[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill-nvidia-nemo-retriever":3,"mdc-8d5ema-key":37,"related-repo-nvidia-nemo-retriever":484,"related-org-nvidia-nemo-retriever":590},{"slug":4,"name":4,"fn":5,"description":6,"org":7,"tags":11,"stars":26,"repoUrl":27,"updatedAt":28,"license":29,"forks":30,"topics":31,"repo":32,"sourceUrl":35,"mdContent":36},"nemo-retriever","search and aggregate across documents","Use when the user wants to search, query, extract, transcribe, describe, quote, filter, or aggregate across documents — PDFs, scanned forms \u002F images (`.jpg` `.png` `.tiff`), Office (`.docx` `.pptx`), text (`.html` `.txt`), audio (`.mp3` `.wav` `.m4a`), or video (`.mp4` `.mov`). Prefer this over native Read \u002F Grep for multi-file or non-PDF corpora. Not for: editing files, web browsing, single-file plain-text lookups, fine-tuning.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},"nvidia","NVIDIA","https:\u002F\u002Fpexgzepcugksgbtrxkhf.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Forg-logos\u002Fnvidia.png",[12,16,19,20,23],{"name":13,"slug":14,"type":15},"PDF","pdf","tag",{"name":17,"slug":18,"type":15},"Documents","documents",{"name":9,"slug":8,"type":15},{"name":21,"slug":22,"type":15},"Search","search",{"name":24,"slug":25,"type":15},"Office","office",2473,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills","2026-07-14T05:29:19.424084","Apache-2.0",281,[],{"repoUrl":27,"stars":26,"forks":30,"topics":33,"description":34},[],"AI agent skills published by NVIDIA","https:\u002F\u002Fgithub.com\u002FNVIDIA\u002Fskills\u002Ftree\u002FHEAD\u002Fskills\u002Fnemo-retriever","---\nname: nemo-retriever\ndescription: \"Use when the user wants to search, query, extract, transcribe, describe, quote, filter, or aggregate across documents — PDFs, scanned forms \u002F images (`.jpg` `.png` `.tiff`), Office (`.docx` `.pptx`), text (`.html` `.txt`), audio (`.mp3` `.wav` `.m4a`), or video (`.mp4` `.mov`). Prefer this over native Read \u002F Grep for multi-file or non-PDF corpora. Not for: editing files, web browsing, single-file plain-text lookups, fine-tuning.\"\nlicense: Apache-2.0\nallowed-tools: Bash Write Read\n---\n\n# nemo-retriever\n\nThe `retriever` CLI indexes a folder of PDFs into LanceDB (`retriever ingest`) and serves vector search over it (`retriever query`). For any task about searching\u002Fanswering questions across a folder of PDFs, use this CLI — do not write a custom RAG.\n\n**Beyond PDFs and beyond semantic search.** `retriever ingest` also handles images, Office, HTML, TXT, audio, and video — see `references\u002Fsetup.md` for the per-format recipe and `references\u002Finstall.md` for the install extras (`[multimedia]`, libreoffice, ffmpeg). For non-semantic operations — page filter, verbatim quote with citation, corpus-level aggregate, chart\u002Fimage caption hits — see `references\u002Fquery.md`. Don't fall back to native Read\u002FGrep\u002FPython on non-PDF inputs.\n\n## Install (if `retriever` is missing)\n\nIf `command -v retriever` returns nothing, follow `references\u002Finstall.md` to install the NeMo Retriever Library before proceeding. It prints `RETRIEVER_VENV=\u003Cpath>`; substitute that path for `\u003CRETRIEVER_VENV>` in every example in this skill (setup, query, troubleshooting, and the CLI references).\n\n## Workflow — read the reference for the current phase, then execute\n\n| Turn type | Read this once | Then execute |\n| :--- | :--- | :--- |\n| **Setup turn** (first turn — `.\u002Flancedb\u002Fnv-ingest.lance` doesn't exist) | `references\u002Fsetup.md` | Build the index |\n| **Query turn** (every subsequent turn — user asks a question) | `references\u002Fquery.md` | One `retriever query` call |\n| Anything errored or returned empty | `references\u002Ftroubleshooting.md` | Apply the named recovery; do not improvise |\n\nFor the full `retriever ingest` \u002F `retriever query` CLI specs, see `references\u002Fcli\u002Fingest.md` and `references\u002Fcli\u002Fquery.md`. You do not need these for routine turns — `\u003CRETRIEVER_VENV>\u002Fbin\u002Fretriever \u003Csubcommand> --help` is faster.