[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-cuopt-model-mapper":3,"mdc-a6le1m-key":40,"related-repo-nvidia-cuopt-model-mapper":1158,"related-org-nvidia-cuopt-model-mapper":1247},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":20,"repoUrl":21,"updatedAt":22,"license":23,"forks":24,"topics":25,"repo":35,"sourceUrl":38,"mdContent":39},"cuopt-model-mapper","map optimization problems to cuOpt models","Map interpreted optimization problems into cuOpt-native models for the fast path with minimal clarifying questions.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19],{"name":13,"slug":14,"type":15},"Optimization","optimization","tag",{"name":17,"slug":18,"type":15},"Data Modeling","data-modeling",{"name":9,"slug":8,"type":15},462,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fcuopt-examples","2026-07-30T05:29:19.401656","Apache-2.0",81,[26,27,28,29,30,14,31,32,33,34],"combinatorial-optimization","gpu","linear-programming","mixed-integer-programming","operations-research","optimization-algorithms","route-optimization","traveling-salesman-problem","vehicle-routing-problem",{"repoUrl":21,"stars":20,"forks":24,"topics":36,"description":37},[26,27,28,29,30,14,31,32,33,34],"NVIDIA cuOpt examples for decision optimization","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fcuopt-examples\u002Ftree\u002FHEAD\u002Fcuopt_on_nemoclaw\u002Fopenclaw-skills\u002Fcuopt-model-mapper","---\nname: cuopt-model-mapper\nversion: \"26.06.01\"\ndescription: Map interpreted optimization problems into cuOpt-native models for the fast path with minimal clarifying questions.\nlicense: Apache-2.0\nmetadata:\n  author: NVIDIA cuOpt Team\n  tags:\n    - cuopt\n    - nemoclaw\n    - orchestration\norigin: skill-evolution\n---\n\n# cuOpt Model Mapper\n\nUse this skill in the **fast path** after the request has already been identified as an optimization problem and the data has been interpreted enough to support model construction.\n\nThis skill takes the working interpretation of the problem and maps it directly into cuOpt-native model objects.\n\n## Purpose\n\nThe fast path should avoid unnecessary architecture.\n\nThis skill exists to:\n- build directly in cuOpt\n- avoid a heavyweight intermediate model\n- ask only the smallest number of blocking questions\n- solve and return the result quickly\n\nThis skill is **not** for replayable\u002Fauditable artifact design. That belongs to a different path.\n\n## Preconditions\n\nUse this skill only when the following are already mostly clear:\n- the request is genuinely an optimization problem\n- the problem family is identified as LP \u002F MILP \u002F QP \u002F routing\n- the uploaded data has been inspected enough to infer likely roles\n- only a small number of unresolved modeling questions remain\n\nIf those conditions are not met, first use:\n- `optimization-intent-router`\n- `optimization-mode-router` if execution mode may matter for replayability, audit, export, or reuse\n- `tabular-optimization-ingestion`\n\n## Core rule\n\nFor the fast path, map directly from the interpreted data into cuOpt structures.\nDo not introduce a replayable intermediate artifact unless the user asks for replayability, auditability, export, or reuse.\n\n**In NemoClaw sandbox:** before building the cuOpt model, confirm\n`cuopt-sandbox` gates completed (probe → select gRPC capability → smoke test).\nDo not build a parallel heuristic assigner \"first\" — cuOpt is the first\nand only solver for assignments\u002Fschedules.\n\n## Workflow\n\n### 1. Confirm the minimum viable formulation\nBefore building, confirm internally:\n- what the decisions are\n- what the objective is\n- what the hard constraints are\n- whether integrality is required\n- whether the problem is numerical optimization or routing\n\nUse the unresolved blocker list from ingestion as the starting point; do not reopen broad exploratory questioning unless the current interpretation is clearly inconsistent.\n\nIf one non-retrievable modeling choice would change the meaning of the solve, ask exactly one concise blocking question.\n\n**Scheduling with no stated objective:** Feasibility (no double-booking,\nrespect unavailability, assign every item) belongs in **hard constraints**.\nDo not deliver a greedy feasible schedule first. Ask one objective\nquestion if needed, or state a default secondary objective (e.g. balance\nslot times, minimize penalty slacks), then solve with cuOpt.\n\nExamples:\n- **\"Do these production quantities need to be whole numbers?\"**\n- **\"Must all demand be met, or can unmet demand be allowed with a penalty?\"**\n- **\"Are these time windows mandatory, or just preferred?\"**\n\n### 2. Choose the cuOpt path\n\n#### If LP \u002F MILP\n- hand off through `numerical-optimization-formulation` for formulation discipline\n- use `cuopt-numerical-optimization-api-python` (or `cuopt-numerical-optimization-api-cli` for MPS inputs)\n- preserve direct mappings from source data into variables, coefficients, RHS values, bounds, and objective terms\n\n#### If QP\n- hand off through `numerical-optimization-formulation` for formulation discipline\n- use `cuopt-numerical-optimization-api-python`\n- preserve direct mappings from source data into variables, quadratic objective terms, RHS values, and bounds\n\n#### If routing\n- hand off through `routing-formulation`\n- use `cuopt-routing-api-python`\n- map source data directly into locations, demands, vehicles, travel metrics, time windows, and pickup-delivery relationships\n\n### 3. Preserve source-to-model traceability lightly\nEven in fast mode, keep enough working traceability to avoid confusion during the same interaction.\n\nAt minimum, be able to state:\n- which tables\u002Fcolumns drove the objective\n- which tables\u002Fcolumns drove major constraints\n- what each decision variable represents\n\nDo this without building a full replayable artifact.