[{"data":1,"prerenderedAt":437},["ShallowReactive",2],{"project-49435675":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":47,"post_text":49,"what":50,"said":101,"maker_in_thread":217,"shape":228,"grid":233,"attempts":400,"same_week":401,"maker_shows":45,"maker_hidden":430,"reader_version":431,"card":432},"49435675","CarWatch","A Raspberry with Qwen my local car AI","title","Show HN: I made a Raspberry with Qwen my local car AI","https:\u002F\u002Fgithub.com\u002FThinkOffApp\u002FCarWatch","github.com","petruspennanen",1787671220,2026,146,68,446,4366,11,"notable","ai","in-car voice assistant","other",[24,25,26,27,28],"local llm","obd-ii","raspberry pi","car manual rag","offline ai","curious",-0.29,{"positive":32,"neutral":33,"mixed":34,"critical":35},3,17,0,14,34,{"performance":38,"docs":40,"idea":41,"pricing":43,"execution":44,"design":46},{"positive":34,"critical":39},4,{"positive":34,"critical":39},{"positive":42,"critical":32},2,{"positive":34,"critical":42},{"positive":34,"critical":45},1,{"positive":34,"critical":34},[45,32,32,34,34,34,34,34,45,34,34,45,45,45,34,45,34,32,45,45,34,32,45,32,48,34,45,32,32,32,45,34,32,48,45,32,34,48,32,32],5,"Found that you can actually run a 35B Qwen model on a Pi with very impressive intelligence and stability. Built connectors for car ODB to read all about car internals, and manufacturer's cloud service for stuff like changing AC or opening\u002F locking doors. Gave it info such as the full car manual. And then hooked it up with my other agents in our discussion room!\n\nSo now it can answer car questions such as \"when should I add oil and what kind of oil?\" and help you fully offline, and when online talk with the agent family that includes all the most powerful models so they know how the car is, pla",{"problem":51,"audience":60,"pricing":61,"motivation":67,"time_to_build":60,"stack":71,"openness":98,"stage":99,"team_size":100},{"value":52,"source":53,"evidence":54},"Answers car questions and reads vehicle data offline, and controls car functions via manufacturer cloud","maker",[55,57],{"id":56},"post",{"id":58,"by":11,"text":59},"49461147","Yes Qwen runs local. Manufacturer cloud workw via existing integrations (they have many) in Home Assistant. The Pi makes a secure tunnel to my home assistant running on a Mac…",null,{"value":62,"source":53,"evidence":63},"Carwatch is open source and free",[64],{"id":65,"by":11,"text":66},"49456188","You can chat with it offline, it serves a dashboard and a chat interface. Carwatch is all open source and free. Groupmind.one is a free service where you can have a private room…",{"value":68,"source":53,"evidence":69},"Give the car a local AI agent that works with the maker's other agents",[70],{"id":56},[72,82,86,90],{"value":73,"source":53,"evidence":74},"Qwen3.6-35B-A3B (UD-Q3_K_S quant)",[75,79],{"id":76,"by":77,"text":78},"49445735","embedding-shape","Even so, in this case, author is using UD-Q3_K_S dynamic weights for Qwen3.6-35B-A3B, it will be dumb. Even the BF16 weights do stupid stuff like missing to confirm all…",{"id":80,"by":11,"text":81},"49455859","I've been very impressed by it's intelligence and lack of hallucinations. The dynamic Q3 is a good balance between accuracy and size keeping the 35B just below 16 GB. It is not…",{"value":83,"source":53,"evidence":84},"Raspberry Pi",[85],{"id":56},{"value":87,"source":53,"evidence":88},"Home Assistant",[89],{"id":58,"by":11,"text":59},{"value":91,"source":92,"evidence":93},"Whisper (speech-to-text)","commenter",[94],{"id":95,"by":96,"text":97},"49439595","0xbadcafebee","Qwen 4B and 9B should be faster and better reasoning than Gemma 4 E4B. Other good options for that much RAM are Gemma 4 12b and 31b. Gemma 4 E4B would be better for native audio,…","open_source","launched","solo",{"praise":102,"objections":111,"requests":156,"questions":174,"alternatives":195},[103,107],{"value":104,"source":53,"evidence":105},"Impressive local model performance on a Pi",[106],{"id":56},{"value":108,"source":53,"evidence":109},"Maker reports low hallucination rate with RAG on manual",[110],{"id":80,"by":11,"text":81},[112,119,130,141],{"value":113,"source":92,"evidence":114},"Model choice too slow and too heavy for Pi memory bandwidth",[115],{"id":116,"by":117,"text":118},"49440191","dofm","The issue is not RAM size. It's memory bandwidth! The 12B and 31B models will be useless on a Pi 5; maybe the 12B can be persuaded to run, but it may not manage more than one…",{"value":120,"source":92,"evidence":121},"Unclear use cases beyond what buttons already do",[122,126],{"id":123,"by":124,"text":125},"49445375","fedeb95","I