[{"data":1,"prerenderedAt":268},["ShallowReactive",2],{"project-46641362":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":44,"what":45,"said":88,"maker_in_thread":159,"shape":170,"grid":175,"attempts":230,"same_week":231,"maker_shows":161,"maker_hidden":261,"reader_version":262,"card":263},"46641362","Gambit","An open-source agent harness for building reliable AI agents","title","Show HN: Gambit, an open-source agent harness for building reliable AI agents","https:\u002F\u002Fgithub.com\u002Fbolt-foundry\u002Fgambit","github.com","randall",1768522405,2026,91,27,752,4366,18,"notable","ai","agent framework","library",[24,25,26,27,28],"ai agents","agent harness","llm orchestration","typescript agents","agent evals","loved",0.2,{"positive":32,"neutral":33,"mixed":34,"critical":34},5,3,1,10,{"docs":37,"idea":39,"privacy":41,"design":42},{"positive":38,"critical":38},0,{"positive":40,"critical":38},2,{"positive":38,"critical":34},{"positive":34,"critical":38},[33,38,33,38,38,38,38,33,38,34,38,32,38,33,38,38,32,38,32,38,32,32,38,38],"Hey HN!\n\nWanted to show our open source agent harness called Gambit.\n\nIf you’re not familiar, agent harnesses are sort of like an operating system for an agent... they handle tool calling, planning, context window management, and don’t require as much developer orchestration.\n\nNormally you might see an agent orchestration framework pipeline like:\n\ncompute -> compute -> compute -> LLM -> compute -> compute -> LLM\n\nwe invert this so with an agent harness, it’s more like:\n\nLLM -> LLM -> LLM -> compute -> LLM -> LLM -> compute -> LLM\n\nEssentially you describe each agent in either a self contained ",{"problem":46,"audience":52,"pricing":57,"motivation":58,"time_to_build":57,"stack":62,"openness":85,"stage":86,"team_size":87},{"value":47,"source":48,"evidence":49},"Building reliable AI agents with less developer orchestration, handling tool calling, planning and context management","maker",[50],{"id":51},"post",{"value":53,"source":54,"evidence":55},"Developers building AI agents","inferred",[56],{"id":51},null,{"value":59,"source":48,"evidence":60},"Prior LLM video editor results were unsatisfying, leading to work on inference-time LLM quality",[61],{"id":51},[63,67,73,81],{"value":64,"source":48,"evidence":65},"TypeScript",[66],{"id":51},{"value":68,"source":48,"evidence":69},"Zod schemas",[70],{"id":71,"by":11,"text":72},"46673251","yeah the way to do that stuff is through zod schemas… input and output schemas. you can set up really complex validation. thanks for checking it out!!",{"value":74,"source":75,"evidence":76},"OpenRouter","commenter",[77],{"id":78,"by":79,"text":80},"46659076","elgrantomate","also, it seems like this works with openrouter, and perhaps OpenAI -- what about Gemini API?",{"value":82,"source":75,"evidence":83},"OpenAI",[84],{"id":78,"by":79,"text":80},"open_source","beta","unknown",{"praise":89,"objections":107,"requests":119,"questions":128,"alternatives":147},[90,96],{"value":91,"source":75,"evidence":92},"Clean separation of logic, code and prompts",[93],{"id":94,"by":79,"text":95},"46658241","I've been playing with this for the past 24 hours or so. I like the atomic containment of the LLM, and the clear separation of logic, code, and prompts. You have some great…",{"value":97,"source":75,"evidence":98},"Good architecture and typed deck composition",[99,103],{"id":100,"by":101,"text":102},"46681251","niyikiza","Nice architecture. The typed deck composition pattern is exactly right for making agent workflows testable. One thing I've been thinking about is that schema validation catches…",{"id":104,"by":105,"text":106},"46641979","sofdao","this is awesome are things like file system baked in? fan of the design of the system. looks great architecturally",[108,112],{"value":109,"source":75,"evidence":110},"Parameter and prompt constraints are hard to get the model to follow",[111],{"id":94,"by":79,"text":95},{"value":113,"source":75,"evidence":114},"PII guarantee claim is