\n\n### 4. Solve with cuOpt\n- use cuOpt whenever available and appropriate for the identified problem family\n- verify solver status\n- verify that the result is sensible against the modeled constraints\n\n### 5. Explain results in user language\nReturn:\n- recommended decisions\n- objective value\n- what the objective means in business terms\n- any important bottlenecks or binding tradeoffs\n- key assumptions that materially affect the answer\n\n## Direct mapping guidance\n\n### Numerical optimization mapping\nTypical direct mappings include:\n- per-row profit\u002Fcost columns → objective coefficients\n- per-row resource consumption columns → constraint coefficients\n- capacity table values → RHS limits\n- min\u002Fmax fields → variable bounds or additional constraints\n- yes\u002Fno\u002Fopen-close decisions → binary variables\n- count decisions → integer variables\n\n### Routing mapping\nTypical direct mappings include:\n- customer\u002Flocation rows → stops\n- coordinate or matrix tables → travel metric inputs\n- demand fields → stop demand\n- vehicle table → fleet definition\n- depot fields → start\u002Fend nodes\n- time window columns → service time constraints\n- pickup\u002Fdelivery ids → paired stop relationships\n\n## Examples\n\n### Example 1: direct LP\u002FMILP mapping\nInterpreted problem:\n- one row per product\n- `profit` column is objective coefficient\n- `labor_hours` and `steel_units` are resource coefficients\n- a separate capacity table provides RHS limits\n\nFast-path mapping:\n- create one decision variable per product\n- maximize total profit\n- add one constraint per capacity\u002Fresource\n- if products must be whole units, use integer variables\n\n### Example 2: direct routing mapping\nInterpreted problem:\n- one row per customer stop\n- vehicle table defines fleet\n- travel-time matrix defines movement cost\n- customer demand and time-window fields are available\n\nFast-path mapping:\n- map stops, fleet, and travel data directly into cuOpt routing inputs\n- ask only the final blocker question if needed, such as whether time windows are hard constraints\n- solve and explain route recommendations\n\n### Example 3: do not overbuild\nInterpreted problem:\n- user wants a one-off answer from uploaded data\n- no replay, export, or audit requirement is mentioned\n\nCorrect behavior:\n- map directly into cuOpt\n- do not stop to build a replayable structured artifact\n- keep only enough traceability to explain the result clearly in the current interaction\n\n## Integrality rule\n\nDo not default to continuous variables when the decisions represent discrete units, counts, assignments, vehicles, workers, items, or yes\u002Fno choices.\n\nAsk if unclear, but if the problem statement strongly implies discreteness, model it as MILP rather than LP.\n\n## Fast-path guardrails\n\n- Do not build a heavyweight intermediate representation for one-off solves.\n- Do not skip a critical clarifying question if it changes the model meaning.\n- Do not ask a long questionnaire when one focused question will unblock the solve.\n- Do not use cuOpt for tasks that are really descriptive analytics.\n- Do not force routing when the problem is actually allocation without path construction.\n- Do not force QP when the objective is linear.\n\n## Result reporting requirements\n\nAlways report at least:\n- solver status\n- objective value\n- brief explanation of what the objective means\n- recommended decisions\n- any assumptions or caveats that materially affect the result\n\nIf relevant, also mention which constraint or resource appears to be most limiting.\n\n## Handoff guidance\n\n- For LP \u002F MILP:\n  - use `numerical-optimization-formulation`\n  - then use `cuopt-numerical-optimization-api-python` (or `cuopt-numerical-optimization-api-cli` for MPS inputs)\n  - follow `cuopt-user-rules`\n  - in sandbox contexts, follow `cuopt-sandbox` (gates + selected gRPC path)\n    before any LP\u002FMILP solve\n\n- For QP:\n  - use `numerical-optimization-formulation`\n  - then use `cuopt-numerical-optimization-api-python`\n  - follow `cuopt-user-rules`\n  - in sandbox contexts, follow `cuopt-sandbox` (gates + selected gRPC path)\n    before any QP solve\n\n- For routing:\n  - use `routing-formulation`\n  - then use `cuopt-routing-api-python`\n  - follow `cuopt-user-rules`\n  - in sandbox contexts, follow `cuopt-sandbox` (gates + REST)\n    before any routing solve\n\n## Success criterion\n\nThis skill succeeds when the agent can go from:\n- interpreted data\n- plus zero or a few blocking clarifications\n\nto:\n- direct cuOpt model construction\n- a validated solve\n- and a clear business-facing answer\n\nwithout introducing unnecessary replay\u002Faudit machinery.\n",{"data":41,"body":50},{"name":4,"version":42,"description":6,"license":23,"metadata":43,"origin":49},"26.06.01",{"author":44,"tags":45},"NVIDIA cuOpt Team",[46,47,48],"cuopt","nemoclaw","orchestration","skill-evolution",{"type":51,"children":52},"root",[53,61,75,80,87,92,97,122,134,140,145,168,173,206,212,217,235,241,248,253,281,286,291,308,313,340,346,353,395,401,428,434,462,468,473,478,496,501,507,525,531,536,564,570,576,581,614,620,624,662,668,674,679,722,727,750,756,760,783,787,805,811,815,828,833,851,857,862,867,873,906,912,917,943,948,954,1106,1112,1117,1130,1135,1153],{"type":54,"tag":55,"props":56,"children":57},"element","h1",{"id":4},[58],{"type":59,"value":60},"text","cuOpt Model Mapper",{"type":54,"tag":62,"props":63,"children":64},"p",{},[65,67,73],{"type":59,"value":66},"Use this skill in the ",{"type":54,"tag":68,"props":69,"children":70},"strong",{},[71],{"type":59,"value":72},"fast path",{"type":59,"value":74}," after the request has already been identified as an optimization problem and the data has been interpreted enough to support model construction.",{"type":54,"tag":62,"props":76,"children":77},{},[78],{"type":59,"value":79},"This skill takes the working interpretation of the problem and maps it directly into cuOpt-native model objects.",{"type":54,"tag":81,"props":82,"children":84},"h2",{"id":83},"purpose",[85],{"type":59,"value":86},"Purpose",{"type":54,"tag":62,"props":88,"children":89},{},[90],{"type":59,"value":91},"The fast path should avoid unnecessary architecture.",{"type":54,"tag":62,"props":93,"children":94},{},[95],{"type":59,"value":96},"This skill exists to:",{"type":54,"tag":98,"props":99,"children":100},"ul",{},[101,107,112,117],{"type":54,"tag":102,"props":103,"children":104},"li",{},[105],{"type":59,"value":106},"build