[{"data":1,"prerenderedAt":591},["ShallowReactive",2],{"project-42174829":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":36,"aspects":37,"strip":46,"post_text":48,"what":49,"said":88,"maker_in_thread":193,"shape":205,"grid":209,"attempts":549,"same_week":554,"maker_shows":40,"maker_hidden":584,"reader_version":585,"card":586},"42174829","FastGraphRAG","Better RAG using good old PageRank","title","Show HN: FastGraphRAG – Better RAG using good old PageRank","https:\u002F\u002Fgithub.com\u002Fcirclemind-ai\u002Ffast-graphrag","github.com","liukidar",1731951793,2024,457,119,82,3515,3,"legend","ai","retrieval augmented generation","library",[24,25,26,27,28],"graphrag","rag","knowledge graph","pagerank","retrieval","curious",-0.11,{"positive":32,"neutral":33,"mixed":34,"critical":35},7,31,1,6,45,{"idea":38,"performance":41,"execution":43,"docs":44,"pricing":45},{"positive":39,"critical":40},4,2,{"positive":34,"critical":42},0,{"positive":42,"critical":42},{"positive":42,"critical":40},{"positive":42,"critical":34},[42,18,42,47,42,42,42,42,47,42,18,47,42,18,42,34,42,34,18,40,18,42,42,42,18,18,47,18,18,18,18,39,18,18,42,47,42,42,34,34],5,"Hey there HN! We’re Antonio, Luca, and Yuhang, and we’re excited to introduce Fast GraphRAG, an open-source RAG approach that leverages knowledge graphs and the 25 years old PageRank for better information retrieval and reasoning.\n\nBuilding a good RAG pipeline these days takes a lot of manual optimizations. Most engineers intuitively start from naive RAG: throw everything in a vector database and hope that semantic search is powerful enough. This can work for use cases where accuracy isn’t too important and hallucinations are tolerable, but it doesn’t work for more difficult queries that invol",{"problem":50,"audience":56,"pricing":61,"motivation":65,"time_to_build":69,"stack":70,"openness":85,"stage":86,"team_size":87},{"value":51,"source":52,"evidence":53},"Naive RAG with vector search fails on multi-hop reasoning and noisy data and is hard to debug","maker",[54],{"id":55},"post",{"value":57,"source":58,"evidence":59},"Engineers building RAG pipelines over their own data","inferred",[60],{"id":55},{"value":62,"source":52,"evidence":63},"Free hosted tier with up to 100 monthly requests, paid plans, self-hosting available",[64],{"id":55},{"value":66,"source":52,"evidence":67},"Naive RAG is inaccurate and hard to debug; existing graph RAG approaches are naive in how they build and explore graphs",[68],{"id":55},null,[71,77,81],{"value":72,"source":52,"evidence":73},"Python",[74],{"id":75,"by":11,"text":76},"42175597","The graph is currently stored using python-igraph. The codebase is designed such that it is easy to integrate any graphdb by writing a light wrapper around it (we will provide…",{"value":78,"source":52,"evidence":79},"python-igraph",[80],{"id":75,"by":11,"text":76},{"value":82,"source":52,"evidence":83},"OpenAI gpt-4o-mini",[84],{"id":75,"by":11,"text":76},"open_source","launched","team",{"praise":89,"objections":102,"requests":132,"questions":151,"alternatives":174},[90],{"value":91,"source":92,"evidence":93},"Useful idea applying PageRank to graph retrieval","commenter",[94,98],{"id":95,"by":96,"text":97},"42177865","anotherpaulg","Super interesting, thanks for sharing. How large a corpus of domain specific text do you need to obtain a useful knowledge graph? Aider has been doing PageRank on the call graph…",{"id":99,"by":100,"text":101},"42179904","jillesvangurp","Cool idea. IMHO traditional information retrieval is the way to go with RAG. Vector search is nice but also slow and expensive and people seem to use it as magic pixie dust. It…",[103,114,125],{"value":104,"source":92,"evidence":105},"Documentation in README is thin on how