\n\nBefore ingesting a mixed folder, inventory extensions (`find \u003Cdir> -name '*.*' | sed 's\u002F.*\\.\u002F\u002F' | sort -u`) — `--input-type=auto` silently drops anything outside the supported set. See `references\u002Ftroubleshooting.md` \"Unsupported file types\".\n\n## Hard limits (apply to every turn)\n\n- **Setup turn**: build the index in one shell command (see `references\u002Fsetup.md`). STOP after the index lands.\n- **Query turn**: at most **2 Bash calls** — 1 `retriever query`, +1 optional targeted text-extract per `references\u002Fquery.md`. Reply and then STOP.\n- **No narration between tool calls.** Tokens you emit between calls become input + cached input for every later turn — quadratic cost. Go straight from reading the summary to writing the JSON file.\n- **Banned**: `TodoWrite`, Glob, Grep, `Read` of whole PDFs, re-running setup, spawning subagents, speculative \"confirmation\" calls.\n\nLong query turns (5+ tool calls, 1M+ cache-read tokens) cost ~5× a disciplined turn and almost always still produce the wrong answer. **Answering partially beats timing out.**\n",{"data":38,"body":40},{"name":4,"description":6,"license":29,"allowed-tools":39},"Bash Write Read",{"type":41,"children":42},"root",[43,50,81,131,145,181,187,310,353,381,387,474],{"type":44,"tag":45,"props":46,"children":47},"element","h1",{"id":4},[48],{"type":49,"value":4},"text",{"type":44,"tag":51,"props":52,"children":53},"p",{},[54,56,63,65,71,73,79],{"type":49,"value":55},"The ",{"type":44,"tag":57,"props":58,"children":60},"code",{"className":59},[],[61],{"type":49,"value":62},"retriever",{"type":49,"value":64}," CLI indexes a folder of PDFs into LanceDB (",{"type":44,"tag":57,"props":66,"children":68},{"className":67},[],[69],{"type":49,"value":70},"retriever ingest",{"type":49,"value":72},") and serves vector search over it (",{"type":44,"tag":57,"props":74,"children":76},{"className":75},[],[77],{"type":49,"value":78},"retriever query",{"type":49,"value":80},"). For any task about searching\u002Fanswering questions across a folder of PDFs, use this CLI — do not write a custom RAG.",{"type":44,"tag":51,"props":82,"children":83},{},[84,90,92,97,99,105,107,113,115,121,123,129],{"type":44,"tag":85,"props":86,"children":87},"strong",{},[88],{"type":49,"value":89},"Beyond PDFs and beyond semantic search.",{"type":49,"value":91}," ",{"type":44,"tag":57,"props":93,"children":95},{"className":94},[],[96],{"type":49,"value":70},{"type":49,"value":98}," also handles images, Office, HTML, TXT, audio, and video — see ",{"type":44,"tag":57,"props":100,"children":102},{"className":101},[],[103],{"type":49,"value":104},"references\u002Fsetup.md",{"type":49,"value":106}," for the per-format recipe and ",{"type":44,"tag":57,"props":108,"children":110},{"className":109},[],[111],{"type":49,"value":112},"references\u002Finstall.md",{"type":49,"value":114}," for the install extras (",{"type":44,"tag":57,"props":116,"children":118},{"className":117},[],[119],{"type":49,"value":120},"[multimedia]",{"type":49,"value":122},", libreoffice, ffmpeg). For non-semantic operations — page filter, verbatim quote with citation, corpus-level aggregate, chart\u002Fimage caption hits — see ",{"type":44,"tag":57,"props":124,"children":126},{"className":125},[],[127],{"type":49,"value":128},"references\u002Fquery.md",{"type":49,"value":130},". Don't fall back to native Read\u002FGrep\u002FPython on non-PDF inputs.",{"type":44,"tag":132,"props":133,"children":135},"h2",{"id":134},"install-if-retriever-is-missing",[136,138,143],{"type":49,"value":137},"Install (if ",{"type":44,"tag":57,"props":139,"children":141},{"className":140},[],[142],{"type":49,"value":62},{"type":49,"value":144}," is missing)",{"type":44,"tag":51,"props":146,"children":147},{},[148,150,156,158,163,165,171,173,179],{"type":49,"value":149},"If ",{"type":44,"tag":57,"props":151,"children":153},{"className":152},[],[154],{"type":49,"value":155},"command -v retriever",{"type":49,"value":157}," returns