directly in cuOpt",{"type":54,"tag":102,"props":108,"children":109},{},[110],{"type":59,"value":111},"avoid a heavyweight intermediate model",{"type":54,"tag":102,"props":113,"children":114},{},[115],{"type":59,"value":116},"ask only the smallest number of blocking questions",{"type":54,"tag":102,"props":118,"children":119},{},[120],{"type":59,"value":121},"solve and return the result quickly",{"type":54,"tag":62,"props":123,"children":124},{},[125,127,132],{"type":59,"value":126},"This skill is ",{"type":54,"tag":68,"props":128,"children":129},{},[130],{"type":59,"value":131},"not",{"type":59,"value":133}," for replayable\u002Fauditable artifact design. That belongs to a different path.",{"type":54,"tag":81,"props":135,"children":137},{"id":136},"preconditions",[138],{"type":59,"value":139},"Preconditions",{"type":54,"tag":62,"props":141,"children":142},{},[143],{"type":59,"value":144},"Use this skill only when the following are already mostly clear:",{"type":54,"tag":98,"props":146,"children":147},{},[148,153,158,163],{"type":54,"tag":102,"props":149,"children":150},{},[151],{"type":59,"value":152},"the request is genuinely an optimization problem",{"type":54,"tag":102,"props":154,"children":155},{},[156],{"type":59,"value":157},"the problem family is identified as LP \u002F MILP \u002F QP \u002F routing",{"type":54,"tag":102,"props":159,"children":160},{},[161],{"type":59,"value":162},"the uploaded data has been inspected enough to infer likely roles",{"type":54,"tag":102,"props":164,"children":165},{},[166],{"type":59,"value":167},"only a small number of unresolved modeling questions remain",{"type":54,"tag":62,"props":169,"children":170},{},[171],{"type":59,"value":172},"If those conditions are not met, first use:",{"type":54,"tag":98,"props":174,"children":175},{},[176,186,197],{"type":54,"tag":102,"props":177,"children":178},{},[179],{"type":54,"tag":180,"props":181,"children":183},"code",{"className":182},[],[184],{"type":59,"value":185},"optimization-intent-router",{"type":54,"tag":102,"props":187,"children":188},{},[189,195],{"type":54,"tag":180,"props":190,"children":192},{"className":191},[],[193],{"type":59,"value":194},"optimization-mode-router",{"type":59,"value":196}," if execution mode may matter for replayability, audit, export, or reuse",{"type":54,"tag":102,"props":198,"children":199},{},[200],{"type":54,"tag":180,"props":201,"children":203},{"className":202},[],[204],{"type":59,"value":205},"tabular-optimization-ingestion",{"type":54,"tag":81,"props":207,"children":209},{"id":208},"core-rule",[210],{"type":59,"value":211},"Core rule",{"type":54,"tag":62,"props":213,"children":214},{},[215],{"type":59,"value":216},"For the fast path, map directly from the interpreted data into cuOpt structures.\nDo not introduce a replayable intermediate artifact unless the user asks for replayability, auditability, export, or reuse.",{"type":54,"tag":62,"props":218,"children":219},{},[220,225,227,233],{"type":54,"tag":68,"props":221,"children":222},{},[223],{"type":59,"value":224},"In NemoClaw sandbox:",{"type":59,"value":226}," before building the cuOpt model, confirm\n",{"type":54,"tag":180,"props":228,"children":230},{"className":229},[],[231],{"type":59,"value":232},"cuopt-sandbox",{"type":59,"value":234}," gates completed (probe → select gRPC capability → smoke test).\nDo not build a parallel heuristic assigner \"first\" — cuOpt is the first\nand only solver for assignments\u002Fschedules.",{"type":54,"tag":81,"props":236,"children":238},{"id":237},"workflow",[239],{"type":59,"value":240},"Workflow",{"type":54,"tag":242,"props":243,"children":245},"h3",{"id":244},"_1-confirm-the-minimum-viable-formulation",[246],{"type":59,"value":247},"1. Confirm the minimum viable formulation",{"type":54,"tag":62,"props":249,"children":250},{},[251],{"type":59,"value":252},"Before building, confirm internally:",{"type":54,"tag":98,"props":254,"children":255},{},[256,261,266,271,276],{"type":54,"tag":102,"props":257,"children":258},{},[259],{"type":59,"value":260},"what the decisions are",{"type":54,"tag":102,"props":262,"children":263},{},[264],{"type":59,"value":265},"what the objective is",{"type":54,"tag":102,"props":267,"children":268},{},[269],{"type":59,"value":270},"what the hard constraints are",{"type":54,"tag":102,"props":272,"children":273},{},[274],{"type":59,"value":275},"whether integrality is required",{"type":54,"tag":102,"props":277,"children":278},{},[279],{"type":59,"value":280},"whether the problem is numerical optimization or routing",{"type":54,"tag":62,"props":282,"children":283},{},[284],{"type":59,"value":285},"Use the unresolved blocker list from ingestion as the starting point; do not reopen broad exploratory questioning unless the current interpretation is clearly inconsistent.",{"type":54,"tag":62,"props":287,"children":288},{},[289],{"type":59,"value":290},"If one non-retrievable modeling choice would change the meaning of the solve, ask exactly one concise blocking question.",{"type":54,"tag":62,"props":292,"children":293},{},[294,299,301,306],{"type":54,"tag":68,"props":295,"children":296},{},[297],{"type":59,"value":298},"Scheduling with no stated objective:",{"type":59,"value":300}," Feasibility (no double-booking,\nrespect unavailability, assign every item) belongs in ",{"type":54,"tag":68,"props":302,"children":303},{},[304],{"type":59,"value":305},"hard constraints",{"type":59,"value":307},".\nDo not deliver a greedy feasible schedule first. Ask one objective\nquestion if needed, or state a default secondary objective (e.g. balance\nslot times, minimize penalty slacks), then solve with cuOpt.",{"type":54,"tag":62,"props":309,"children":310},{},[311],{"type":59,"value":312},"Examples:",{"type":54,"tag":98,"props":314,"children":315},{},[316,324,332],{"type":54,"tag":102,"props":317,"children":318},{},[319],{"type":54,"tag":68,"props":320,"children":321},{},[322],{"type":59,"value":323},"\"Do these production quantities need to be whole numbers?\"",{"type":54,"tag":102,"props":325,"children":326},{},[327],{"type":54,"tag":68,"props":328,"children":329},{},[330],{"type":59,"value":331},"\"Must all demand be met, or can unmet demand be allowed with a penalty?