aeon-huggingface-trending
Daily filtered scan over HF's three trending surfaces — models, datasets, spaces — cluster-ranked rather than raw-download ranked.
Endpoints
curl -s "https://huggingface.co/api/models?sort=trending&direction=-1&limit=30"
curl -s "https://huggingface.co/api/datasets?sort=trending&direction=-1&limit=30"
curl -s "https://huggingface.co/api/spaces?sort=trending&direction=-1&limit=30"
Filters
- Empty repos, no commits, no model card → drop.
- Same-week quantization of an already-trending model → demoted to "Quantizations" tail.
- Fork without README delta vs upstream → drop.
- Authors with > 5 trending entries in 24h → demoted (typically aggregators).
- License unclear or missing → flagged inline, not dropped.
"Why notable" line
Per surfaced entry, a one-sentence tag: new architecture / size step / context-window jump / notable author affiliation / license change. If no concrete reason exists, the entry says "no clear why — popular but unremarkable" rather than inventing one.
Output
Three sections — Models, Datasets, Spaces — each with the surviving picks, "why notable", license, and download/view count. Tail section for quantizations.
Rules
- "Why notable" is a hard requirement. No reason → no surface.
- License flags appear inline (Apache 2.0, MIT, OpenRAIL-M, custom-no-commercial). Operators trading on model output care.
- Spaces section often beats Models for builder-tier signal — a working demo is stronger validation than a card claim.
1---2name: aeon-huggingface-trending3description: |4 Trending Hugging Face models, datasets, and spaces — filtered by license sanity, dedup vs same-week5 quantizations, with a "why notable" line per pick (architecture shift, size step, license change,6 notable author). Surfaces what's actually shifting rather than just popular.7 Triggers: "trending on HF", "what models are hot", "huggingface trending", "new spaces today",8 "best new datasets".9---1011# aeon-huggingface-trending1213Daily filtered scan over HF's three trending surfaces — models, datasets, spaces — cluster-ranked rather than raw-download ranked.1415## Endpoints1617```bash18curl -s "https://huggingface.co/api/models?sort=trending&direction=-1&limit=30"19curl -s "https://huggingface.co/api/datasets?sort=trending&direction=-1&limit=30"20curl -s "https://huggingface.co/api/spaces?sort=trending&direction=-1&limit=30"21```2223## Filters2425- Empty repos, no commits, no model card → drop.26- Same-week quantization of an already-trending model → demoted to "Quantizations" tail.27- Fork without README delta vs upstream → drop.28- Authors with > 5 trending entries in 24h → demoted (typically aggregators).29- License unclear or missing → flagged inline, not dropped.3031## "Why notable" line3233Per surfaced entry, a one-sentence tag: new architecture / size step / context-window jump / notable author affiliation / license change. If no concrete reason exists, the entry says "no clear why — popular but unremarkable" rather than inventing one.3435## Output3637Three sections — Models, Datasets, Spaces — each with the surviving picks, "why notable", license, and download/view count. Tail section for quantizations.3839## Rules4041- "Why notable" is a hard requirement. No reason → no surface.42- License flags appear inline (Apache 2.0, MIT, OpenRAIL-M, custom-no-commercial). Operators trading on model output care.43- Spaces section often beats Models for builder-tier signal — a working demo is stronger validation than a card claim.
BankrBot