{
  "schema_version": "1.1",
  "id": "s8:https://www.youtube.com/watch?v=lyL5QhgIOxc",
  "slug": "serving-2-million-models-without-melting-scaling-the-hugging-face-hub-ar-06hazr6",
  "url": "https://feed7.dev/p/serving-2-million-models-without-melting-scaling-the-hugging-face-hub-ar-06hazr6",
  "title": "Serving 2 Million Models Without Melting: Scaling the Hugging Face Hub — Arek Borucki, Hugging Face",
  "why_included": "Hugging Face’s Hub shows how to scale model search: precompute tokens, separate metadata from artifacts, isolate heavy reads, and autoscale from workload signals.",
  "summary": "Hugging Face reports **3 million public models**, **1 million datasets**, and more than **14 million users**. Model metadata lives in MongoDB Atlas, artifacts live in object storage, and search uses precomputed tokens plus Lucene-backed Atlas Search.",
  "practical_implication": "For large agent-facing catalogs, optimize the read path separately from writes and binary storage. Route stale-tolerant reads and heavy aggregation away from the primary, isolate reporting, and scale both application pods and cluster nodes.",
  "agent_context": "Hugging Face reports **3 million public models**, **1 million datasets**, and more than **14 million users**. Model metadata lives in MongoDB Atlas, artifacts live in object storage, and search uses precomputed tokens plus Lucene-backed Atlas Search.\n\nFor large agent-facing catalogs, optimize the read path separately from writes and binary storage. Route stale-tolerant reads and heavy aggregation away from the primary, isolate reporting, and scale both application pods and cluster nodes.\n\nThese choices reflect Hugging Face’s workload, not a universal blueprint. Sharding is still planned, and moving from CPU- and memory-based HPA to **KEDA workload metrics** is described as upcoming work.",
  "source": {
    "name": "AI Engineer",
    "url": "https://www.youtube.com/watch?v=lyL5QhgIOxc",
    "published_at": "2026-07-28T13:41:11.000Z"
  },
  "source_class": "video",
  "content_type": "Video",
  "layer": "infra",
  "domains": [
    "data"
  ],
  "topics": [
    "retrieval",
    "open-models"
  ],
  "verification": {
    "status": "source_linked",
    "label": "Source Linked",
    "method": "source_feed",
    "verified_at": null
  },
  "uncertainty": [
    "These choices reflect Hugging Face’s workload, not a universal blueprint. Sharding is still planned, and moving from CPU- and memory-based HPA to **KEDA workload metrics** is described as upcoming work."
  ],
  "connected_context": {
    "meaning": "This contributes a concrete read-heavy catalog architecture rather than another model-serving or retrieval technique: metadata, binaries, search, reporting, and stale-tolerant reads are scaled as distinct paths. It is useful as workload-specific evidence, not a finished blueprint, because sharding and workload-based autoscaling remain planned rather than demonstrated.",
    "corpus_size": 262,
    "generated_at": "2026-07-29T10:05:40.850Z",
    "connections": []
  },
  "lifecycle": "Current",
  "published_at": "2026-07-28T13:41:11.000Z",
  "modified_at": "2026-07-28T13:41:11.000Z",
  "supersedes": [],
  "expires_at": null,
  "formats": {
    "html": "https://feed7.dev/p/serving-2-million-models-without-melting-scaling-the-hugging-face-hub-ar-06hazr6",
    "json": "https://feed7.dev/p/serving-2-million-models-without-melting-scaling-the-hugging-face-hub-ar-06hazr6.json",
    "markdown": "https://feed7.dev/p/serving-2-million-models-without-melting-scaling-the-hugging-face-hub-ar-06hazr6.md"
  }
}