\"",{"type":54,"tag":102,"props":333,"children":334},{},[335],{"type":54,"tag":68,"props":336,"children":337},{},[338],{"type":59,"value":339},"\"Are these time windows mandatory, or just preferred?\"",{"type":54,"tag":242,"props":341,"children":343},{"id":342},"_2-choose-the-cuopt-path",[344],{"type":59,"value":345},"2. Choose the cuOpt path",{"type":54,"tag":347,"props":348,"children":350},"h4",{"id":349},"if-lp-milp",[351],{"type":59,"value":352},"If LP \u002F MILP",{"type":54,"tag":98,"props":354,"children":355},{},[356,369,390],{"type":54,"tag":102,"props":357,"children":358},{},[359,361,367],{"type":59,"value":360},"hand off through ",{"type":54,"tag":180,"props":362,"children":364},{"className":363},[],[365],{"type":59,"value":366},"numerical-optimization-formulation",{"type":59,"value":368}," for formulation discipline",{"type":54,"tag":102,"props":370,"children":371},{},[372,374,380,382,388],{"type":59,"value":373},"use ",{"type":54,"tag":180,"props":375,"children":377},{"className":376},[],[378],{"type":59,"value":379},"cuopt-numerical-optimization-api-python",{"type":59,"value":381}," (or ",{"type":54,"tag":180,"props":383,"children":385},{"className":384},[],[386],{"type":59,"value":387},"cuopt-numerical-optimization-api-cli",{"type":59,"value":389}," for MPS inputs)",{"type":54,"tag":102,"props":391,"children":392},{},[393],{"type":59,"value":394},"preserve direct mappings from source data into variables, coefficients, RHS values, bounds, and objective terms",{"type":54,"tag":347,"props":396,"children":398},{"id":397},"if-qp",[399],{"type":59,"value":400},"If QP",{"type":54,"tag":98,"props":402,"children":403},{},[404,414,423],{"type":54,"tag":102,"props":405,"children":406},{},[407,408,413],{"type":59,"value":360},{"type":54,"tag":180,"props":409,"children":411},{"className":410},[],[412],{"type":59,"value":366},{"type":59,"value":368},{"type":54,"tag":102,"props":415,"children":416},{},[417,418],{"type":59,"value":373},{"type":54,"tag":180,"props":419,"children":421},{"className":420},[],[422],{"type":59,"value":379},{"type":54,"tag":102,"props":424,"children":425},{},[426],{"type":59,"value":427},"preserve direct mappings from source data into variables, quadratic objective terms, RHS values, and bounds",{"type":54,"tag":347,"props":429,"children":431},{"id":430},"if-routing",[432],{"type":59,"value":433},"If routing",{"type":54,"tag":98,"props":435,"children":436},{},[437,447,457],{"type":54,"tag":102,"props":438,"children":439},{},[440,441],{"type":59,"value":360},{"type":54,"tag":180,"props":442,"children":444},{"className":443},[],[445],{"type":59,"value":446},"routing-formulation",{"type":54,"tag":102,"props":448,"children":449},{},[450,451],{"type":59,"value":373},{"type":54,"tag":180,"props":452,"children":454},{"className":453},[],[455],{"type":59,"value":456},"cuopt-routing-api-python",{"type":54,"tag":102,"props":458,"children":459},{},[460],{"type":59,"value":461},"map source data directly into locations, demands, vehicles, travel metrics, time windows, and pickup-delivery relationships",{"type":54,"tag":242,"props":463,"children":465},{"id":464},"_3-preserve-source-to-model-traceability-lightly",[466],{"type":59,"value":467},"3. Preserve source-to-model traceability lightly",{"type":54,"tag":62,"props":469,"children":470},{},[471],{"type":59,"value":472},"Even in fast mode, keep enough working traceability to avoid confusion during the same interaction.",{"type":54,"tag":62,"props":474,"children":475},{},[476],{"type":59,"value":477},"At minimum, be able to state:",{"type":54,"tag":98,"props":479,"children":480},{},[481,486,491],{"type":54,"tag":102,"props":482,"children":483},{},[484],{"type":59,"value":485},"which tables\u002Fcolumns drove the objective",{"type":54,"tag":102,"props":487,"children":488},{},[489],{"type":59,"value":490},"which tables\u002Fcolumns drove major constraints",{"type":54,"tag":102,"props":492,"children":493},{},[494],{"type":59,"value":495},"what each decision variable represents",{"type":54,"tag":62,"props":497,"children":498},{},[499],{"type":59,"value":500},"Do this without building a full replayable artifact.",{"type":54,"tag":242,"props":502,"children":504},{"id":503},"_4-solve-with-cuopt",[505],{"type":59,"value":506},"4. Solve with cuOpt",{"type":54,"tag":98,"props":508,"children":509},{},[510,515,520],{"type":54,"tag":102,"props":511,"children":512},{},[513],{"type":59,"value":514},"use cuOpt whenever available and appropriate for the identified problem family",{"type":54,"tag":102,"props":516,"children":517},{},[518],{"type":59,"value":519},"verify solver status",{"type":54,"tag":102,"props":521,"children":522},{},[523],{"type":59,"value":524},"verify that the result is sensible against the modeled constraints",{"type":54,"tag":242,"props":526,"children":528},{"id":527},"_5-explain-results-in-user-language",[529],{"type":59,"value":530},"5. Explain results in user language",{"type":54,"tag":62,"props":532,"children":533},{},[534],{"type":59,"value":535},"Return:",{"type":54,"tag":98,"props":537,"children":538},{},[539,544,549,554,559],{"type":54,"tag":102,"props":540,"children":541},{},[542],{"type":59,"value":543},"recommended decisions",{"type":54,"tag":102,"props":545,"children":546},{},[547],{"type":59,"value":548},"objective value",{"type":54,"tag":102,"props":550,"children":551},{},[552],{"type":59,"value":553},"what the objective means in business terms",{"type":54,"tag":102,"props":555,"children":556},{},[557],{"type":59,"value":558},"any important bottlenecks or binding tradeoffs",{"type":54,"tag":102,"props":560,"children":561},{},[562],{"type":59,"value":563},"key assumptions that materially affect the answer",{"type":54,"tag":81,"props":565,"children":567},{"id":566},"direct-mapping-guidance",[568],{"type":59,"value":569},"Direct mapping guidance",{"type":54,"tag":242,"props":571,"children":573},{"id":572},"numerical-optimization-mapping",[574],{"type":59,"value":575},"Numerical optimization mapping",{"type":54,"tag":62,"props":577,"children":578},{},[579],{"type":59,"value":580},"Typical direct mappings include:",{"type":54,"tag":98,"props":582,"children":583},{},[584,589,594,599,604,609],{"type":54,"tag":102,"props":585,"children":586},{},[587],{"type":59,"value":588},"per-row