Foryu
Find people researching areas similar to you (HuggingFace)
Last verified:
What is Foryu?
Foryu is a browser-based Hugging Face “taste twin” finder that estimates which users are most similar to you based on the models you like. You type a Hugging Face handle, and it expands your like graph two hops out, then ranks nearby users by cosine similarity to your taste profile.
The tool is designed to run entirely in your browser. The site says the only bundled data file is a vocabulary of the 10k most-liked Hugging Face models, along with like counts and IDF weights; your input, candidate search, and ranking are all done client-side.
Its core method is live graph expansion: it fetches your public likes from the Hugging Face API, takes distinctive likes, finds likers of those models, then looks at those users’ likes and scores the overlap. The site describes this as taking roughly 10 seconds, with about 70 API calls, and no backend owned by the tool.
It is best suited for Hugging Face users who want to discover similar people, compare tastes, or find peers in the ML community. The website also notes important limits: likes are only partial signals, only the first page of likers is sampled, datasets and Spaces are excluded, and the recommender stops if your likes do not intersect the vocabulary.
Foryu pricing
Pricing model: Freemium
The website presents Foryu as a weekend project and does not list any paid plans, subscriptions, or usage fees. The core experience is browser-based and client-side, and the site does not describe a free tier because the tool appears to be offered as a free web demo rather than a tiered product. No pricing table, premium plan, or included paid features are shown on the website.
Foryu pros
- Runs entirely in the browser
- No backend required
- Uses public Hugging Face data
- Finds similar users by taste
- Two-hop graph expansion
- Cosine-similarity ranking
- Uses IDF weighting
- Focuses on distinctive likes
- Works from a simple handle input
- Fast enough for interactive use
- Only ships a small vocab file
- Avoids server-side account storage
- Can surface niche peer matches
- Ranks by shared model attention
- Explains its matching method clearly
Foryu cons
- Requires a Hugging Face handle
- Depends on public HF API availability
- Around 70 API calls per lookup
- Can take 5–15 seconds
- Samples only the first page of likers
- Biases toward recent fans of popular models
- Excludes datasets and Spaces
- Fails when likes do not intersect vocab
- Likes are imperfect preference signals
- No backend means limited server-side features
- Only covers models in the shipped vocabulary
- May miss less visible users
- Not a general-purpose recommendation engine
- Only works for Hugging Face social graphs
- Results reflect similarity, not endorsement
Frequently asked questions about Foryu
What does Foryu do?
Foryu finds your closest “ML twins” on Hugging Face by comparing your liked models with other users’ likes. You enter a Hugging Face handle, and the tool ranks other users by how similar their taste is to yours.
How does Foryu calculate similarity?
It fetches a user’s public likes from Hugging Face, keeps the most distinctive ones, expands the graph two hops out through likers and their likes, and then ranks candidates with cosine similarity against your taste vector. The site says IDF weighting is used to help emphasize distinctive preferences.
Does Foryu run on a server?
No. The site says the ranking, candidate search, and input handling all run in your browser, and that the only shipped data file is vocab.json. The tool does not rely on its own backend for the recommendation logic.
How long does a search take?
The website says results typically take about 10 seconds, with a broader range of roughly 5 to 15 seconds. That timing comes from the live public API lookups used during matching.
What data does Foryu use?
Foryu uses public Hugging Face likes and a bundled vocabulary of the 10k most-liked Hugging Face models, plus their like counts and IDF scores. It excludes datasets and Spaces from the matching process.
Why might Foryu not find any matches?
The site says the recommender bails out if your likes do not intersect the vocabulary. That can happen if your likes are too unusual, outside the supported model set, or not visible in the data it samples.
Does Foryu treat likes as endorsements?
No. The site explicitly warns that Hugging Face likes are bookmarks as much as endorsements, so the similarity score reflects shared attention more than shared opinion.
Why can the results be biased?
Foryu only samples the first page of likers for each model, which can overrepresent recent fans of popular models. The site says IDF weighting partially corrects that bias, but it does not remove it entirely.
What happens if I use a new or obscure handle?
If the handle’s likes are not well represented in the vocabulary or do not connect strongly through the sampled graph, the results may be sparse or the tool may return no recommendation. The matching is strongest when your likes overlap with the bundled model set.
Who is Foryu for?
It is aimed at Hugging Face users who want to discover similar people, compare model tastes, or explore the ML community through shared likes. The site frames it as a lightweight browser tool for taste-based user discovery.