overstated",[115],{"id":116,"by":117,"text":118},"46685068","yencabulator","> - Rubric based grading to guarantee you (for instance) don’t leak PII accidentally That does not sound like a \"guarantee\", at all.",[120,124],{"value":121,"source":75,"evidence":122},"Gemini API support",[123],{"id":78,"by":79,"text":80},{"value":125,"source":75,"evidence":126},"File loading as input",[127],{"id":94,"by":79,"text":95},[129,136,143],{"value":130,"source":75,"evidence":131},"How context is handled when agents call agents",[132],{"id":133,"by":134,"text":135},"46656955","benban","nice work. the idea of breaking agents into short-lived executors with explicit inputs\u002Foutputs makes a lot of sense - most failures i've seen come from agents staying alive too…",{"value":137,"source":75,"evidence":138},"Comparison to Mastra",[139],{"id":140,"by":141,"text":142},"46642428","Trufa","Is this an alternative to https:\u002F\u002Fmastra.ai\u002Fdocs How would it compare?",{"value":144,"source":75,"evidence":145},"Integration with cryptographic warrant authorization",[146],{"id":100,"by":101,"text":102},[148,152],{"value":149,"source":75,"evidence":150},"Mastra",[151],{"id":140,"by":141,"text":142},{"value":153,"source":75,"evidence":154},"LangChain",[155],{"id":156,"by":157,"text":158},"46641679","franciscomello","This looks quite interesting in terms of the architecture. Seems like a fresh take on stuff like Langchain, which at least last time I checked sucks.",{"comments":160,"answered":161,"key":162},11,7,[163,164,167],{"id":71,"text":72},{"id":165,"text":166},"46683016","So two things. 1\u002F crypto signing is totally the right way to think about this. 2\u002F I'm limiting prompt injection by using chain of command:…",{"id":168,"text":169},"46685937","yaaaaa exactly. You're totally on the same wavelength as me. Let's be friends lol",{"commenters":171,"max_depth":172,"maker_replies":160,"first_comment_mins":173,"lifespan_hours":174},12,4,41,95,[176,178,180,182,183,186,187,191,192,194,196,199,201,204,206,209,212,213,215,217,219,223,226,228],[94,38,33,177,79,38],"project\u002Fmixed\u002Fdocs",[71,38,38,179,11,34],"maker\u002Fneutral\u002Fdocs",[100,34,33,181,101,38],"project\u002Fneutral\u002Fidea",[165,34,38,181,11,34],[184,34,38,185,101,38],"46685256","commenter\u002Fneutral\u002Fnone",[168,34,38,185,11,34],[188,40,38,189,190,38],"46648780","other\u002Fneutral\u002Fnone","iainctduncan",[140,33,33,181,141,38],[193,33,38,181,11,34],"46642459",[116,172,34,195,117,38],"project\u002Fcritical\u002Fprivacy",[197,172,38,198,11,34],"46685929","project\u002Fneutral\u002Fprivacy",[133,32,32,200,134,38],"project\u002Fpositive\u002Fidea",[202,32,38,203,11,34],"46673261","project\u002Fneutral\u002Fnone",[78,205,33,203,79,38],6,[207,205,38,208,11,34],"46685944","maker\u002Fpositive\u002Fnone",[210,161,38,189,211,38],"46642393","tomhow",[156,161,32,200,157,38],[214,161,38,208,11,34],"46641999",[104,161,32,216,105,38],"project\u002Fpositive\u002Fdesign",[218,161,38,203,11,34],"46641997",[220,161,32,221,222,38],"46642089","project\u002Fpositive\u002Fnone","alberson",[224,161,32,221,225,38],"46642099","pych",[227,161,38,208,11,34],"46642183",[229,161,38,208,11,34],"46642431",[],[232,237,241,246,251,256],{"id":233,"name":234,"points":235,"category":236},"46706528","ChartGPU",670,"data",{"id":238,"name":239,"points":240,"category":20},"46713106","Sweep Next-Edit 1.5B",534,{"id":242,"name":243,"points":244,"category":245},"46673809","Calquio",314,"finance_crypto",{"id":247,"name":248,"points":249,"category":250},"46696283","Coi",225,"languages",{"id":252,"name":253,"points":254,"category":255},"46709543","Rails UI",208,"dev_tools",{"id":257,"name":258,"points":259,"category":260},"46671952","Dock",177,"productivity",false,"v2",{"id":264,"name":5,"line":6,"ai":261,"cat":20,"year":13,"points":14,"comments":15,"rec":29,"s":30,"top":18,"kw":265,"maker":11,"domain":10,"tier":19,"attempts":38,"strip":266,"url":9,"mix":267,"praise":91,"objection":109,"sub":21},46641362,[24,25,26,27,28],"303000030105030050505500",[32,33,34,34],1793991140645]