profit\u002Fcost columns → objective coefficients",{"type":54,"tag":102,"props":590,"children":591},{},[592],{"type":59,"value":593},"per-row resource consumption columns → constraint coefficients",{"type":54,"tag":102,"props":595,"children":596},{},[597],{"type":59,"value":598},"capacity table values → RHS limits",{"type":54,"tag":102,"props":600,"children":601},{},[602],{"type":59,"value":603},"min\u002Fmax fields → variable bounds or additional constraints",{"type":54,"tag":102,"props":605,"children":606},{},[607],{"type":59,"value":608},"yes\u002Fno\u002Fopen-close decisions → binary variables",{"type":54,"tag":102,"props":610,"children":611},{},[612],{"type":59,"value":613},"count decisions → integer variables",{"type":54,"tag":242,"props":615,"children":617},{"id":616},"routing-mapping",[618],{"type":59,"value":619},"Routing mapping",{"type":54,"tag":62,"props":621,"children":622},{},[623],{"type":59,"value":580},{"type":54,"tag":98,"props":625,"children":626},{},[627,632,637,642,647,652,657],{"type":54,"tag":102,"props":628,"children":629},{},[630],{"type":59,"value":631},"customer\u002Flocation rows → stops",{"type":54,"tag":102,"props":633,"children":634},{},[635],{"type":59,"value":636},"coordinate or matrix tables → travel metric inputs",{"type":54,"tag":102,"props":638,"children":639},{},[640],{"type":59,"value":641},"demand fields → stop demand",{"type":54,"tag":102,"props":643,"children":644},{},[645],{"type":59,"value":646},"vehicle table → fleet definition",{"type":54,"tag":102,"props":648,"children":649},{},[650],{"type":59,"value":651},"depot fields → start\u002Fend nodes",{"type":54,"tag":102,"props":653,"children":654},{},[655],{"type":59,"value":656},"time window columns → service time constraints",{"type":54,"tag":102,"props":658,"children":659},{},[660],{"type":59,"value":661},"pickup\u002Fdelivery ids → paired stop relationships",{"type":54,"tag":81,"props":663,"children":665},{"id":664},"examples",[666],{"type":59,"value":667},"Examples",{"type":54,"tag":242,"props":669,"children":671},{"id":670},"example-1-direct-lpmilp-mapping",[672],{"type":59,"value":673},"Example 1: direct LP\u002FMILP mapping",{"type":54,"tag":62,"props":675,"children":676},{},[677],{"type":59,"value":678},"Interpreted problem:",{"type":54,"tag":98,"props":680,"children":681},{},[682,687,698,717],{"type":54,"tag":102,"props":683,"children":684},{},[685],{"type":59,"value":686},"one row per product",{"type":54,"tag":102,"props":688,"children":689},{},[690,696],{"type":54,"tag":180,"props":691,"children":693},{"className":692},[],[694],{"type":59,"value":695},"profit",{"type":59,"value":697}," column is objective coefficient",{"type":54,"tag":102,"props":699,"children":700},{},[701,707,709,715],{"type":54,"tag":180,"props":702,"children":704},{"className":703},[],[705],{"type":59,"value":706},"labor_hours",{"type":59,"value":708}," and ",{"type":54,"tag":180,"props":710,"children":712},{"className":711},[],[713],{"type":59,"value":714},"steel_units",{"type":59,"value":716}," are resource coefficients",{"type":54,"tag":102,"props":718,"children":719},{},[720],{"type":59,"value":721},"a separate capacity table provides RHS limits",{"type":54,"tag":62,"props":723,"children":724},{},[725],{"type":59,"value":726},"Fast-path mapping:",{"type":54,"tag":98,"props":728,"children":729},{},[730,735,740,745],{"type":54,"tag":102,"props":731,"children":732},{},[733],{"type":59,"value":734},"create one decision variable per product",{"type":54,"tag":102,"props":736,"children":737},{},[738],{"type":59,"value":739},"maximize total profit",{"type":54,"tag":102,"props":741,"children":742},{},[743],{"type":59,"value":744},"add one constraint per capacity\u002Fresource",{"type":54,"tag":102,"props":746,"children":747},{},[748],{"type":59,"value":749},"if products must be whole units, use integer variables",{"type":54,"tag":242,"props":751,"children":753},{"id":752},"example-2-direct-routing-mapping",[754],{"type":59,"value":755},"Example 2: direct routing mapping",{"type":54,"tag":62,"props":757,"children":758},{},[759],{"type":59,"value":678},{"type":54,"tag":98,"props":761,"children":762},{},[763,768,773,778],{"type":54,"tag":102,"props":764,"children":765},{},[766],{"type":59,"value":767},"one row per customer stop",{"type":54,"tag":102,"props":769,"children":770},{},[771],{"type":59,"value":772},"vehicle table defines fleet",{"type":54,"tag":102,"props":774,"children":775},{},[776],{"type":59,"value":777},"travel-time matrix defines movement cost",{"type":54,"tag":102,"props":779,"children":780},{},[781],{"type":59,"value":782},"customer demand and time-window fields are available",{"type":54,"tag":62,"props":784,"children":785},{},[786],{"type":59,"value":726},{"type":54,"tag":98,"props":788,"children":789},{},[790,795,800],{"type":54,"tag":102,"props":791,"children":792},{},[793],{"type":59,"value":794},"map stops, fleet, and travel data directly into cuOpt routing inputs",{"type":54,"tag":102,"props":796,"children":797},{},[798],{"type":59,"value":799},"ask only the final blocker question if needed, such as whether time windows are hard constraints",{"type":54,"tag":102,"props":801,"children":802},{},[803],{"type":59,"value":804},"solve and explain route recommendations",{"type":54,"tag":242,"props":806,"children":808},{"id":807},"example-3-do-not-overbuild",[809],{"type":59,"value":810},"Example 3: do not overbuild",{"type":54,"tag":62,"props":812,"children":813},{},[814],{"type":59,"value":678},{"type":54,"tag":98,"props":816,"children":817},{},[818,823],{"type":54,"tag":102,"props":819,"children":820},{},[821],{"type":59,"value":822},"user wants a one-off answer from uploaded data",{"type":54,"tag":102,"props":824,"children":825},{},[826],{"type":59,"value":827},"no replay, export, or audit requirement is mentioned",{"type":54,"tag":62,"props":829,"children":830},{},[831],{"type":59,"value":832},"Correct behavior:",{"type":54,"tag":98,"props":834,"children":835},{},[836,841,846],{"type":54,"tag":102,"props":837,"children":838},{},[839],{"type":59,"value":840},"map directly into cuOpt",{"type":54,"tag":102,"props":842,"children":843},{},[844],{"type":59,"value":845},"do