nothing, follow ",{"type":44,"tag":57,"props":159,"children":161},{"className":160},[],[162],{"type":49,"value":112},{"type":49,"value":164}," to install the NeMo Retriever Library before proceeding. It prints ",{"type":44,"tag":57,"props":166,"children":168},{"className":167},[],[169],{"type":49,"value":170},"RETRIEVER_VENV=\u003Cpath>",{"type":49,"value":172},"; substitute that path for ",{"type":44,"tag":57,"props":174,"children":176},{"className":175},[],[177],{"type":49,"value":178},"\u003CRETRIEVER_VENV>",{"type":49,"value":180}," in every example in this skill (setup, query, troubleshooting, and the CLI references).",{"type":44,"tag":132,"props":182,"children":184},{"id":183},"workflow-read-the-reference-for-the-current-phase-then-execute",[185],{"type":49,"value":186},"Workflow — read the reference for the current phase, then execute",{"type":44,"tag":188,"props":189,"children":190},"table",{},[191,216],{"type":44,"tag":192,"props":193,"children":194},"thead",{},[195],{"type":44,"tag":196,"props":197,"children":198},"tr",{},[199,206,211],{"type":44,"tag":200,"props":201,"children":203},"th",{"align":202},"left",[204],{"type":49,"value":205},"Turn type",{"type":44,"tag":200,"props":207,"children":208},{"align":202},[209],{"type":49,"value":210},"Read this once",{"type":44,"tag":200,"props":212,"children":213},{"align":202},[214],{"type":49,"value":215},"Then execute",{"type":44,"tag":217,"props":218,"children":219},"tbody",{},[220,255,288],{"type":44,"tag":196,"props":221,"children":222},{},[223,242,250],{"type":44,"tag":224,"props":225,"children":226},"td",{"align":202},[227,232,234,240],{"type":44,"tag":85,"props":228,"children":229},{},[230],{"type":49,"value":231},"Setup turn",{"type":49,"value":233}," (first turn — ",{"type":44,"tag":57,"props":235,"children":237},{"className":236},[],[238],{"type":49,"value":239},".\u002Flancedb\u002Fnv-ingest.lance",{"type":49,"value":241}," doesn't exist)",{"type":44,"tag":224,"props":243,"children":244},{"align":202},[245],{"type":44,"tag":57,"props":246,"children":248},{"className":247},[],[249],{"type":49,"value":104},{"type":44,"tag":224,"props":251,"children":252},{"align":202},[253],{"type":49,"value":254},"Build the index",{"type":44,"tag":196,"props":256,"children":257},{},[258,268,276],{"type":44,"tag":224,"props":259,"children":260},{"align":202},[261,266],{"type":44,"tag":85,"props":262,"children":263},{},[264],{"type":49,"value":265},"Query turn",{"type":49,"value":267}," (every subsequent turn — user asks a question)",{"type":44,"tag":224,"props":269,"children":270},{"align":202},[271],{"type":44,"tag":57,"props":272,"children":274},{"className":273},[],[275],{"type":49,"value":128},{"type":44,"tag":224,"props":277,"children":278},{"align":202},[279,281,286],{"type":49,"value":280},"One ",{"type":44,"tag":57,"props":282,"children":284},{"className":283},[],[285],{"type":49,"value":78},{"type":49,"value":287}," call",{"type":44,"tag":196,"props":289,"children":290},{},[291,296,305],{"type":44,"tag":224,"props":292,"children":293},{"align":202},[294],{"type":49,"value":295},"Anything errored or returned empty",{"type":44,"tag":224,"props":297,"children":298},{"align":202},[299],{"type":44,"tag":57,"props":300,"children":302},{"className":301},[],[303],{"type":49,"value":304},"references\u002Ftroubleshooting.md",{"type":44,"tag":224,"props":306,"children":307},{"align":202},[308],{"type":49,"value":309},"Apply the named recovery; do not improvise",{"type":44,"tag":51,"props":311,"children":312},{},[313,315,320,322,327,329,335,337,343,345,351],{"type":49,"value":314},"For the full ",{"type":44,"tag":57,"props":316,"children":318},{"className":317},[],[319],{"type":49,"value":70},{"type":49,"value":321}," \u002F ",{"type":44,"tag":57,"props":323,"children":325},{"className":324},[],[326],{"type":49,"value":78},{"type":49,"value":328}," CLI specs, see ",{"type":44,"tag":57,"props":330,"children":332},{"className":331},[],[333],{"type":49,"value":334},"references\u002Fcli\u002Fingest.md",{"type":49,"value":336}," and ",{"type":44,"tag":57,"props":338,"children":340},{"className":339},[],[341],{"type":49,"value":342},"references\u002Fcli\u002Fquery.md",{"type":49,"value":344},". You do not need these for routine turns — ",{"type":44,"tag":57,"props":346,"children":348},{"className":347},[],[349],{"type":49,"value":350},"\u003CRETRIEVER_VENV>\u002Fbin\u002Fretriever \u003Csubcommand> --help",{"type":49,"value":352}," is faster.",{"type":44,"tag":51,"props":354,"children":355},{},[356,358,364,366,372,374,379],{"type":49,"value":357},"Before ingesting a mixed folder, inventory extensions (",{"type":44,"tag":57,"props":359,"children":361},{"className":360},[],[362],{"type":49,"value":363},"find \u003Cdir> -name '*.*' | sed 's\u002F.*\\.\u002F\u002F' | sort -u",{"type":49,"value":365},") — ",{"type":44,"tag":57,"props":367,"children":369},{"className":368},[],[370],{"type":49,"value":371},"--input-type=auto",{"type":49,"value":373}," silently drops anything outside the supported set. See ",{"type":44,"tag":57,"props":375,"children":377},{"className":376},[],[378],{"type":49,"value":304},{"type":49,"value":380}," \"Unsupported file types\".",{"type":44,"tag":132,"props":382,"children":384},{"id":383},"hard-limits-apply-to-every-turn",[385],{"type":49,"value":386},"Hard limits (apply to every turn)",{"type":44,"tag":388,"props":389,"children":390},"ul",{},[391,408,438,448],{"type":44,"tag":392,"props":393,"children":394},"li",{},[395,399,401,406],{"type":44,"tag":85,"props":396,"children":397},{},[398],{"type":49,"value":231},{"type":49,"value":400},": build the index in one shell command (see ",{"type":44,"tag":57,"props":402,"children":404},{"className":403},[],[405],{"type":49,"value":104},{"type":49,"value":407},"). STOP after the index lands.",{"type":44,"tag":392,"props":409,"children":410},{},[411,415,417,422,424,429,431,436],{"type":44,"tag":85,"props":412,"children":413},{},[414],{"type":49,"value":265},{"type":49,"value":416},": at most ",{"type":44,"tag":85,"props":418,"children":419},{},[420],{"type":49,"value":421},"2 Bash calls",{"type":49,"value":423}," — 1 ",{"type":44,"tag":57,"props":425,"children":427},{"className":426},[],[428],{"type":49,"value":78},{"type":49,"value":430},", +1 optional targeted text-extract per ",{"type":44,"tag":57,"props":432,"children":434},{"className":433},[],[435],{"type":49,"value":128},{"type":49,"value":437},". Reply and then STOP.",{"type":44,"tag":392,"props":439,"children":440},{},[441,446],{"type":44,"tag":85,"props":442,"children":443},{},[444],{"type":49,"value":445},"No narration between tool calls.",{"type":49,"value":447}," Tokens you emit between calls become input + cached input for every later turn — quadratic cost. Go straight from reading the summary to writing the JSON file.",{"type":44,"tag":392,"props":449,"children":450},{},[451,456,458,464,466,472],{"type":44,"tag":85,"props":452,"children":453},{},[454],{"type":49,"value":455},"Banned",{"type":49,"value":457},": ",{"type":44,"tag":57,"props":459,"children":461},{"className":460},[],[462],{"type":49,"value":463},"TodoWrite",{"type":49,"value":465},", Glob, Grep, ",{"type":44,"tag":57,"props":467,"children":469},{"className":468},[],[470],{"type":49,"value":471},"Read",{"type":49,"value":473}," of whole PDFs, re-running setup, spawning subagents, speculative \"confirmation\" calls.",{"type":44,"tag":51,"props":475,"children":476},{},[477,479],{"type":49,"value":478},"Long query turns (5+ tool calls, 1M+ cache-read tokens) cost ~5× a disciplined turn and almost always still produce the wrong answer. ",{"type":44,"tag":85,"props":480,"children":481},{},[482],{"type":49,"value":483},"Answering partially beats timing out.",{"items":485,"total":589},[486,503,517,531,543,560,575],{"slug":487,"name":487,"fn":488,"description":489,"org":490,"tags":491,"stars":26,"repoUrl":27,"updatedAt":502},"accelerated-computing-cudf","accelerate data processing with cuDF","Official NVIDIA-authored guidance for NVIDIA cuDF GPU DataFrames, pandas acceleration, dask-cuDF, ETL, joins, groupby, CSV\u002FParquet I\u002FO, nullable semantics, and multi-GPU DataFrame workloads.