think you should provide better use case scenarios that make clear why this project is useful. I can already change AC and lock doors with two buttons, why it should be better…",{"id":127,"by":128,"text":129},"49446402","sshussain270","At one point, every machine on earth is going to be its own agent, asking and answering questions and expressing concerns. Its like a good example of that, but I think the…",{"value":131,"source":92,"evidence":132},"Local models unreliable on specific car details",[133,137],{"id":134,"by":135,"text":136},"49440155","ehnto","Fair warning, I have found local models and frontier models to be very bad at the specifics when it comes to cars. Small differences like month and year model can impact oil…",{"id":138,"by":139,"text":140},"49445803","ssl-3","It makes stuff up about my car all the time, even when provided with factory and\u002For aftermarket service manuals to peruse in a custom GPT. That car is a ridiculously-common Honda…",{"value":142,"source":92,"evidence":143},"Readme includes unverified claims and reads as AI-written",[144,148,152],{"id":145,"by":146,"text":147},"49439897","Reubachi","I'm currently integrating one for simple diagnostic readouts to integrated small touchscreen. The main goal being a retro looking display that mainly dumps ODB data\u002Fmonitors, but…",{"id":149,"by":150,"text":151},"49440022","VTimofeenko","From readme: \"\"\" Unverified against the real car — it will confirm or refute itself on the next drive \"\"\" Looks like a ton of this is aspirational. For your project -- do you…",{"id":153,"by":154,"text":155},"49439907","tessierashpool","I scanned this README looking for the part written by a human and gave up when I realized there wasn't one. the Mermaid diagram doesn't even render.",[157,163,167],{"value":158,"source":53,"evidence":159},"Clearer use cases with status indicators",[160],{"id":161,"by":11,"text":162},"49460933","Thanks point taken, will add clear use cases with proven \u002F in testing \u002F planned status indicators!",{"value":164,"source":53,"evidence":165},"Support for more car brands via cloud integrations",[166],{"id":58,"by":11,"text":59},{"value":168,"source":92,"evidence":169},"Try a smaller, faster reasoning model",[170,171],{"id":116,"by":117,"text":118},{"id":172,"by":11,"text":173},"49469680","Hmm could try Ornith - wouldn't mind if answers came a bit faster. Probably hard to reach Qwens intelligence level though.",[175,185],{"value":176,"source":92,"evidence":177},"How does it handle trim\u002Fyear ambiguity and show sources?",[178,182],{"id":179,"by":180,"text":181},"49443031","jiangriver66","I’d be most interested in how it handles uncertainty when the answer depends on trim\u002Fyear or service history. For a car assistant, a useful default might be “I found the relevant…",{"id":183,"by":11,"text":184},"49456296","It refers to the manual page, that's really all you can do in that case especially offline.",{"value":186,"source":92,"evidence":187},"What hardware is needed for OBD data?",[188,192],{"id":189,"by":190,"text":191},"49439639","yjftsjthsd-h","What's the hardware situation to get OBD data?",{"id":193,"by":11,"text":194},"49456569","Any standard OBD-II ELM32 will work, they are like $15.. Mine is a Vgate iCar Pro over bluetooth. A USB ELM327 v1.5 also works",[196,203,210],{"value":197,"source":92,"evidence":198},"Freematics One+ for OBD and GPS",[199],{"id":200,"by":201,"text":202},"49443835","rpcope1","I have no idea on OP, but as far as a waypointer, have been in the area of building bespoke automotive telemetry solutions for a while now (including reversing some existing…",{"value":204,"source":92,"evidence":205},"Perplexity for fact-checking",[206],{"id":207,"by":208,"text":209},"49446709","ygouzerh","For prompts that needs fact-checking, I like these days to use Perplexity directly instead these days. It's way faster than the default websearch tool + give a link to the…",{"value":211,"source":92,"evidence":212},"Pitstop app",[213],{"id":214,"by":215,"text":216},"49440918","Robdel12","I will shamelessly plug https:\u002F\u002Fpitstop.app. I've been working on this for years and I work on\u002Frace cars regularly. I wouldn't fully trust LLMs with this :p BUT this project is…",{"comments":218,"answered":219,"key":220},18,9,[221,224,225],{"id":222,"text":223},"49455924","The setup here would be that your Sol would talk with CarWatch asking about the state of different car parts, service indicators and CarWatch would give a prioritized replacement…",{"id":80,"text":81},{"id":226,"text":227},"49460910","Hmm