not stop to build a replayable structured artifact",{"type":54,"tag":102,"props":847,"children":848},{},[849],{"type":59,"value":850},"keep only enough traceability to explain the result clearly in the current interaction",{"type":54,"tag":81,"props":852,"children":854},{"id":853},"integrality-rule",[855],{"type":59,"value":856},"Integrality rule",{"type":54,"tag":62,"props":858,"children":859},{},[860],{"type":59,"value":861},"Do not default to continuous variables when the decisions represent discrete units, counts, assignments, vehicles, workers, items, or yes\u002Fno choices.",{"type":54,"tag":62,"props":863,"children":864},{},[865],{"type":59,"value":866},"Ask if unclear, but if the problem statement strongly implies discreteness, model it as MILP rather than LP.",{"type":54,"tag":81,"props":868,"children":870},{"id":869},"fast-path-guardrails",[871],{"type":59,"value":872},"Fast-path guardrails",{"type":54,"tag":98,"props":874,"children":875},{},[876,881,886,891,896,901],{"type":54,"tag":102,"props":877,"children":878},{},[879],{"type":59,"value":880},"Do not build a heavyweight intermediate representation for one-off solves.",{"type":54,"tag":102,"props":882,"children":883},{},[884],{"type":59,"value":885},"Do not skip a critical clarifying question if it changes the model meaning.",{"type":54,"tag":102,"props":887,"children":888},{},[889],{"type":59,"value":890},"Do not ask a long questionnaire when one focused question will unblock the solve.",{"type":54,"tag":102,"props":892,"children":893},{},[894],{"type":59,"value":895},"Do not use cuOpt for tasks that are really descriptive analytics.",{"type":54,"tag":102,"props":897,"children":898},{},[899],{"type":59,"value":900},"Do not force routing when the problem is actually allocation without path construction.",{"type":54,"tag":102,"props":902,"children":903},{},[904],{"type":59,"value":905},"Do not force QP when the objective is linear.",{"type":54,"tag":81,"props":907,"children":909},{"id":908},"result-reporting-requirements",[910],{"type":59,"value":911},"Result reporting requirements",{"type":54,"tag":62,"props":913,"children":914},{},[915],{"type":59,"value":916},"Always report at least:",{"type":54,"tag":98,"props":918,"children":919},{},[920,925,929,934,938],{"type":54,"tag":102,"props":921,"children":922},{},[923],{"type":59,"value":924},"solver status",{"type":54,"tag":102,"props":926,"children":927},{},[928],{"type":59,"value":548},{"type":54,"tag":102,"props":930,"children":931},{},[932],{"type":59,"value":933},"brief explanation of what the objective means",{"type":54,"tag":102,"props":935,"children":936},{},[937],{"type":59,"value":543},{"type":54,"tag":102,"props":939,"children":940},{},[941],{"type":59,"value":942},"any assumptions or caveats that materially affect the result",{"type":54,"tag":62,"props":944,"children":945},{},[946],{"type":59,"value":947},"If relevant, also mention which constraint or resource appears to be most limiting.",{"type":54,"tag":81,"props":949,"children":951},{"id":950},"handoff-guidance",[952],{"type":59,"value":953},"Handoff guidance",{"type":54,"tag":98,"props":955,"children":956},{},[957,1014,1060],{"type":54,"tag":102,"props":958,"children":959},{},[960,962],{"type":59,"value":961},"For LP \u002F MILP:",{"type":54,"tag":98,"props":963,"children":964},{},[965,974,991,1002],{"type":54,"tag":102,"props":966,"children":967},{},[968,969],{"type":59,"value":373},{"type":54,"tag":180,"props":970,"children":972},{"className":971},[],[973],{"type":59,"value":366},{"type":54,"tag":102,"props":975,"children":976},{},[977,979,984,985,990],{"type":59,"value":978},"then use ",{"type":54,"tag":180,"props":980,"children":982},{"className":981},[],[983],{"type":59,"value":379},{"type":59,"value":381},{"type":54,"tag":180,"props":986,"children":988},{"className":987},[],[989],{"type":59,"value":387},{"type":59,"value":389},{"type":54,"tag":102,"props":992,"children":993},{},[994,996],{"type":59,"value":995},"follow ",{"type":54,"tag":180,"props":997,"children":999},{"className":998},[],[1000],{"type":59,"value":1001},"cuopt-user-rules",{"type":54,"tag":102,"props":1003,"children":1004},{},[1005,1007,1012],{"type":59,"value":1006},"in sandbox contexts, follow ",{"type":54,"tag":180,"props":1008,"children":1010},{"className":1009},[],[1011],{"type":59,"value":232},{"type":59,"value":1013}," (gates + selected gRPC path)\nbefore any LP\u002FMILP solve",{"type":54,"tag":102,"props":1015,"children":1016},{},[1017,1019],{"type":59,"value":1018},"For QP:",{"type":54,"tag":98,"props":1020,"children":1021},{},[1022,1031,1040,1049],{"type":54,"tag":102,"props":1023,"children":1024},{},[1025,1026],{"type":59,"value":373},{"type":54,"tag":180,"props":1027,"children":1029},{"className":1028},[],[1030],{"type":59,"value":366},{"type":54,"tag":102,"props":1032,"children":1033},{},[1034,1035],{"type":59,"value":978},{"type":54,"tag":180,"props":1036,"children":1038},{"className":1037},[],[1039],{"type":59,"value":379},{"type":54,"tag":102,"props":1041,"children":1042},{},[1043,1044],{"type":59,"value":995},{"type":54,"tag":180,"props":1045,"children":1047},{"className":1046},[],[1048],{"type":59,"value":1001},{"type":54,"tag":102,"props":1050,"children":1051},{},[1052,1053,1058],{"type":59,"value":1006},{"type":54,"tag":180,"props":1054,"children":1056},{"className":1055},[],[1057],{"type":59,"value":232},{"type":59,"value":1059}," (gates + selected gRPC path)\nbefore any QP solve",{"type":54,"tag":102,"props":1061,"children":1062},{},[1063,1065],{"type":59,"value":1064},"For routing:",{"type":54,"tag":98,"props":1066,"children":1067},{},[1068,1077,1086,1095],{"type":54,"tag":102,"props":1069,"children":1070},{},[1071,1072],{"type":59,"value":373},{"type":54,"tag":180,"props":1073,"children":1075},{"className":1074},[],[1076],{"type":59,"value":446},{"type":54,"tag":102,"props":1078,"children":1079},{},[1080,1081],{"type":59,"value":978},{"type":54,"tag":180,"props":1082,"children":1084},{"className":1083},[],[1085],{"type":59,"value":456},{"type":54,"tag":102,"props":1087,"children":1088},{},[1089,1090],{"type":59,"value":995},{"type":54,"tag":180,"props":1091,"children":1093},{"className":1092},[],[1094],{"type":59,"value":1001},{"type":54,"tag":102,"props":1096,"children":1097},{},[1098,1099,1104],{"type":59,"value":1006},{"type":54,"tag":180,"props":1100,"children":1102},{"className":1101},[],[1103],{"type":59,"value":232},{"type":59,"value":1105}," (gates + REST)\nbefore any routing solve",{"type":54,"tag":81,"props":1107,"children":1109},{"id":1108},"success-criterion",[1110],{"type":59,"value":1111},"Success criterion",{"type":54,"tag":62,"props":1113,"children":1114},{},[1115],{"type":59,"value":1116},"This skill succeeds when the agent can go from:",{"type":54,"tag":98,"props":1118,"children":1119},{},[1120,1125],{"type":54,"tag":102,"props":1121,"children":1122},{},[1123],{"type":59,"value":1124},"interpreted data",{"type":54,"tag":102,"props":1126,"children":1127},{},[1128],{"type":59,"value":1129},"plus zero or a few blocking clarifications",{"type":54,"tag":62,"props":1131,"children":1132},{},[1133],{"type":59,"value":1134},"to:",{"type":54,"tag":98,"props":1136,"children":1137},{},[1138,1143,1148],{"type":54,"tag":102,"props":1139,"children":1140},{},[1141],{"type":59,"value":1142},"direct cuOpt model construction",{"type":54,"tag":102,"props":1144,"children":1145},{},[1146],{"type":59,"value":1147},"a validated solve",{"type":54,"tag":102,"props":1149,"children":1150},{},[1151],{"type":59,"value":1152},"and a clear business-facing answer",{"type":54,"tag":62,"props":1154,"children":1155},{},[1156],{"type":59,"value":1157},"without introducing unnecessary replay\u002Faudit machinery.",{"items":1159,"total":1246},[1160,1175,1181,1195,1208,1223,1237],{"slug":1161,"name":1161,"fn":1162,"description":1163,"org":1164,"tags":1165,"stars":20,"repoUrl":21,"updatedAt":1174},"cuopt-debugging","debug NVIDIA cuOpt optimization problems","Troubleshoot cuOpt LP\u002FMILP problems including errors, wrong results, infeasible solutions, performance issues, and status codes. Use when the user says something isn't working, gets unexpected results, or needs help diagnosing issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1166,1169,1172,1173],{"name":1167,"slug":1168,"type":15},"Debugging","debugging",{"name":1170,"slug":1171,"type":15},"Mathematics","mathematics",{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},"2026-07-23T05:43:32.51428",{"slug":4,"name":4,"fn":5,"description":6,"org":1176,"tags":1177,"stars":20,"repoUrl":21,"updatedAt":22},{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1178,1179,1180],{"name":17,"slug":18,"type":15},{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},{"slug":232,"name":232,"fn":1182,"description":1183,"org":1184,"tags":1185,"stars":20,"repoUrl":21,"updatedAt":1194},"run cuOpt optimization in NemoClaw sandbox","Run cuOpt in the NemoClaw sandbox — probe\u002Fsmoke gates, prefer cancelable Python gRPC jobs, use legacy remote execution only when that API is unavailable, then vendored cuOpt skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1186,1189,1190,1191],{"name":1187,"slug":1188,"type":15},"CLI","cli",{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},{"name":1192,"slug":1193,"type":15},"Simulation","simulation","2026-07-30T05:29:17.131364",{"slug":1196,"name":1196,"fn":1197,"description":1198,"org":1199,"tags":1200,"stars":20,"repoUrl":21,"updatedAt":1207},"generic-max-supply","plan supply chain models with cuOpt","Multi-period supply chain planning model: data files, BOM structure, variable\u002Fconstraint reference for the max-supply base model.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1201,1202,1203,1204],{"name":17,"slug":18,"type":15},{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},{"name":1205,"slug":1206,"type":15},"Supply Chain","supply-chain","2026-07-14T05:32:29.39577",{"slug":1209,"name":1209,"fn":1210,"description":1211,"org":1212,"tags":1213,"stars":20,"repoUrl":21,"updatedAt":1222},"optimization-from-data-orchestrator","orchestrate optimization data and solving","Coordinate uploaded data plus a natural-language question into interpretation, clarification, cuOpt solve, and a user-facing answer.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1214,1217,1220,1221],{"name":1215,"slug":1216,"type":15},"Automation","automation",{"name":1218,"slug":1219,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},"2026-07-14T05:32:59.680482",{"slug":185,"name":185,"fn":1224,"description":1225,"org":1226,"tags":1227,"stars":20,"repoUrl":21,"updatedAt":1236},"classify optimization and analytics requests","Classify whether a data-backed request is LP, MILP, QP, routing, or non-optimization analytics.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1228,1231,1234,1235],{"name":1229,"slug":1230,"type":15},"Analytics","analytics",{"name":1232,"slug":1233,"type":15},"Data Analysis","data-analysis",{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},"2026-07-14T05:32:53.475243",{"slug":194,"name":194,"fn":1238,"description":1239,"org":1240,"tags":1241,"stars":20,"repoUrl":21,"updatedAt":1245},"route optimization requests to cuOpt solvers","Choose fast direct-to-cuOpt solve versus replayable or auditable model artifact mode.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1242,1243,1244],{"name":1215,"slug":1216,"type":15},{"name":9,"slug":8,"type":15},{"name":13,"slug":14,"type":15},"2026-07-30T05:29:18.162351",8,{"items":1248,"total":1403},[1249,1267,1285,1296,1308,1320,1333,1347,1360,1371,1385,1394],{"slug":1250,"name":1250,"fn":1251,"description":1252,"org":1253,"tags":1254,"stars":1264,"repoUrl":1265,"updatedAt":1266},"nemoclaw-user-guide","retrieve NemoClaw documentation and configuration","Guides human users' AI agents to the NemoClaw docs MCP server and canonical Fern documentation in Markdown form. Use when users ask how to install, configure, operate, troubleshoot, secure, or learn NemoClaw with an AI coding assistant. Trigger keywords - nemoclaw docs, use nemoclaw with ai agent, nemoclaw mcp docs, nemoclaw install help, nemoclaw quickstart, nemoclaw markdown docs, llms.txt, agent skills.