it works",[106,110],{"id":107,"by":108,"text":109},"42177360","larodi","Wonder why this all - here on HN - is not part of the readme .md which says absolutely nothing about how and why this all would work. The whole approach to representing the work,…",{"id":111,"by":112,"text":113},"42182753","andai","I might be the wrong target audience (I do have a great interest in this, but I am not doing it at a professional level) but I feel the GitHub could explain things a bit better —…",{"value":115,"source":92,"evidence":116},"Very similar to existing projects like nano graphrag and HippoRAG without credit",[117,121],{"id":118,"by":119,"text":120},"42178597","dantodor","It looks awfully similar to nano graphrag, but I fail to see any credits to it.",{"id":122,"by":123,"text":124},"42176273","zwaps","What is the difference to HippoRAG, which seems to be the same approach but came our earlier?",{"value":126,"source":92,"evidence":127},"Pricing is unclear between free tier and self-hosting",[128],{"id":129,"by":130,"text":131},"42177753","davidsainez","Cool! But I'm confused on your pricing. The github page says first 100 requests are free but the landing page says to self host if you want to use for free. I signed up and used…",[133,140],{"value":134,"source":92,"evidence":135},"Add Neo4j and other graph database connectors",[136],{"id":137,"by":138,"text":139},"42178080","ignomaniac","does FastRAG integrate with other graph databases like neo4J ?",{"value":141,"source":92,"evidence":142},"Benchmark results on multi-hop retrieval datasets",[143,147],{"id":144,"by":145,"text":146},"42181689","gillesjacobs","Do you have any retrieval and generation metric scores (eg, KILT or NQ datasets)? I know benchmark datasets are not the be-all-end-all, but a halfway decent score and…",{"id":148,"by":149,"text":150},"42180433","lordofgibbons","Thanks for releasing this! Have you gotten a chance to run any multi-hop retrieval benchmarks? It would be very useful to be able to compare this method to other establishes RAG…",[152,160,167],{"value":153,"source":92,"evidence":154},"How does it compare to HippoRAG and LightRAG",[155,156],{"id":122,"by":123,"text":124},{"id":157,"by":158,"text":159},"42182260","redwood","I'm curious how the implementation compares with LightRAG (https:\u002F\u002Fgithub.com\u002FHKUDS\u002FLightRAG) ?",{"value":161,"source":92,"evidence":162},"How well does it answer broad questions",[163],{"id":164,"by":165,"text":166},"42180268","fudged71","How is it at answering broad questions? Communities and clustering were specifically for that purpose of agglomerating, right?",{"value":168,"source":92,"evidence":169},"Can it be used with LLMs other than OpenAI",[170],{"id":171,"by":172,"text":173},"42175293","adultSwim","Can this be used with LLMs other than the OpenAI API?",[175,179,185,189],{"value":176,"source":52,"evidence":177},"Microsoft GraphRAG",[178],{"id":55},{"value":180,"source":52,"evidence":181},"HippoRAG",[182],{"id":183,"by":11,"text":184},"42178258","LLMs are only used to construct the graph, to navigate it we use an algorithmic approach. As of now, what we do is very similar to HippoRAG…",{"value":186,"source":92,"evidence":187},"LightRAG",[188],{"id":157,"by":158,"text":159},{"value":190,"source":92,"evidence":191},"nano graphrag",[192],{"id":118,"by":119,"text":120},{"comments":194,"answered":35,"key":195},10,[196,199,202],{"id":197,"text":198},"42178100","This is super interesting! Thanks for sharing. Here we are talking of graphs in the milions nodes\u002Fedges, so efficiency is not that big of a deal, since anyway things are gonna be…",{"id":200,"text":201},"42178206","Thanks for sharing! These are all very helpful insights! We'll keep this in mind :)",{"id":203,"text":204},"42178056","We have tried from small novels to full documentations of some milion tokens and both seem to create interesting graphs, it would be great to hear some feedback as more people…",{"commenters":206,"max_depth":32,"maker_replies":194,"first_comment_mins":207,"lifespan_hours":208},77,30,100,[210,214,216,218,222,226,229,232,234,238,241,244,246,249,253,256,258,260,263,265,268,270,272,275,278,281,285,289,293,296,299,302,306,308,309,311,314,317,319,321,323,327,329,331,332,336,339,343,346,349,352,355,357,360,364,366,370,375,378,381,385,389,392,396,399,402,404,406,408,411,414,417,419,421,423,428,430,432,435,440,443,447,449,451,455,458,460,462,464,466,468,471,475,479,482,483,485,488,490,493,496,500,504,508,510,512,514,517,519,521,524,527,531,534,536,539,541,543,548],[211,42,42,212,213,42],"42177973","commenter\u002Fneutral\u002Fidea","michelpp",[197,42,42,215,11,34],"maker\u002Fneutral\u002Fidea",[217,42,42,212,213,42],"42183404",[219,42,42,220,221,42],"42188168","commenter\u002Fpositive\u002Fidea","lmeyerov",[223,42,18,224,225,42],"42179763","project\u002Fneutral\u002Fidea","arkokoley",[227,34,42,212,228,42],"42177909","LASR",[230,34,42,212,231,42],"42178169","sramam",[200,34,42,233,11,34],"maker\u002Fpositive\u002Fnone",[235,34,47,236,237,42],"42178645","maker\u002Fpositive\u002Fidea","antves",[239,34,42,220,240,42],"42178703","sdesol",[242,34,42,212,243,42],"42179361","yaj54",[245,34,42,212,145,42],"42181552",[247,34,42,212,248,42],"42182937","tweezy",[250,34,42,251,252,42],"42190145","commenter\u002Fcritical\u002Fidea","oedemis",[254,34,42,212,255,42],"42179704","siquick",[257,34,42,212,112,42],"42182734",[259,34,42,212,240,42],"42184700",[261,34,42,212,262,42],"42185565","firejake308",[264,34,42,212,240,42],"42185710",[266,34,42,212,267,42],"42187577","spdustin",[269,34,42,212,112,42],"42186383",[271,34,42,212,240,42],"42187734",[273,34,42,212,274,42],"42183748","mhuffman",[276,34,42,277,228,42],"42187084","commenter\u002Fneutral\u002Fexecution",[279,34,47,236,280,42],"42184367","isoprophlex",[282,34,42,283,284,42],"42186316","other\u002Fneutral\u002Fnone","katelatte",[286,34,42,287,288,42],"42193843","other\u002Fneutral\u002Fidea","itissid",[290,40,42,291,292,42],"42184665","other\u002Fcritical\u002Fidea","inboulder",[294,40,42,212,295,42],"42185135","samsonradu",[297,40,42,220,298,42],"42187213","jsenn",[300,40,42,212,301,42],"42187832","queueueue",[303,18,18,304,305,42],"42176588","project\u002Fmixed\u002Fperformance","AIorNot",[307,18,47,233,237,42],"42176844",[95,39,18,224,96,42],[203,39,42,310,11,34],"maker\u002Fneutral\u002Fnone",[312,39,42,283,313,42],"42185733","ukuina",[315,39,42,283,316,42],"42190012","LiveTheDream",[318,39,42,283,313,42],"42199047",[320,39,42,283,316,42],"42199783",[322,39,42,283,316,42],"42199959",[324,47,18,325,326,42],"42175488","project\u002Fneutral\u002Fexecution","deepsquirrelnet",[75,47,42,328,11,34],"maker\u002Fneutral\u002Fexecution",[330,47,42,283,284,42],"42186268",[99,35,42,291,100,42],[333,32,34,334,335,42],"42180308","project\u002Fcritical\u002Fidea","krawczstef",[144,337,18,338,145,42],8,"maker\u002Fneutral\u002Fperformance",[340,341,42,283,342,42],"42180275",9,"yccheok",[344,341,42,287,345,42],"42180549","martinkallstrom",[347,341,42,287,348,42],"42180583","dheerkt",[350,341,42,287,351,42],"42182362","Tsarp",[353,194,42,283,354,42],"42190493","scorps",[157,356,18,215,158,42],11,[111,358,34,359,112,42],12,"project\u002Fcritical\u002Fdocs",[361,362,18,224,363,42],"42177693",13,"peppertree",[365,362,18,215,237,42],"42177805",[367,368,18,215,369,42],"42186220",14,"staticautomatic",[371,372,18,373,374,42],"42182257",15,"project\u002Fneutral\u002Fperformance","ah27182",[129,376,34,377,130,42],16,"project\u002Fcritical\u002Fpricing",