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[492,495,498,499],{"name":493,"slug":494,"type":15},"Data Analysis","data-analysis",{"name":496,"slug":497,"type":15},"Data Engineering","data-engineering",{"name":9,"slug":8,"type":15},{"name":500,"slug":501,"type":15},"Performance","performance","2026-07-14T05:28:43.176466",{"slug":504,"name":504,"fn":505,"description":506,"org":507,"tags":508,"stars":26,"repoUrl":27,"updatedAt":516},"aiq-deploy","deploy and manage NVIDIA AI-Q infrastructure","Use when asked to install, deploy, run, validate, troubleshoot, or stop NVIDIA AI-Q Blueprint infrastructure.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[509,512,515],{"name":510,"slug":511,"type":15},"Deployment","deployment",{"name":513,"slug":514,"type":15},"Infrastructure","infrastructure",{"name":9,"slug":8,"type":15},"2026-07-14T05:29:06.667109",{"slug":518,"name":518,"fn":519,"description":520,"org":521,"tags":522,"stars":26,"repoUrl":27,"updatedAt":530},"aiq-research","conduct deep research with AI-Q","Use when asked to run deep research or AI-Q research through a reachable NVIDIA AI-Q Blueprint backend.\n",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[523,526,527],{"name":524,"slug":525,"type":15},"Agents","agents",{"name":9,"slug":8,"type":15},{"name":528,"slug":529,"type":15},"Research","research","2026-07-14T05:28:06.816956",{"slug":532,"name":532,"fn":533,"description":534,"org":535,"tags":536,"stars":26,"repoUrl":27,"updatedAt":542},"amc-run-sample-calibration","run AMC sample dataset calibration","Run end-to-end calibration on the shipped sample dataset (sdg_08_2_sample_data_010926.zip) against a running AMC microservice. Use when user says 'test sample dataset', 'run sample calibration', 'verify AMC install', or 'launch and test'.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[537,538,539],{"name":493,"slug":494,"type":15},{"name":9,"slug":8,"type":15},{"name":540,"slug":541,"type":15},"Testing","testing","2026-07-17T05:29:03.913266",{"slug":544,"name":544,"fn":545,"description":546,"org":547,"tags":548,"stars":26,"repoUrl":27,"updatedAt":559},"amc-run-video-calibration","calibrate video datasets with AutoMagicCalib","Calibrate a new dataset from pre-recorded video files via the AutoMagicCalib REST API. Use when user has local MP4s and says 'calibrate my videos', 'run AMC on these videos', or similar. For RTSP\u002Flive streams, use amc-run-rtsp-calibration instead.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[549,552,555,556],{"name":550,"slug":551,"type":15},"Automation","automation",{"name":553,"slug":554,"type":15},"Imaging","imaging",{"name":9,"slug":8,"type":15},{"name":557,"slug":558,"type":15},"Video","video","2026-07-17T05:28:53.905004",{"slug":561,"name":561,"fn":562,"description":563,"org":564,"tags":565,"stars":26,"repoUrl":27,"updatedAt":574},"amc-setup-calibration-stack","deploy AutoMagicCalib microservice with Docker","Launch AutoMagicCalib microservice and web UI from NGC release images via Docker Compose. Use when user says 'deploy auto calibration', 'launch auto calibration', 'launch AMC', 'start MS+UI', or 'set up auto-magic-calib'. Requires NGC API key.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[566,567,570,571],{"name":510,"slug":511,"type":15},{"name":568,"slug":569,"type":15},"Docker","docker",{"name":9,"slug":8,"type":15},{"name":572,"slug":573,"type":15},"Operations","operations","2026-07-17T05:28:56.913999",{"slug":576,"name":576,"fn":577,"description":578,"org":579,"tags":580,"stars":26,"repoUrl":27,"updatedAt":588},"cudaq-guide","develop quantum applications with CUDA-Q","CUDA-Q onboarding guide for installation, test programs, GPU simulation, QPU hardware, and quantum applications.",{"slug":8,"name":9,"logoUrl":10,"githubOrg":9},[581,582,585],{"name":9,"slug":8,"type":15},{"name":583,"slug":584,"type":15},"Quantum