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1255,1258,1261],{"name":1256,"slug":1257,"type":15},"Documentation","documentation",{"name":1259,"slug":1260,"type":15},"MCP","mcp",{"name":1262,"slug":1263,"type":15},"Search","search",21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":1268,"name":1268,"fn":1269,"description":1270,"org":1271,"tags":1272,"stars":1282,"repoUrl":1283,"updatedAt":1284},"mcore-build-and-dependency","manage Megatron-LM development environments","Container-based dev environment setup and dependency management for Megatron-LM. Covers acquiring and launching the CI container, uv package management, and updating uv.lock.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1273,1276,1279],{"name":1274,"slug":1275,"type":15},"Containers","containers",{"name":1277,"slug":1278,"type":15},"Deployment","deployment",{"name":1280,"slug":1281,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":1286,"name":1286,"fn":1287,"description":1288,"org":1289,"tags":1290,"stars":1282,"repoUrl":1283,"updatedAt":1295},"mcore-bump-base-image","update NVIDIA PyTorch base images","Bump the NVIDIA PyTorch base image (`nvcr.io\u002Fnvidia\u002Fpytorch:YY.MM-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker\u002F.ngc_version.dev` and GitLab CI in `.gitlab\u002Fstages\u002F01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1291,1294],{"name":1292,"slug":1293,"type":15},"CI\u002FCD","ci-cd",{"name":1277,"slug":1278,"type":15},"2026-07-14T05:25:59.97109",{"slug":1297,"name":1297,"fn":1298,"description":1299,"org":1300,"tags":1301,"stars":1282,"repoUrl":1283,"updatedAt":1307},"mcore-cicd","manage CI\u002FCD pipelines for Megatron-LM","CI\u002FCD reference for Megatron-LM. Covers CI pipeline structure, PR scope labels, triggering internal GitLab CI (which force-pushes the current branch to a pull-request\u002FBRANCH ref — always dry-run and verify the destination first; never run against shared or protected branches), and CI failure investigation.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1302,1303,1304],{"name":1292,"slug":1293,"type":15},{"name":1277,"slug":1278,"type":15},{"name":1305,"slug":1306,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":1309,"name":1309,"fn":1310,"description":1311,"org":1312,"tags":1313,"stars":1282,"repoUrl":1283,"updatedAt":1319},"mcore-create-issue","investigate CI failures and create issues","Investigate a failing GitHub Actions run or job and create a GitHub issue for the failure.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1314,1315,1316],{"name":1167,"slug":1168,"type":15},{"name":1305,"slug":1306,"type":15},{"name":1317,"slug":1318,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":1321,"name":1321,"fn":1322,"description":1323,"org":1324,"tags":1325,"stars":1282,"repoUrl":1283,"updatedAt":1332},"mcore-linting-and-formatting","lint and format Megatron-LM code","Linting and formatting for Megatron-LM. Covers running autoformat.sh, tools (ruff, black, isort, pylint, mypy), and code style rules.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1326,1329],{"name":1327,"slug":1328,"type":15},"Best Practices","best-practices",{"name":1330,"slug":1331,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":1334,"name":1334,"fn":1335,"description":1336,"org":1337,"tags":1338,"stars":1282,"repoUrl":1283,"updatedAt":1346},"mcore-migrate-gpt-to-hybrid","migrate Megatron-LM models to HybridModel","Migration guide for moving Megatron Core GPTModel checkpoints, model providers, training commands, and layer mappings to HybridModel.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1339,1342,1345],{"name":1340,"slug":1341,"type":15},"Machine Learning","machine-learning",{"name":1343,"slug":1344,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":1348,"name":1348,"fn":1349,"description":1350,"org":1351,"tags":1352,"stars":1282,"repoUrl":1283,"updatedAt":1359},"mcore-onboard-gb200-1node-tests","onboard functional tests for GB200","Onboard 1-node GitHub MR functional tests for GB200 from existing mr-scoped 2-node tests.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1353,1356],{"name":1354,"slug":1355,"type":15},"QA","qa",{"name":1357,"slug":1358,"type":15},"Testing","testing","2026-07-14T05:25:53.673039",{"slug":1361,"name":1361,"fn":1362,"description":1363,"org":1364,"tags":1365,"stars":1282,"repoUrl":1283,"updatedAt":1370},"mcore-run-on-slurm","launch distributed training jobs on SLURM","How to launch distributed Megatron-LM training jobs on a SLURM cluster. Covers a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules across hardware and parallelism modes, container conventions, monitoring, and per-rank failure diagnosis.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1366,1367],{"name":1277,"slug":1278,"type":15},{"name":1368,"slug":1369,"type":15},"Infrastructure","infrastructure","2026-07-14T05:25:49.362534",{"slug":1372,"name":1372,"fn":1373,"description":1374,"org":1375,"tags":1376,"stars":1282,"repoUrl":1283,"updatedAt":1384},"mcore-split-pr","split pull requests to reduce review load","Split a PR into multiple PRs to reduce the number of required CODEOWNERS reviewer groups.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1377,1380,1381],{"name":1378,"slug":1379,"type":15},"Code Review","code-review",{"name":1305,"slug":1306,"type":15},{"name":1382,"slug":1383,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":1386,"name":1386,"fn":1387,"description":1388,"org":1389,"tags":1390,"stars":1282,"repoUrl":1283,"updatedAt":1393},"mcore-testing","run and manage Megatron-LM tests","Test system for Megatron-LM. Covers test layout, recipe YAML structure, adding and running unit and functional tests, golden values, marker filters, and CI parity.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1391,1392],{"name":1354,"slug":1355,"type":15},{"name":1357,"slug":1358,"type":15},"2026-07-14T05:25:54.928983",{"slug":1395,"name":1395,"fn":1396,"description":1397,"org":1398,"tags":1399,"stars":1282,"repoUrl":1283,"updatedAt":1402},"nightly-sync","manage nightly main-to-dev sync workflows","Domain knowledge for the nightly main-to-dev sync workflow. Covers merge strategy, CI architecture, failure investigation, and known issues.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[1400,1401],{"name":1215,"slug":1216,"type":15},{"name":1292,"slug":1293,"type":15},"2026-07-30T05:29:03.275638",496]