we must be doing something different, I think the official accuracy measurement for the Q3 is more than 90%. Maybe some bug affecting your test?",{"commenters":229,"max_depth":230,"maker_replies":218,"first_comment_mins":231,"lifespan_hours":232},40,6,246,76,[234,236,240,243,246,250,254,256,258,260,261,263,266,268,272,273,275,278,280,284,288,290,292,295,298,301,303,304,307,309,310,313,316,318,320,321,323,324,325,326,329,331,333,336,338,340,342,344,346,350,353,355,358,361,364,366,368,372,374,378,379,382,383,388,391,396,398,399],[134,34,45,235,135,34],"project\u002Fcritical\u002Fperformance",[237,34,32,238,239,34],"49440333","project\u002Fneutral\u002Fdocs","javier123454321",[241,34,32,238,242,34],"49442587","efskap",[244,34,32,238,245,34],"49444342","HPsquared",[247,34,32,248,249,34],"49450189","project\u002Fneutral\u002Fidea","Kim_Bruning",[251,34,34,252,253,34],"49442675","other\u002Fneutral\u002Fnone","whalesalad",[255,34,34,252,139,34],"49443759",[257,34,34,252,253,34],"49444973",[138,34,34,259,139,34],"other\u002Fcritical\u002Fperformance",[207,34,34,252,208,34],[262,34,34,252,245,34],"49444354",[264,34,34,252,265,34],"49444693","King-Aaron",[222,34,34,267,11,45],"project\u002Fpositive\u002Fidea",[269,34,32,270,271,34],"49442892","project\u002Fneutral\u002Fperformance","alexandra_au",[76,34,45,235,77,34],[80,34,34,274,11,45],"project\u002Fpositive\u002Fperformance",[276,34,34,277,77,34],"49455955","commenter\u002Fcritical\u002Fperformance",[226,34,34,279,11,45],"commenter\u002Fneutral\u002Fperformance",[281,45,45,282,283,34],"49443351","project\u002Fcritical\u002Fpricing","nacs",[285,45,34,286,287,34],"49443845","commenter\u002Fneutral\u002Fnone","tzs",[65,45,34,289,11,45],"project\u002Fpositive\u002Fpricing",[153,42,45,291,154,34],"project\u002Fcritical\u002Fdocs",[293,42,45,291,294,34],"49440059","almogo",[296,42,34,252,297,34],"49440821","thoughtbefore",[299,42,34,300,11,45],"49483247","project\u002Fpositive\u002Fdocs",[145,32,45,302,146,34],"project\u002Fcritical\u002Fexecution",[149,32,45,291,150,34],[305,32,34,306,11,45],"49456410","project\u002Fpositive\u002Fexecution",[200,32,32,308,201,34],"project\u002Fneutral\u002Fexecution",[58,32,34,306,11,45],[311,39,45,282,312,34],"49443058","walrus01",[314,39,34,252,315,34],"49443213","blululu",[317,39,45,235,312,34],"49443385",[123,48,45,319,124,34],"project\u002Fcritical\u002Fidea",[161,48,34,248,11,45],[322,230,32,270,117,34],"49439307",[95,230,32,270,96,34],[116,230,45,235,117,34],[172,230,34,270,11,45],[327,230,32,270,328,34],"49441210","aamargulies",[330,230,34,274,11,45],"49456437",[214,332,48,267,215,34],7,[334,332,34,252,335,34],"49442337","apinstein",[337,332,34,252,215,34],"49467378",[127,339,45,319,128,34],8,[341,339,32,248,208,34],"49446641",[343,339,34,248,11,45],"49460975",[345,219,34,306,11,45],"49483230",[347,348,32,349,201,34],"49443926",10,"project\u002Fneutral\u002Fnone",[351,348,32,349,352,34],"49444115","meatmanek",[354,348,34,267,11,45],"49461048",[356,18,45,319,357,34],"49440030","hypfer",[359,18,34,252,360,34],"49445352","voidUpdate",[362,18,34,252,363,34],"49462862","anshorei",[365,18,34,252,360,34],"49463177",[367,18,32,248,208,34],"49446697",[369,370,48,267,371,34],"49443177",12,"dowonseo",[373,370,34,300,11,45],"49461052",[375,376,45,291,377,34],"49446150",13,"amelius",[189,35,32,308,190,34],[380,35,32,308,381,34],"49439761","Alive-in-2025",[193,35,34,308,11,45],[384,385,48,386,387,34],"49442586",15,"project\u002Fpositive\u002Fnone","ellis0n",[389,385,34,390,11,45],"49456216","maker\u002Fpositive\u002Fnone",[392,393,32,394,395,34],"49444207",16,"project\u002Fneutral\u002Fdesign","bitwize",[397,393,34,349,11,45],"49461067",[179,33,32,248,180,34],[183,33,34,238,11,45],[],[402,407,412,416,421,425],{"id":403,"name":404,"points":405,"category":406},"49373456","simedw.com",598,"creative",{"id":408,"name":409,"points":410,"category":411},"49378768","Huzzah",384,"dev_tools",{"id":413,"name":414,"points":415,"category":406},"49419351","rack-02",240,{"id":417,"name":418,"points":419,"category":420},"49437049","LatticeDB",190,"data",{"id":422,"name":423,"points":424,"category":420},"49421806","PicoMQ",159,{"id":426,"name":427,"points":428,"category":429},"49410814","Thoreau BASIC",156,"languages",false,"v2",{"id":433,"name":5,"line":6,"ai":430,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":30,"top":18,"kw":434,"maker":11,"domain":10,"tier":19,"attempts":34,"strip":435,"url":9,"mix":436,"praise":104,"objection":113,"sub":21},49435675,[24,25,26,27,28],"1330000010011101031103135013331035130533",[32,33,34,35],1793991164750]