by":112,"text":116},"38371101","Interesting to see that the semantic chunking in the tools library is a wrapper around GPT-4. Asks GPT for the python code and executes it:…",{"value":118,"source":65,"evidence":119},"Little differentiation from LlamaIndex or Langchain",[120,124],{"id":121,"by":122,"text":123},"38377548","rmonvfer","How is this any different from LlamaIndex [1]? [1] https:\u002F\u002Fwww.llamaindex.ai",{"id":104,"by":105,"text":106},[126,133],{"value":127,"source":65,"evidence":128},"Expose more relevance and ranking control at application level",[129,132],{"id":130,"by":69,"text":131},"38371803","Got it. I'd encourage you to expose more of that functionality at the level of your application if possible. I think there is a lot of potential in using more than just cosine…",{"id":68,"by":69,"text":70},{"value":134,"source":65,"evidence":135},"Support for more data sources such as Redis",[136],{"id":137,"by":138,"text":139},"38374480","imperio59","Why MySQL and not PostgreSQL or Redis on the roadmap for sources?",[141,148],{"value":142,"source":65,"evidence":143},"How does it improve retrieval compared to a plain vector DB",[144],{"id":145,"by":146,"text":147},"38370662","J_Shelby_J","How does the improve upon retrieval compared to just using any vector db and semantic search?",{"value":149,"source":65,"evidence":150},"Why use this instead of Haystack",[151],{"id":152,"by":153,"text":154},"38370607","alchemist1e9","If someone is about to start their project using Haystack would you suggest they instead look at Neumtry?",[156,160,164,168],{"value":157,"source":65,"evidence":158},"LlamaIndex",[159],{"id":121,"by":122,"text":123},{"value":161,"source":65,"evidence":162},"Haystack",[163],{"id":152,"by":153,"text":154},{"value":165,"source":65,"evidence":166},"Langchain",[167],{"id":104,"by":105,"text":106},{"value":169,"source":65,"evidence":170},"MemGPT",[171],{"id":172,"by":173,"text":174},"38370898","omarfarooq","Have you guys connected with MemGPT?",{"comments":34,"answered":176,"key":177},4,[178,181,184],{"id":179,"text":180},"38370572","Relevance calculations are handled by the vector db but we try to improve such relevance with the use of metadata (you will see how our components have \"selectors\" so that…",{"id":182,"text":183},"38373272","Gotta start somewhere! Take a look at this one as well! https:\u002F\u002Fnews.ycombinator.com\u002Fitem?id=37824547",{"id":185,"text":186},"38370678","Well, of course I'm biased on the answer :). But to give a not-so-biased answer, I would first try to understand what the project is about and whether RAG is a priority in it. If…",{"commenters":188,"max_depth":176,"maker_replies":34,"first_comment_mins":189,"lifespan_hours":190},17,117,26,[192,194,196,198,200,204,207,209,211,213,217,219,220,221,223,226,230,232,235,237,238,240,241,244,247,249,252,253,254,256],[115,32,38,193,112,32],"project\u002Fcritical\u002Fexecution",[89,32,44,195,90,32],"maker\u002Fneutral\u002Fexecution",[111,32,38,197,112,32],"project\u002Fcritical\u002Fprivacy",[199,32,44,195,90,32],"38372370",[201,32,44,202,203,32],"38372557","project\u002Fneutral\u002Fexecution","treprinum",[205,32,38,197,206,32],"38375953","smokel",[100,38,38,208,101,32],"project\u002Fcritical\u002Fidea",[104,38,32,210,105,32],"commenter\u002Fcritical\u002Fidea",[212,38,32,210,90,32],"38385500",[214,40,32,215,216,32],"38370452","other\u002Fneutral\u002Fnone","westurner",[68,44,44,218,69,32],"project\u002Fneutral\u002Fperformance",[179,44,32,202,11,38],[130,44,44,218,69,32],[145,176,44,222,146,32],"project\u002Fneutral\u002Fidea",[224,176,44,225,90,32],"38370807","maker\u002Fneutral\u002Fidea",[227,34,38,228,229,32],"38373208","maker\u002Fcritical\u002Fidea","matmulbro",[182,34,32,231,11,38],"maker\u002Fneutral\u002Fnone",[233,34,32,215,234,32],"38374467","obiefernandez",[121,236,44,222,122,32],6,[93,236,44,225,90,32],[152,239,44,222,153,32],7,[185,239,32,225,11,38],[242,239,32,243,153,32],"38372340","commenter\u002Fneutral\u002Fidea",[245,239,32,243,246,32],"38377088","zansara",[248,239,44,225,153,32],"38383183",[137,250,44,251,138,32],8,"project\u002Fneutral\u002Fnone",[75,250,32,231,11,38],[78,250,32,231,11,38],[172,255,44,251,173,32],9,[257,255,44,231,90,32],"38371943",[],[260,265,270,274,279,284],{"id":261,"name":262,"points":263,"category":264},"38327017","Ad Accelerator",713,"consumer_social",{"id":266,"name":267,"points":268,"category":269},"38316936","UneeBee",459,"learning_science",{"id":271,"name":272,"points":273,"category":269},"38365638","Pickcode",420,{"id":275,"name":276,"points":277,"category":278},"38285228","screenshot-to-code",358,"dev_tools",{"id":280,"name":281,"points":282,"category":283},"38360776","Stella Nera",154,"hardware",{"id":285,"name":286,"points":287,"category":264},"38295535","David Adler personal landing page",152,false,"v2",{"id":291,"name":5,"line":6,"ai":288,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":30,"top":18,"kw":292,"maker":11,"domain":10,"tier":19,"attempts":32,"strip":293,"url":9,"mix":294,"praise":86,"objection":97,"sub":21},38368570,[24,25,26,27,28],"131331100030333100333000330033",[32,33,32,34],1794036645026]