Computing","quantum-computing",{"name":586,"slug":587,"type":15},"Simulation","simulation","2026-07-14T05:26:58.898253",305,{"items":591,"total":740},[592,608,624,635,647,661,674,688,699,708,722,731],{"slug":593,"name":593,"fn":594,"description":595,"org":596,"tags":597,"stars":605,"repoUrl":606,"updatedAt":607},"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},[598,601,604],{"name":599,"slug":600,"type":15},"Documentation","documentation",{"name":602,"slug":603,"type":15},"MCP","mcp",{"name":21,"slug":22,"type":15},21777,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FNemoClaw","2026-07-20T06:00:01.461044",{"slug":609,"name":609,"fn":610,"description":611,"org":612,"tags":613,"stars":621,"repoUrl":622,"updatedAt":623},"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},[614,617,618],{"name":615,"slug":616,"type":15},"Containers","containers",{"name":510,"slug":511,"type":15},{"name":619,"slug":620,"type":15},"Python","python",17049,"https:\u002F\u002Fgithub.com\u002FNVIDIA\u002FMegatron-LM","2026-07-27T06:06:11.249662",{"slug":625,"name":625,"fn":626,"description":627,"org":628,"tags":629,"stars":621,"repoUrl":622,"updatedAt":634},"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},[630,633],{"name":631,"slug":632,"type":15},"CI\u002FCD","ci-cd",{"name":510,"slug":511,"type":15},"2026-07-14T05:25:59.97109",{"slug":636,"name":636,"fn":637,"description":638,"org":639,"tags":640,"stars":621,"repoUrl":622,"updatedAt":646},"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},[641,642,643],{"name":631,"slug":632,"type":15},{"name":510,"slug":511,"type":15},{"name":644,"slug":645,"type":15},"GitHub","github","2026-07-27T06:06:12.278222",{"slug":648,"name":648,"fn":649,"description":650,"org":651,"tags":652,"stars":621,"repoUrl":622,"updatedAt":660},"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},[653,656,657],{"name":654,"slug":655,"type":15},"Debugging","debugging",{"name":644,"slug":645,"type":15},{"name":658,"slug":659,"type":15},"Triage","triage","2026-07-14T05:25:57.442089",{"slug":662,"name":662,"fn":663,"description":664,"org":665,"tags":666,"stars":621,"repoUrl":622,"updatedAt":673},"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},[667,670],{"name":668,"slug":669,"type":15},"Best Practices","best-practices",{"name":671,"slug":672,"type":15},"Code Analysis","code-analysis","2026-07-14T05:25:56.18433",{"slug":675,"name":675,"fn":676,"description":677,"org":678,"tags":679,"stars":621,"repoUrl":622,"updatedAt":687},"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},[680,683,686],{"name":681,"slug":682,"type":15},"Machine Learning","machine-learning",{"name":684,"slug":685,"type":15},"Migration","migration",{"name":9,"slug":8,"type":15},"2026-07-17T06:07:11.777011",{"slug":689,"name":689,"fn":690,"description":691,"org":692,"tags":693,"stars":621,"repoUrl":622,"updatedAt":698},"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},[694,697],{"name":695,"slug":696,"type":15},"QA","qa",{"name":540,"slug":541,"type":15},"2026-07-14T05:25:53.673039",{"slug":700,"name":700,"fn":701,"description":702,"org":703,"tags":704,"stars":621,"repoUrl":622,"updatedAt":707},"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},[705,706],{"name":510,"slug":511,"type":15},{"name":513,"slug":514,"type":15},"2026-07-14T05:25:49.362534",{"slug":709,"name":709,"fn":710,"description":711,"org":712,"tags":713,"stars":621,"repoUrl":622,"updatedAt":721},"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},[714,717,718],{"name":715,"slug":716,"type":15},"Code Review","code-review",{"name":644,"slug":645,"type":15},{"name":719,"slug":720,"type":15},"Pull Requests","pull-requests","2026-07-14T05:26:01.226578",{"slug":723,"name":723,"fn":724,"description":725,"org":726,"tags":727,"stars":621,"repoUrl":622,"updatedAt":730},"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},[728,729],{"name":695,"slug":696,"type":15},{"name":540,"slug":541,"type":15},"2026-07-14T05:25:54.928983",{"slug":732,"name":732,"fn":733,"description":734,"org":735,"tags":736,"stars":621,"repoUrl":622,"updatedAt":739},"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},[737,738],{"name":550,"slug":551,"type":15},{"name":631,"slug":632,"type":15},"2026-07-30T05:29:03.275638",496]