[379,376,18,380,237,42],"42177848","maker\u002Fneutral\u002Fpricing",[382,383,42,283,384,42],"42180236",17,"robrenaud",[386,387,18,224,388,42],"42181314",18,"2dvisio",[390,387,42,283,391,42],"42217736","darkotic",[393,394,42,283,395,42],"42175882",19,"fareesh",[397,394,42,283,398,42],"42175938","chefandy",[400,394,42,283,401,42],"42177728","aspenmayer",[403,394,42,283,398,42],"42178899",[405,394,42,283,401,42],"42178927",[407,394,42,283,398,42],"42180239",[409,394,42,283,410,42],"42176092","nkristoffersen",[412,394,42,287,413,42],"42176183","treprinum",[415,394,42,287,416,42],"42176214","avereveard",[418,394,18,224,237,42],"42176243",[122,420,18,224,123,42],20,[422,420,18,224,237,42],"42176512",[424,425,18,426,427,42],"42176406",21,"project\u002Fneutral\u002Fdocs","adamgordonbell",[429,425,18,224,237,42],"42176656",[431,425,18,426,427,42],"42177343",[148,433,47,434,149,42],22,"project\u002Fpositive\u002Fperformance",[436,437,18,438,439,42],"42175707",23,"project\u002Fneutral\u002Fnone","stephantul",[441,437,42,442,11,34],"42175735","maker\u002Fneutral\u002Fdocs",[444,437,42,445,446,42],"42175783","commenter\u002Fneutral\u002Fnone","dvaun",[448,437,42,445,439,42],"42177347",[164,450,18,373,165,42],24,[452,453,18,224,454,42],"42175987",25,"handfuloflight",[456,453,47,457,237,42],"42176068","project\u002Fpositive\u002Fidea",[118,459,34,334,119,42],26,[137,461,18,438,138,42],27,[463,461,42,438,11,34],"42178112",[465,461,18,310,138,42],"42178133",[467,461,42,310,11,34],"42178233",[469,461,18,310,470,42],"42180457","onel",[472,473,42,287,474,42],"42175895",28,"vasa_",[476,477,18,438,478,42],"42177141",29,"barlimp",[480,477,47,481,237,42],"42177188","project\u002Fpositive\u002Fnone",[171,207,18,438,172,42],[484,207,18,438,237,42],"42175313",[486,207,18,426,487,42],"42177308","youngNed",[489,207,42,442,11,34],"42178014",[491,207,18,438,492,42],"42180125","antman",[494,33,18,224,495,42],"42185304","ravishar313",[497,498,18,224,499,42],"42187439",32,"splitrocket",[501,502,42,283,503,42],"42188394",33,"omgCPhuture",[505,506,42,287,507,42],"42175160",34,"dartos",[509,506,42,212,237,42],"42175255",[511,506,42,224,11,34],"42175322",[513,506,47,457,507,42],"42175455",[515,506,42,445,516,42],"42175811","IanCal",[518,506,42,287,398,42],"42175998",[520,506,42,445,507,42],"42176743",[522,506,42,283,523,42],"42184293","mdp2021",[525,506,42,445,526,42],"42186981","mehh",[528,529,42,283,530,42],"42180810",35,"pgt",[532,529,42,283,533,42],"42182102","Loic",[107,535,34,359,108,42],36,[537,535,42,283,538,42],"42177570","azinman2",[540,535,42,445,108,42],"42183016",[542,535,18,442,237,42],"42177640",[544,545,34,546,547,42],"42177807",37,"project\u002Fcritical\u002Fnone","bionhoward",[183,545,42,438,11,34],[550,553],{"id":551,"t":552,"points":18,"comments":18},"42022517",1730503169,{"id":4,"t":12,"points":14,"comments":15},[555,560,565,570,574,579],{"id":556,"name":557,"points":558,"category":559},"42156977","lanparty.house",1117,"consumer_social",{"id":561,"name":562,"points":563,"category":564},"42185668","Black Hole Vision",316,"creative",{"id":566,"name":567,"points":568,"category":569},"42173119","Tips.io",291,"dev_tools",{"id":571,"name":572,"points":573,"category":569},"42197754","Trailimap",279,{"id":575,"name":576,"points":577,"category":578},"42149044","closing.wtf",273,"finance_crypto",{"id":580,"name":581,"points":582,"category":583},"42168166","Nova",271,"languages",false,"v2",{"id":587,"name":5,"line":6,"ai":584,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":30,"top":18,"kw":588,"maker":11,"domain":10,"tier":19,"attempts":40,"strip":589,"url":9,"mix":590,"praise":91,"objection":104,"sub":21},42174829,[24,25,26,27,28],"0305000050350301013230003353333433050011",[32,33,34,35],1794036664518]